Compare commits

...
Author SHA1 Message Date
claudeandClaude Opus 5.5 f881e2cd54 DUEL_DESIGN: list secondPlayerCitations among house-rule keys
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-18 00:03:44 -05:00
claudeandClaude Opus 5.5 5121fd0418 DUEL_DESIGN v1.3: balance v1 = secondPlayerCitations 1 + turnCap 40; Location rider rejected (#64)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-18 00:03:38 -05:00
claudeandClaude Opus 5.5 6e97e2c8ee DUEL_DESIGN v1.2: secondPlayerCitations 1 (balance v1, #64), targets, Location finding
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 23:43:18 -05:00
claudeandClaude Opus 5.5 6182dff1cd DUEL_DESIGN v1.1: night-review nits (public Deleted, Garrison vs Featured, clock, rivalry wording), align with duel-cards #70
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 21:54:35 -05:00
claudeandClaude Opus 5.5 bac186470d DUEL_DESIGN.md v1: rules, keywords, templates, Connected, modes, tone, worked cards
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 21:35:07 -05:00
claudeandClaude Opus 5.5 2af2e7f06f Night 2 handbook: night-lead runs on tower-plus
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 13:08:17 -05:00
claudeandClaude Opus 5.5 57dc1e60aa Night 2 handbook: ui-pvp runs on inspiron-3670
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 13:07:59 -05:00
claudeandClaude Opus 5.5 dbf385d217 Night 2: data analyst scripts behind reviews/data.md
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 13:07:31 -05:00
claudeandClaude Opus 5.5 e34b09d06d Night 2: plan, design brief, handbook and plan reviews
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 13:05:42 -05:00
claudeanderic 4f7a8a7c74 Night 1: Edit War card game, eom.dev accounts, peer-hosted modes, Android and iOS apps (#22)
Overnight build by 17 Claude agents (2026-11-16 22:45 → 01:30 EST), coordinated in [claude/wikideck milestone 63](https://gitea.eom.dev/claude/wikideck/milestone/63). **For Eric's review. Do not merge without it.** The full write-up is in `docs/night/REPORT.md`.

**What's in it** (38 reviewed PRs, squash-merged into `night/2026-11-16`):
- **Edit War:**
  - a 3-lane, 6-turn card game built on what WikiDeck cards actually hold
  - a pure, deterministic, off-chain rules engine (`shared/game/`, with the API in `docs/game/ENGINE_API.md`); rules are data and v1.1 is the default
  - a mobile-first Play UI with a lobby, deck builder, four loaner starter decks, a match against the AI, results and a tutorial
  - a landing page (fixes #3)
- **Design and balance:** `docs/game/DESIGN.md`, `CARD_DATA.md` (236 sampled articles) and `BALANCE.md` (bot simulations). Starter pairings went from 23–77% on v1 to roughly even on v1.1.
- **eom.dev accounts:** Dex OIDC with PKCE, plus off-chain wallet linking, where the client builds the message it signs (`server/account/`, `src/auth/`).
- **Peer-hosted game modes:** modes are JSON data, never code. Node API at `/api/game/*`, with host credit and leaderboards. Three example modes (`shared/game/modes.ts`, `server/game/`, `docs/game/HOSTING.md`).
- **Mobile:** Capacitor 7 Android app (a debug APK played a full match against the AI in the emulator), and an iOS project waiting on Xcode (`docs/night/ANDROID.md`, `docs/night/IOS.md`).
- **Testnet tooling:** `deploy/night/` (LAN-only, 3 nodes) and an optional `WIKIDECK_USER_AGENT` (the default is unchanged).

**Not touched:** consensus and protocol (`server/chain.ts`, `shared/protocol.ts`, `server/testnet/{consensus,ledger,…}.ts`, `native/`), the live deployment and mainnet. Matches run off-chain and never move or burn NFTs.

**Checks:** the final integration gate passed in `node:24-alpine`: `npm test` 241/242 with 0 fail (1 skipped), and `npm run build` passes. The repo needs Node 24, because Node 20 fails 12 tests falsely.

**Decisions for Eric** (see REPORT.md, Known issues and Next steps):
1. whether the live node should set `WIKIDECK_USER_AGENT`
2. the Xcode licence, which you should accept yourself
3. #49 in the fork, a bot change held after the freeze
4. on-chain host rewards (proposal only, in HOSTING.md)

🤖 Generated with [Claude Code](https://claude.com/claude-code)Reviewed-on: WikiDeck/wikideck#22

Co-authored-by: Claude <claude@eom.dev>
2026-11-17 17:40:58 +00:00
eric 636f5b9c77 V2.1 (#21)
Working version Reviewed-on: WikiDeck/wikideck#21
2026-11-01 22:07:51 +00:00
eric 1edf2e6200 Merge pull request 'More cards' (#20) from v2 into main
Reviewed-on: WikiDeck/wikideck#20
2026-10-19 22:57:02 +00:00
283 changed files with 40528 additions and 553 deletions
+17
View File
@@ -1,3 +1,20 @@
node_modules
dist
.git
data
native/build
.env
.env.*
!.env.example
**/*.wallet.json
**/*backup*.json
**/*.sqlite
**/*.sqlite-*
**/*.pem
**/*.key
**/*.p12
**/*.pfx
*.log
.vscode
.idea
+25 -1
View File
@@ -12,8 +12,23 @@ WORKDIR /app
COPY package.json package-lock.json ./
RUN npm ci
COPY . .
ARG WIKIDECK_NETWORK=mainnet
ARG WIKIDECK_GENESIS_TIME=1793491200
ENV VITE_WIKIDECK_NETWORK=$WIKIDECK_NETWORK
ENV VITE_WIKIDECK_GENESIS_TIME=$WIKIDECK_GENESIS_TIME
RUN npm run build
FROM build AS test
RUN apk add --no-cache libstdc++
ENV WIKIDECK_NETWORK=testnet \
WIKIDECK_GENESIS_TIME=1792411200 \
VITE_WIKIDECK_NETWORK=testnet \
VITE_WIKIDECK_GENESIS_TIME=1792411200
RUN npm run build
COPY --from=randomx-build /native/build/wikideck-randomx /app/native/wikideck-randomx
ENV RANDOMX_BINARY=/app/native/wikideck-randomx
CMD ["npm", "test"]
FROM node:24-alpine AS production-deps
WORKDIR /app
COPY package.json package-lock.json ./
@@ -21,17 +36,25 @@ RUN npm ci --omit=dev --ignore-scripts && npm cache clean --force
FROM node:24-alpine AS runtime
RUN apk add --no-cache libstdc++
ARG WIKIDECK_NETWORK=mainnet
ARG WIKIDECK_GENESIS_TIME=1793491200
ENV NODE_ENV=production \
HOST=0.0.0.0 \
PORT=3001 \
WIKIDECK_DATA_DIR=/app/data \
RANDOMX_BINARY=/app/native/wikideck-randomx
RANDOMX_BINARY=/app/native/wikideck-randomx \
WIKIDECK_NETWORK=$WIKIDECK_NETWORK \
WIKIDECK_GENESIS_TIME=$WIKIDECK_GENESIS_TIME \
WIKIDECK_IMAGE_NETWORK=$WIKIDECK_NETWORK \
WIKIDECK_IMAGE_GENESIS_TIME=$WIKIDECK_GENESIS_TIME
WORKDIR /app
COPY --from=production-deps /app/node_modules ./node_modules
COPY package.json ./
COPY --from=build /app/dist ./dist
COPY server ./server
COPY shared ./shared
COPY scripts/node-backup.ts ./scripts/node-backup.ts
COPY scripts/ops-check.ts ./scripts/ops-check.ts
COPY --from=randomx-build /native/build/wikideck-randomx ./native/wikideck-randomx
RUN mkdir -p /app/data && chown node:node /app/data
USER node
@@ -40,3 +63,4 @@ HEALTHCHECK --interval=30s --timeout=5s --start-period=60s --retries=3 \
CMD ["node", "-e", "fetch('http://127.0.0.1:3001/healthz', {signal: AbortSignal.timeout(4000)}).then(r => process.exit(r.ok ? 0 : 1)).catch(() => process.exit(1))"]
# Node 24 strips the server's erasable TypeScript; no tsx/dev dependencies at runtime.
CMD ["node", "--experimental-strip-types", "server/index.ts"]
+4
View File
@@ -0,0 +1,4 @@
FROM node:24-alpine
WORKDIR /app
COPY package.json ./
RUN npm install --package-lock-only --ignore-scripts
+72 -14
View File
@@ -1,10 +1,53 @@
# Wikideck
Wikideck is an experimental, self-hosted RandomX testnet where each mined block creates a native NFT from a random English Wikipedia article. A page may be minted more than once; each NFT commits its own article and mint-time statistics snapshot, including a Wikimedia-hosted picture. It includes a React interface, a validating node, local P-256 wallets, transfers, and current off-chain Wikipedia statistics.
Wikideck is an experimental blockchain game about **mining, collecting, and trading illustrated Wikipedia cards**. Mine a block with the native RandomX CPU miner to earn a card, build a collection in a local wallet, and trade cards directly with another player.
**This is a local test network, not a public mainnet or an investment product.** No public bootnodes or hosted network are provided. Test NFTs have no promised value. Do not use funds or valuable assets. This project is not affiliated with Wikipedia or the Wikimedia Foundation.
**Mainnet-first, still NO-GO:** the default image, runtime profile, and Compose service are mainnet, but the network is experimental, unaudited, and not launched. A local build is not production approval. See the [card-only production plan](docs/card-only-production-plan.md) and [mainnet launch gates](docs/mainnet-launch.md).
## Run the three-node testnet (Docker)
Mining uses **RandomX, the proof-of-work algorithm also used by Monero**, designed to keep general-purpose consumer CPUs competitive. Wikideck is an independent blockchain, not part of Monero or its mining/security network. Hardware, total network hashpower and difficulty determine results; no particular mining reward rate or profitability is guaranteed.
**Mainnet is a gated candidate, not an operating public network or an investment product.** There are no hosted nodes, public bootnodes, or automatic peer discovery. No external audit is planned; the operator accepts that residual risk. Cards have no promised monetary value. Do not use funds or valuable assets. This project is not affiliated with Wikipedia or the Wikimedia Foundation.
## Mine, collect, and trade cards
- **Mine Wikipedia cards.** Each accepted block mints a native NFT for the miner from a random illustrated English Wikipedia article. The card records the article snapshot and mint-time statistics. The same page can appear on more than one distinct card. The reward is a card representing a page—not ownership of the page, its copyright, or a payment to Wikipedia.
- **Build collections.** Browse chain discoveries or focus on a subject, series, or theme. Cards are controlled by your encrypted local wallet.
- **Trade directly.** Look up another public wallet and exchange up to eight cards on each side in one signed, atomic trade. Offers are shared between the two players; the app does not list cards for sale or process payments.
- **Gameplay is not included yet.** The current release supports mining, collecting, transfers, and direct trades. Playable card games are future work.
**Licensing status:** this checkout does not yet contain a project LICENSE file. Source availability alone does not grant open-source modification/redistribution rights. Wikipedia text/images and dependencies retain their separate licenses.
## What's available now
| Available in the mainnet-first codebase | Not included |
| --- | --- |
| RandomX mining and illustrated Wikipedia card rewards | Playable card games |
| Encrypted local wallets, collections, transfers, and atomic card trades | Production network launch |
| Mainnet candidate by default; testnet only as an explicit local development fixture | Public bootstrap network or production readiness |
## Default: mainnet candidate (Docker)
The default mainnet genesis is fixed at **November 1, 2026, 00:00 UTC**
(`1793491200`). Rebuilding the current Dockerfile requires no special genesis
argument and no wait until January. Publish the rebuilt image as
`ericomeehan/wikideck:latest` before using the three-node Ansible playbook; it reads
the chain ID automatically and configures the risk acknowledgment and internal
peers. This updates source defaults, not an already published image or an existing
ledger's chain identity. Create/unlock a wallet in the deployed HTTPS UI and run
the native miner to produce cards and confirm transfers/trades.
Install Docker Desktop with Linux containers and Docker Compose v2. From the repository root:
```sh
docker compose up --build -d
docker compose ps
```
The default Compose service is mainnet-first and binds only to **http://127.0.0.1:3201** behind a separately configured HTTPS edge. It intentionally exits until the exact chain approval, explicit unaudited-risk acceptance, genesis time, HTTPS origin, and at least two valid public HTTPS peers are configured. Do not set these to placeholders just to make the process start. See [mainnet launch gates](docs/mainnet-launch.md). The node does not mine automatically. Stop it with `docker compose down`; the ledger volume remains. **Do not add `-v` unless you intend to delete that data.**
For isolated development, use the explicitly testnet-only [three-node fixture](#explicit-testnet-development-fixture-docker). `npm run dev` also pins both the API and Vite browser build to that testnet fixture; it does not silently change a production build.
## Explicit testnet development fixture (Docker)
Install Docker Desktop with Linux containers and Docker Compose v2. From the repository root, start three connected nodes:
@@ -22,9 +65,9 @@ Stop the nodes with:
docker compose -f compose.testnet.yaml down
```
The volumes persist across stops and rebuilds. **Do not add `-v` unless you intend to delete all three ledgers.** Back up wallets separately. Protocol v4 uses a new chain identity; earlier v2/v3 ledgers and encrypted backups are not compatible. To keep old Docker data intact while trying v4, use a separate Compose project name, for example `docker compose -p wikideck-v4 -f compose.testnet.yaml up --build -d`, and use the same `-p wikideck-v4` for subsequent Compose commands. The default `docker compose up --build -d` uses [compose.yaml](compose.yaml): it starts one isolated node on port 3001 with a different volume.
The target block interval is **15 minutes**. The volumes persist across stops and rebuilds. **Do not add `-v` unless you intend to delete all three ledgers.** Back up wallets separately. Protocol v6 adds atomic multi-card trades and a new chain identity; earlier v2–v5 ledgers and encrypted backups are not compatible. To keep old Docker data intact while trying v6, use a separate Compose project name, for example `docker compose -p wikideck-v6 -f compose.testnet.yaml up --build -d`, and use the same `-p wikideck-v6` for subsequent Compose commands. The default `docker compose up --build -d` uses [compose.yaml](compose.yaml): it starts one isolated node on port 3001 with a different volume.
For an existing host install, keep the old data directory intact and set `WIKIDECK_DATA_DIR` to a new empty directory before starting protocol v4. The node intentionally fails closed on an old ledger; it does not reset or migrate it.
For an existing host install, keep the old data directory intact and set `WIKIDECK_DATA_DIR` to a new empty directory before starting protocol v6. The node intentionally fails closed on an old-protocol ledger; it does not reset or migrate it. A v6 JSON ledger is imported into SQLite on first startup and left in place as a recovery copy. `npm start` defaults to the gated mainnet identity; use `npm run dev` only for local testnet development.
## Run on your host
@@ -47,21 +90,21 @@ The node listens at **http://127.0.0.1:3001** by default. For development, `npm
## Wallets and backups
The browser creates a local, recoverable P-256 wallet. Open the wallet panel and use **Wallet backup & recovery** to download an encrypted backup. Choose a unique password of at least 12 characters, keep the file outside the repository, and store the password separately. Test recovery in another browser profile before relying on the backup.
Open the wallet panel to explicitly create an encrypted local P-256 wallet, unlock an existing vault, or encrypt a legacy plaintext wallet without changing its identity. Nothing is silently generated or unlocked at startup. Choose a strong local password of at least 12 characters. **Wallet backup & recovery** downloads a separately password-encrypted portable backup; keep the file outside the repository and its password separately, and test recovery in another browser profile.
The password encrypts the downloaded backup; it does **not** encrypt the key stored in browser IndexedDB. Same-origin scripts and access to the browser profile may expose that key. Clearing site data without a usable backup loses access. Wallets are scoped to an origin (scheme, hostname, and port), so a backup must be restored at another origin to use the same identity there. A restored wallet controls the same address; it does not move assets between chains.
New/migrated local vaults encrypt the key in an IndexedDB database scoped to the exact chain ID; v2 encrypted backups authenticate that same chain ID. Original fixed-testnet v1 backups remain recoverable only on that original testnet. The unlocked key remains exportable in tab memory for signing and backup. Manual locking, five-minute inactivity, a four-hour absolute lifetime, reload, or hiding/leaving the tab ends the session. Same-origin compromise can still capture passwords or abuse an unlocked signer. Forgotten local passwords cannot be reset: recover from a usable backup. Legacy disk copies may retain old plaintext keys even after migration. Wallets are scoped to an origin (scheme, hostname, and port), so a backup must be restored at another origin to use the same identity there. Recovery does not move assets between chains. See [encrypted wallet lifecycle and remaining risks](docs/wallet-security.md).
The current testnet wallet is separate from any wallet used by an older demo. There is no automatic migration. See [consensus, wallets, and operational limits](docs/testnet-consensus.md) and the in-app backup warnings for more detail.
Each genesis has a separate browser wallet store, and backups cannot cross chain identities. Testnet remains only as an explicit development fixture; there is no automatic migration from older demos. See [consensus, wallets, and operational limits](docs/testnet-consensus.md) and the in-app backup warnings for more detail.
## Mine only by explicit choice
## Mine Wikipedia cards with RandomX
Mining uses the native CPU RandomX engine. The browser page never mines or launches a native program. First build the engine, download an encrypted wallet backup, then run the CLI from an interactive terminal and enter the password at its hidden prompt:
Mining uses the native CPU RandomX engine. First build the engine, download an encrypted wallet backup, then run the CLI from an interactive terminal and enter the password at its hidden prompt:
```sh
npm run miner -- --wallet "path/to/encrypted-backup.json" --consent --node http://127.0.0.1:3101 --blocks 3
npm run miner:testnet -- --wallet "path/to/encrypted-backup.json" --consent --node http://127.0.0.1:3101 --blocks 3
```
Use port 3001 for the default single host/Docker node, or 3101–3103 for the three-node Docker testnet. The command above uses node-a. The miner defaults to one worker, light mode, and one block; `--blocks 3` requests up to three accepted blocks. Press **Ctrl+C** to stop it. Do not put the backup password in a command, environment variable, or repository file. The backup and password are decrypted locally and are not sent to the node.
The example is intentionally testnet-only and uses node-a; the dedicated command pins that profile and its fixed genesis. For mainnet, use `npm run miner` with a chain-bound mainnet backup and the configured HTTPS node origin after satisfying the mainnet launch gates. The miner defaults to one worker, light mode, and one block; `--blocks 3` requests up to three accepted blocks. Press **Ctrl+C** to stop it. Do not put the backup password in a command, environment variable, or repository file. The backup and password are decrypted locally and are not sent to the node.
Each template request asks English Wikipedia for up to ten random pages and retries at most three times to find one with an image and readable extract. The same page may be selected again. The node also captures page size, language-link count, revision, pageviews, and derived scores for the mint-time snapshot. If Wikipedia cannot supply the article or its metadata, template creation returns an error; there is no fixed-article fallback. Validators check and commit the included snapshot without contacting Wikipedia during consensus validation.
@@ -72,13 +115,22 @@ Light mode needs a 256 MiB cache plus scratchpad and process overhead. `--fast`
- **Discover** shows preview examples until the node has mined blocks; previews are not owned NFTs. Minted cards can reference the same page, but each carries its own captured stats snapshot.
- **My collection** shows NFTs owned by the active local wallet on the connected node's canonical chain.
- **Block explorer** distinguishes block IDs, NFT token IDs, and RandomX proof hashes. A submitted transfer is pending until a block confirms it. Reorganizations can change canonical ownership and confirmations.
- **Trade cards** lets you look up another public wallet by receive address, select up to eight cards on each side, prepare a signed offer, and accept another player's offer. Offers are shared as JSON for the other player to review; they are not submitted until accepted. The two parties sign the same card list, and validators apply the whole bundle atomically or reject it entirely. Offers are not reservations: before acceptance the app re-checks ownership, maturity, sequences, and pending activity; stale offers must be recreated.
- Each minted card shows current Wikipedia statistics alongside its immutable at-mint snapshot: pageviews, article size, language links, revision, and derived scores. Current values may change; at-mint values are committed to the block. Missing pageview data stays unavailable, not zero.
- The appearance selector offers Catppuccin Mocha, Macchiato, Frappé, and Latte themes.
The browser interface trusts the connected node's reported state. It does not independently verify the chain or provide an SPV client. Transfers are irreversible once confirmed; check the recipient address carefully.
## Mainnet candidate · launch gated
**Production status: NO-GO pending network, capacity, deployment, and recovery evidence.** External audits are not planned; the operator accepts the residual risk, and the candidate must not be described as audited or certified. See the [security review](docs/security-review.md) and [operator launch checklist](docs/operator-launch.md). A read-only `launch:preflight` script reports blockers and local gate status without launching or granting approval.
The mainnet profile is the default and has a deployment-pinned network/genesis identity, separate ledger directory, chain-ID-scoped browser wallet, and chain-bound backups. The alternate [Compose definition](compose.mainnet.yaml) is also explicitly mainnet. Testnet is available only through the dedicated development/test harness. **No production network has been launched or certified by this work.** Startup requires explicit chain-specific approval, explicit unaudited-risk acceptance, the chosen genesis time in the built image, HTTPS origin, and multiple configured public peers. See [the mainnet launch gates and production checklist](docs/mainnet-launch.md).
## Tests and development checks
Offline preparation tools: `analyze:consensus` reports deterministic retarget/timestamp scenarios; `benchmark:randomx` runs a bounded native benchmark only with explicit `--consent`. Neither launches a node nor changes consensus or approval settings. See [calibration evidence and limitations](docs/consensus-calibration.md) before interpreting rates or illustrative targets.
```sh
npm test
npm run build
@@ -87,6 +139,8 @@ npm run test:native
`npm test` runs the application test suite, including fast test-only proof tests. `npm run test:native` runs the real upstream RandomX known-vector test and requires a native build configured with tests enabled; the documented `npm run native:configure` command enables them. The Docker build also builds RandomX and runs its native test.
The Dockerfile also provides a **test** target containing Node 24, development dependencies, the tests, and the compiled Linux-native RandomX binary. Build that target and run the resulting image to exercise the full native HTTP suite without relying on a host-platform binary. TypeScript compilation enforces erasable syntax because the production runtime strips types directly.
To explicitly test propagation and transfer behavior across the three local Docker nodes, first build the host native engine and start the three-node testnet, then run:
```sh
@@ -97,9 +151,13 @@ This mines three real blocks using disposable identities. It changes the testnet
## Network and safety limits
Nodes use only explicitly configured peers; there is no discovery service or public bootstrap network. The peer, storage, and HTTP configuration is documented in [docs/testnet-consensus.md](docs/testnet-consensus.md). The implementation is experimental: it has no fees or economic security, and remains vulnerable to resource exhaustion, isolation/eclipsing, censorship, majority-hashpower attacks, and reorganizations. Storage limits are stop limits rather than a scalable production design. HTTPS, container hardening, and passing tests do not constitute a security audit.
[Deployment and recovery preparation](docs/operations-recovery.md) includes explicit trusted-proxy configuration, opt-in loopback-only diagnostics and `ops:check`, plus stopped-node `node:backup` create/verify/new-directory restore tooling. New backups contain the ledger only; creation refuses to proceed if legacy financial journals or bindings remain, so archive those separately before retiring old deployments. Tools do not configure TLS/alerts or authorize production launch/restore.
Before considering any broader deployment, the protocol and cryptography would need independent audits, adversarial network testing, hardened abuse controls, and tested recovery procedures. Do not expose a public node or use valuable assets on this experimental network.
[Stage-1 node resource controls](docs/node-resources.md) bound global RPC admission, response bytes and canonical page sizes. The UI reuses paged history; legacy full-state responses stop beyond 128 blocks. A consent-gated loopback-only smoke-load script is available as `test:rpc-load`. This does not resolve long-chain/DAG storage limits or establish production capacity.
Nodes use only explicitly configured peers; there is no discovery service or public bootstrap network. The peer, storage, and HTTP configuration is documented in [docs/testnet-consensus.md](docs/testnet-consensus.md). The implementation is experimental: it has no established economic security and remains vulnerable to resource exhaustion, isolation/eclipsing, censorship, majority-hashpower attacks, and reorganizations. The ledger now uses transactional SQLite/WAL storage and sparse per-block NFT-state deltas, but nodes still retain DAG block records in memory, rebuild state roots over the full NFT set, and revalidate all proofs at startup. It is not yet an archival-scale production design. HTTPS, container hardening, and passing tests do not constitute a security audit.
Before considering any broader deployment, complete adversarial network testing, hardened abuse controls, and recovery procedures. External protocol/cryptography audits are not planned; this leaves accepted residual risk. Do not expose a public node or use valuable assets on this experimental network.
## Attribution
+101
View File
@@ -0,0 +1,101 @@
# Using Android gitignore template: https://github.com/github/gitignore/blob/HEAD/Android.gitignore
# Built application files
*.apk
*.aar
*.ap_
*.aab
# Files for the ART/Dalvik VM
*.dex
# Java class files
*.class
# Generated files
bin/
gen/
out/
# Uncomment the following line in case you need and you don't have the release build type files in your app
# release/
# Gradle files
.gradle/
build/
# Local configuration file (sdk path, etc)
local.properties
# Proguard folder generated by Eclipse
proguard/
# Log Files
*.log
# Android Studio Navigation editor temp files
.navigation/
# Android Studio captures folder
captures/
# IntelliJ
*.iml
.idea/workspace.xml
.idea/tasks.xml
.idea/gradle.xml
.idea/assetWizardSettings.xml
.idea/dictionaries
.idea/libraries
# Android Studio 3 in .gitignore file.
.idea/caches
.idea/modules.xml
# Comment next line if keeping position of elements in Navigation Editor is relevant for you
.idea/navEditor.xml
# Keystore files
# Uncomment the following lines if you do not want to check your keystore files in.
#*.jks
#*.keystore
# External native build folder generated in Android Studio 2.2 and later
.externalNativeBuild
.cxx/
# Google Services (e.g. APIs or Firebase)
# google-services.json
# Freeline
freeline.py
freeline/
freeline_project_description.json
# fastlane
fastlane/report.xml
fastlane/Preview.html
fastlane/screenshots
fastlane/test_output
fastlane/readme.md
# Version control
vcs.xml
# lint
lint/intermediates/
lint/generated/
lint/outputs/
lint/tmp/
# lint/reports/
# Android Profiling
*.hprof
# Cordova plugins for Capacitor
capacitor-cordova-android-plugins
# Copied web assets
app/src/main/assets/public
# Generated Config files
app/src/main/assets/capacitor.config.json
app/src/main/assets/capacitor.plugins.json
app/src/main/res/xml/config.xml
+2
View File
@@ -0,0 +1,2 @@
/build/*
!/build/.npmkeep
+54
View File
@@ -0,0 +1,54 @@
apply plugin: 'com.android.application'
android {
namespace "dev.eom.wikideck"
compileSdk rootProject.ext.compileSdkVersion
defaultConfig {
applicationId "dev.eom.wikideck"
minSdkVersion rootProject.ext.minSdkVersion
targetSdkVersion rootProject.ext.targetSdkVersion
versionCode 1
versionName "1.0"
testInstrumentationRunner "androidx.test.runner.AndroidJUnitRunner"
aaptOptions {
// Files and dirs to omit from the packaged assets dir, modified to accommodate modern web apps.
// Default: https://android.googlesource.com/platform/frameworks/base/+/282e181b58cf72b6ca770dc7ca5f91f135444502/tools/aapt/AaptAssets.cpp#61
ignoreAssetsPattern '!.svn:!.git:!.ds_store:!*.scc:.*:!CVS:!thumbs.db:!picasa.ini:!*~'
}
}
buildTypes {
release {
minifyEnabled false
proguardFiles getDefaultProguardFile('proguard-android.txt'), 'proguard-rules.pro'
}
}
}
repositories {
flatDir{
dirs '../capacitor-cordova-android-plugins/src/main/libs', 'libs'
}
}
dependencies {
implementation fileTree(include: ['*.jar'], dir: 'libs')
implementation "androidx.appcompat:appcompat:$androidxAppCompatVersion"
implementation "androidx.coordinatorlayout:coordinatorlayout:$androidxCoordinatorLayoutVersion"
implementation "androidx.core:core-splashscreen:$coreSplashScreenVersion"
implementation project(':capacitor-android')
testImplementation "junit:junit:$junitVersion"
androidTestImplementation "androidx.test.ext:junit:$androidxJunitVersion"
androidTestImplementation "androidx.test.espresso:espresso-core:$androidxEspressoCoreVersion"
implementation project(':capacitor-cordova-android-plugins')
}
apply from: 'capacitor.build.gradle'
try {
def servicesJSON = file('google-services.json')
if (servicesJSON.text) {
apply plugin: 'com.google.gms.google-services'
}
} catch(Exception e) {
logger.info("google-services.json not found, google-services plugin not applied. Push Notifications won't work")
}
+20
View File
@@ -0,0 +1,20 @@
// DO NOT EDIT THIS FILE! IT IS GENERATED EACH TIME "capacitor update" IS RUN
android {
compileOptions {
sourceCompatibility JavaVersion.VERSION_21
targetCompatibility JavaVersion.VERSION_21
}
}
apply from: "../capacitor-cordova-android-plugins/cordova.variables.gradle"
dependencies {
implementation project(':capacitor-app')
implementation project(':capacitor-browser')
}
if (hasProperty('postBuildExtras')) {
postBuildExtras()
}
+21
View File
@@ -0,0 +1,21 @@
# Add project specific ProGuard rules here.
# You can control the set of applied configuration files using the
# proguardFiles setting in build.gradle.
#
# For more details, see
# http://developer.android.com/guide/developing/tools/proguard.html
# If your project uses WebView with JS, uncomment the following
# and specify the fully qualified class name to the JavaScript interface
# class:
#-keepclassmembers class fqcn.of.javascript.interface.for.webview {
# public *;
#}
# Uncomment this to preserve the line number information for
# debugging stack traces.
#-keepattributes SourceFile,LineNumberTable
# If you keep the line number information, uncomment this to
# hide the original source file name.
#-renamesourcefileattribute SourceFile
@@ -0,0 +1,26 @@
package com.getcapacitor.myapp;
import static org.junit.Assert.*;
import android.content.Context;
import androidx.test.ext.junit.runners.AndroidJUnit4;
import androidx.test.platform.app.InstrumentationRegistry;
import org.junit.Test;
import org.junit.runner.RunWith;
/**
* Instrumented test, which will execute on an Android device.
*
* @see <a href="http://d.android.com/tools/testing">Testing documentation</a>
*/
@RunWith(AndroidJUnit4.class)
public class ExampleInstrumentedTest {
@Test
public void useAppContext() throws Exception {
// Context of the app under test.
Context appContext = InstrumentationRegistry.getInstrumentation().getTargetContext();
assertEquals("com.getcapacitor.app", appContext.getPackageName());
}
}
+51
View File
@@ -0,0 +1,51 @@
<?xml version="1.0" encoding="utf-8"?>
<manifest xmlns:android="http://schemas.android.com/apk/res/android">
<application
android:allowBackup="true"
android:icon="@mipmap/ic_launcher"
android:label="@string/app_name"
android:roundIcon="@mipmap/ic_launcher_round"
android:supportsRtl="true"
android:usesCleartextTraffic="true"
android:theme="@style/AppTheme">
<activity
android:windowSoftInputMode="adjustResize"
android:configChanges="orientation|keyboardHidden|keyboard|screenSize|locale|smallestScreenSize|screenLayout|uiMode|navigation"
android:name=".MainActivity"
android:label="@string/title_activity_main"
android:theme="@style/AppTheme.NoActionBarLaunch"
android:launchMode="singleTask"
android:exported="true">
<intent-filter>
<action android:name="android.intent.action.MAIN" />
<category android:name="android.intent.category.LAUNCHER" />
</intent-filter>
<!-- SSO redirect (RFC 8252 private-use scheme, registered with Dex as dev.eom.wikideck:/oauth/callback) -->
<intent-filter>
<action android:name="android.intent.action.VIEW" />
<category android:name="android.intent.category.DEFAULT" />
<category android:name="android.intent.category.BROWSABLE" />
<data android:scheme="dev.eom.wikideck" />
</intent-filter>
</activity>
<provider
android:name="androidx.core.content.FileProvider"
android:authorities="${applicationId}.fileprovider"
android:exported="false"
android:grantUriPermissions="true">
<meta-data
android:name="android.support.FILE_PROVIDER_PATHS"
android:resource="@xml/file_paths"></meta-data>
</provider>
</application>
<!-- Permissions -->
<uses-permission android:name="android.permission.INTERNET" />
</manifest>
@@ -0,0 +1,5 @@
package dev.eom.wikideck;
import com.getcapacitor.BridgeActivity;
public class MainActivity extends BridgeActivity {}
Binary file not shown.

After

Width:  |  Height:  |  Size: 7.5 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 3.9 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.0 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 14 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 17 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 7.7 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 4.0 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.6 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 13 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 17 KiB

@@ -0,0 +1,34 @@
<vector xmlns:android="http://schemas.android.com/apk/res/android"
xmlns:aapt="http://schemas.android.com/aapt"
android:width="108dp"
android:height="108dp"
android:viewportHeight="108"
android:viewportWidth="108">
<path
android:fillType="evenOdd"
android:pathData="M32,64C32,64 38.39,52.99 44.13,50.95C51.37,48.37 70.14,49.57 70.14,49.57L108.26,87.69L108,109.01L75.97,107.97L32,64Z"
android:strokeColor="#00000000"
android:strokeWidth="1">
<aapt:attr name="android:fillColor">
<gradient
android:endX="78.5885"
android:endY="90.9159"
android:startX="48.7653"
android:startY="61.0927"
android:type="linear">
<item
android:color="#44000000"
android:offset="0.0" />
<item
android:color="#00000000"
android:offset="1.0" />
</gradient>
</aapt:attr>
</path>
<path
android:fillColor="#FFFFFF"
android:fillType="nonZero"
android:pathData="M66.94,46.02L66.94,46.02C72.44,50.07 76,56.61 76,64L32,64C32,56.61 35.56,50.11 40.98,46.06L36.18,41.19C35.45,40.45 35.45,39.3 36.18,38.56C36.91,37.81 38.05,37.81 38.78,38.56L44.25,44.05C47.18,42.57 50.48,41.71 54,41.71C57.48,41.71 60.78,42.57 63.68,44.05L69.11,38.56C69.84,37.81 70.98,37.81 71.71,38.56C72.44,39.3 72.44,40.45 71.71,41.19L66.94,46.02ZM62.94,56.92C64.08,56.92 65,56.01 65,54.88C65,53.76 64.08,52.85 62.94,52.85C61.8,52.85 60.88,53.76 60.88,54.88C60.88,56.01 61.8,56.92 62.94,56.92ZM45.06,56.92C46.2,56.92 47.13,56.01 47.13,54.88C47.13,53.76 46.2,52.85 45.06,52.85C43.92,52.85 43,53.76 43,54.88C43,56.01 43.92,56.92 45.06,56.92Z"
android:strokeColor="#00000000"
android:strokeWidth="1" />
</vector>
@@ -0,0 +1,170 @@
<?xml version="1.0" encoding="utf-8"?>
<vector xmlns:android="http://schemas.android.com/apk/res/android"
android:width="108dp"
android:height="108dp"
android:viewportHeight="108"
android:viewportWidth="108">
<path
android:fillColor="#26A69A"
android:pathData="M0,0h108v108h-108z" />
<path
android:fillColor="#00000000"
android:pathData="M9,0L9,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M19,0L19,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M29,0L29,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M39,0L39,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M49,0L49,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M59,0L59,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M69,0L69,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M79,0L79,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M89,0L89,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M99,0L99,108"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,9L108,9"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,19L108,19"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,29L108,29"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,39L108,39"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,49L108,49"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,59L108,59"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,69L108,69"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,79L108,79"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,89L108,89"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M0,99L108,99"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M19,29L89,29"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M19,39L89,39"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M19,49L89,49"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M19,59L89,59"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M19,69L89,69"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M19,79L89,79"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M29,19L29,89"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M39,19L39,89"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M49,19L49,89"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M59,19L59,89"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M69,19L69,89"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
<path
android:fillColor="#00000000"
android:pathData="M79,19L79,89"
android:strokeColor="#33FFFFFF"
android:strokeWidth="0.8" />
</vector>
Binary file not shown.

After

Width:  |  Height:  |  Size: 3.9 KiB

@@ -0,0 +1,12 @@
<?xml version="1.0" encoding="utf-8"?>
<androidx.coordinatorlayout.widget.CoordinatorLayout xmlns:android="http://schemas.android.com/apk/res/android"
xmlns:app="http://schemas.android.com/apk/res-auto"
xmlns:tools="http://schemas.android.com/tools"
android:layout_width="match_parent"
android:layout_height="match_parent"
tools:context=".MainActivity">
<WebView
android:layout_width="match_parent"
android:layout_height="match_parent" />
</androidx.coordinatorlayout.widget.CoordinatorLayout>
@@ -0,0 +1,5 @@
<?xml version="1.0" encoding="utf-8"?>
<adaptive-icon xmlns:android="http://schemas.android.com/apk/res/android">
<background android:drawable="@color/ic_launcher_background"/>
<foreground android:drawable="@mipmap/ic_launcher_foreground"/>
</adaptive-icon>
@@ -0,0 +1,5 @@
<?xml version="1.0" encoding="utf-8"?>
<adaptive-icon xmlns:android="http://schemas.android.com/apk/res/android">
<background android:drawable="@color/ic_launcher_background"/>
<foreground android:drawable="@mipmap/ic_launcher_foreground"/>
</adaptive-icon>
Binary file not shown.

After

Width:  |  Height:  |  Size: 2.7 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 3.4 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 4.2 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 1.8 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.1 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.7 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 3.9 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 4.9 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 6.4 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 6.5 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.6 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 10 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.2 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 15 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 16 KiB

@@ -0,0 +1,8 @@
<?xml version="1.0" encoding="utf-8"?>
<resources>
<color name="colorPrimary">#1e1e2e</color>
<color name="colorPrimaryDark">#11111b</color>
<color name="colorAccent">#cba6f7</color>
<!-- Behind the status and navigation bars: matches the dark app header. -->
<color name="wikideckChrome">#11111b</color>
</resources>
@@ -0,0 +1,4 @@
<?xml version="1.0" encoding="utf-8"?>
<resources>
<color name="ic_launcher_background">#FFFFFF</color>
</resources>
@@ -0,0 +1,7 @@
<?xml version='1.0' encoding='utf-8'?>
<resources>
<string name="app_name">WikiDeck</string>
<string name="title_activity_main">WikiDeck</string>
<string name="package_name">dev.eom.wikideck</string>
<string name="custom_url_scheme">dev.eom.wikideck</string>
</resources>
@@ -0,0 +1,25 @@
<?xml version="1.0" encoding="utf-8"?>
<resources>
<!-- Base application theme. -->
<style name="AppTheme" parent="Theme.AppCompat.Light.DarkActionBar">
<!-- Customize your theme here. -->
<item name="colorPrimary">@color/colorPrimary</item>
<item name="colorPrimaryDark">@color/colorPrimaryDark</item>
<item name="colorAccent">@color/colorAccent</item>
</style>
<style name="AppTheme.NoActionBar" parent="Theme.AppCompat.DayNight.NoActionBar">
<item name="windowActionBar">false</item>
<item name="windowNoTitle">true</item>
<item name="android:background">@null</item>
<!-- Dark behind the status/navigation bars (edge-to-edge margins) to match the app chrome. -->
<item name="android:windowBackground">@color/wikideckChrome</item>
<item name="android:windowLightStatusBar">false</item>
</style>
<style name="AppTheme.NoActionBarLaunch" parent="Theme.SplashScreen">
<item name="android:background">@drawable/splash</item>
</style>
</resources>
@@ -0,0 +1,5 @@
<?xml version="1.0" encoding="utf-8"?>
<paths xmlns:android="http://schemas.android.com/apk/res/android">
<external-path name="my_images" path="." />
<cache-path name="my_cache_images" path="." />
</paths>
@@ -0,0 +1,18 @@
package com.getcapacitor.myapp;
import static org.junit.Assert.*;
import org.junit.Test;
/**
* Example local unit test, which will execute on the development machine (host).
*
* @see <a href="http://d.android.com/tools/testing">Testing documentation</a>
*/
public class ExampleUnitTest {
@Test
public void addition_isCorrect() throws Exception {
assertEquals(4, 2 + 2);
}
}
+12
View File
@@ -0,0 +1,12 @@
#!/bin/sh
# Build the WikiDeck debug APK against the private Night-1 testnet.
# Needs ANDROID_HOME and a JDK 21 (JAVA_HOME). Output: android/app/build/outputs/apk/debug/app-debug.apk
set -eu
cd "$(dirname "$0")/.."
export VITE_WIKIDECK_NETWORK="${VITE_WIKIDECK_NETWORK:-testnet}"
export VITE_WIKIDECK_GENESIS_TIME="${VITE_WIKIDECK_GENESIS_TIME:-1792411200}"
export VITE_NODE_URL="${VITE_NODE_URL:-http://192.168.1.32:3101}"
npm run build
npx cap sync android
cd android && ./gradlew --no-daemon -q assembleDebug
ls -l app/build/outputs/apk/debug/app-debug.apk
+29
View File
@@ -0,0 +1,29 @@
// Top-level build file where you can add configuration options common to all sub-projects/modules.
buildscript {
repositories {
google()
mavenCentral()
}
dependencies {
classpath 'com.android.tools.build:gradle:8.7.2'
classpath 'com.google.gms:google-services:4.4.2'
// NOTE: Do not place your application dependencies here; they belong
// in the individual module build.gradle files
}
}
apply from: "variables.gradle"
allprojects {
repositories {
google()
mavenCentral()
}
}
task clean(type: Delete) {
delete rootProject.buildDir
}
+9
View File
@@ -0,0 +1,9 @@
// DO NOT EDIT THIS FILE! IT IS GENERATED EACH TIME "capacitor update" IS RUN
include ':capacitor-android'
project(':capacitor-android').projectDir = new File('../node_modules/@capacitor/android/capacitor')
include ':capacitor-app'
project(':capacitor-app').projectDir = new File('../node_modules/@capacitor/app/android')
include ':capacitor-browser'
project(':capacitor-browser').projectDir = new File('../node_modules/@capacitor/browser/android')
+22
View File
@@ -0,0 +1,22 @@
# Project-wide Gradle settings.
# IDE (e.g. Android Studio) users:
# Gradle settings configured through the IDE *will override*
# any settings specified in this file.
# For more details on how to configure your build environment visit
# http://www.gradle.org/docs/current/userguide/build_environment.html
# Specifies the JVM arguments used for the daemon process.
# The setting is particularly useful for tweaking memory settings.
org.gradle.jvmargs=-Xmx1536m
# When configured, Gradle will run in incubating parallel mode.
# This option should only be used with decoupled projects. More details, visit
# http://www.gradle.org/docs/current/userguide/multi_project_builds.html#sec:decoupled_projects
# org.gradle.parallel=true
# AndroidX package structure to make it clearer which packages are bundled with the
# Android operating system, and which are packaged with your app's APK
# https://developer.android.com/topic/libraries/support-library/androidx-rn
android.useAndroidX=true
Binary file not shown.
+7
View File
@@ -0,0 +1,7 @@
distributionBase=GRADLE_USER_HOME
distributionPath=wrapper/dists
distributionUrl=https\://services.gradle.org/distributions/gradle-8.11.1-all.zip
networkTimeout=10000
validateDistributionUrl=true
zipStoreBase=GRADLE_USER_HOME
zipStorePath=wrapper/dists
Vendored Executable
+252
View File
@@ -0,0 +1,252 @@
#!/bin/sh
#
# Copyright © 2015-2021 the original authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# SPDX-License-Identifier: Apache-2.0
#
##############################################################################
#
# Gradle start up script for POSIX generated by Gradle.
#
# Important for running:
#
# (1) You need a POSIX-compliant shell to run this script. If your /bin/sh is
# noncompliant, but you have some other compliant shell such as ksh or
# bash, then to run this script, type that shell name before the whole
# command line, like:
#
# ksh Gradle
#
# Busybox and similar reduced shells will NOT work, because this script
# requires all of these POSIX shell features:
# * functions;
# * expansions «$var», «${var}», «${var:-default}», «${var+SET}»,
# «${var#prefix}», «${var%suffix}», and «$( cmd )»;
# * compound commands having a testable exit status, especially «case»;
# * various built-in commands including «command», «set», and «ulimit».
#
# Important for patching:
#
# (2) This script targets any POSIX shell, so it avoids extensions provided
# by Bash, Ksh, etc; in particular arrays are avoided.
#
# The "traditional" practice of packing multiple parameters into a
# space-separated string is a well documented source of bugs and security
# problems, so this is (mostly) avoided, by progressively accumulating
# options in "$@", and eventually passing that to Java.
#
# Where the inherited environment variables (DEFAULT_JVM_OPTS, JAVA_OPTS,
# and GRADLE_OPTS) rely on word-splitting, this is performed explicitly;
# see the in-line comments for details.
#
# There are tweaks for specific operating systems such as AIX, CygWin,
# Darwin, MinGW, and NonStop.
#
# (3) This script is generated from the Groovy template
# https://github.com/gradle/gradle/blob/HEAD/platforms/jvm/plugins-application/src/main/resources/org/gradle/api/internal/plugins/unixStartScript.txt
# within the Gradle project.
#
# You can find Gradle at https://github.com/gradle/gradle/.
#
##############################################################################
# Attempt to set APP_HOME
# Resolve links: $0 may be a link
app_path=$0
# Need this for daisy-chained symlinks.
while
APP_HOME=${app_path%"${app_path##*/}"} # leaves a trailing /; empty if no leading path
[ -h "$app_path" ]
do
ls=$( ls -ld "$app_path" )
link=${ls#*' -> '}
case $link in #(
/*) app_path=$link ;; #(
*) app_path=$APP_HOME$link ;;
esac
done
# This is normally unused
# shellcheck disable=SC2034
APP_BASE_NAME=${0##*/}
# Discard cd standard output in case $CDPATH is set (https://github.com/gradle/gradle/issues/25036)
APP_HOME=$( cd -P "${APP_HOME:-./}" > /dev/null && printf '%s
' "$PWD" ) || exit
# Use the maximum available, or set MAX_FD != -1 to use that value.
MAX_FD=maximum
warn () {
echo "$*"
} >&2
die () {
echo
echo "$*"
echo
exit 1
} >&2
# OS specific support (must be 'true' or 'false').
cygwin=false
msys=false
darwin=false
nonstop=false
case "$( uname )" in #(
CYGWIN* ) cygwin=true ;; #(
Darwin* ) darwin=true ;; #(
MSYS* | MINGW* ) msys=true ;; #(
NONSTOP* ) nonstop=true ;;
esac
CLASSPATH=$APP_HOME/gradle/wrapper/gradle-wrapper.jar
# Determine the Java command to use to start the JVM.
if [ -n "$JAVA_HOME" ] ; then
if [ -x "$JAVA_HOME/jre/sh/java" ] ; then
# IBM's JDK on AIX uses strange locations for the executables
JAVACMD=$JAVA_HOME/jre/sh/java
else
JAVACMD=$JAVA_HOME/bin/java
fi
if [ ! -x "$JAVACMD" ] ; then
die "ERROR: JAVA_HOME is set to an invalid directory: $JAVA_HOME
Please set the JAVA_HOME variable in your environment to match the
location of your Java installation."
fi
else
JAVACMD=java
if ! command -v java >/dev/null 2>&1
then
die "ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.
Please set the JAVA_HOME variable in your environment to match the
location of your Java installation."
fi
fi
# Increase the maximum file descriptors if we can.
if ! "$cygwin" && ! "$darwin" && ! "$nonstop" ; then
case $MAX_FD in #(
max*)
# In POSIX sh, ulimit -H is undefined. That's why the result is checked to see if it worked.
# shellcheck disable=SC2039,SC3045
MAX_FD=$( ulimit -H -n ) ||
warn "Could not query maximum file descriptor limit"
esac
case $MAX_FD in #(
'' | soft) :;; #(
*)
# In POSIX sh, ulimit -n is undefined. That's why the result is checked to see if it worked.
# shellcheck disable=SC2039,SC3045
ulimit -n "$MAX_FD" ||
warn "Could not set maximum file descriptor limit to $MAX_FD"
esac
fi
# Collect all arguments for the java command, stacking in reverse order:
# * args from the command line
# * the main class name
# * -classpath
# * -D...appname settings
# * --module-path (only if needed)
# * DEFAULT_JVM_OPTS, JAVA_OPTS, and GRADLE_OPTS environment variables.
# For Cygwin or MSYS, switch paths to Windows format before running java
if "$cygwin" || "$msys" ; then
APP_HOME=$( cygpath --path --mixed "$APP_HOME" )
CLASSPATH=$( cygpath --path --mixed "$CLASSPATH" )
JAVACMD=$( cygpath --unix "$JAVACMD" )
# Now convert the arguments - kludge to limit ourselves to /bin/sh
for arg do
if
case $arg in #(
-*) false ;; # don't mess with options #(
/?*) t=${arg#/} t=/${t%%/*} # looks like a POSIX filepath
[ -e "$t" ] ;; #(
*) false ;;
esac
then
arg=$( cygpath --path --ignore --mixed "$arg" )
fi
# Roll the args list around exactly as many times as the number of
# args, so each arg winds up back in the position where it started, but
# possibly modified.
#
# NB: a `for` loop captures its iteration list before it begins, so
# changing the positional parameters here affects neither the number of
# iterations, nor the values presented in `arg`.
shift # remove old arg
set -- "$@" "$arg" # push replacement arg
done
fi
# Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
DEFAULT_JVM_OPTS='"-Xmx64m" "-Xms64m"'
# Collect all arguments for the java command:
# * DEFAULT_JVM_OPTS, JAVA_OPTS, JAVA_OPTS, and optsEnvironmentVar are not allowed to contain shell fragments,
# and any embedded shellness will be escaped.
# * For example: A user cannot expect ${Hostname} to be expanded, as it is an environment variable and will be
# treated as '${Hostname}' itself on the command line.
set -- \
"-Dorg.gradle.appname=$APP_BASE_NAME" \
-classpath "$CLASSPATH" \
org.gradle.wrapper.GradleWrapperMain \
"$@"
# Stop when "xargs" is not available.
if ! command -v xargs >/dev/null 2>&1
then
die "xargs is not available"
fi
# Use "xargs" to parse quoted args.
#
# With -n1 it outputs one arg per line, with the quotes and backslashes removed.
#
# In Bash we could simply go:
#
# readarray ARGS < <( xargs -n1 <<<"$var" ) &&
# set -- "${ARGS[@]}" "$@"
#
# but POSIX shell has neither arrays nor command substitution, so instead we
# post-process each arg (as a line of input to sed) to backslash-escape any
# character that might be a shell metacharacter, then use eval to reverse
# that process (while maintaining the separation between arguments), and wrap
# the whole thing up as a single "set" statement.
#
# This will of course break if any of these variables contains a newline or
# an unmatched quote.
#
eval "set -- $(
printf '%s\n' "$DEFAULT_JVM_OPTS $JAVA_OPTS $GRADLE_OPTS" |
xargs -n1 |
sed ' s~[^-[:alnum:]+,./:=@_]~\\&~g; ' |
tr '\n' ' '
)" '"$@"'
exec "$JAVACMD" "$@"
+94
View File
@@ -0,0 +1,94 @@
@rem
@rem Copyright 2015 the original author or authors.
@rem
@rem Licensed under the Apache License, Version 2.0 (the "License");
@rem you may not use this file except in compliance with the License.
@rem You may obtain a copy of the License at
@rem
@rem https://www.apache.org/licenses/LICENSE-2.0
@rem
@rem Unless required by applicable law or agreed to in writing, software
@rem distributed under the License is distributed on an "AS IS" BASIS,
@rem WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
@rem See the License for the specific language governing permissions and
@rem limitations under the License.
@rem
@rem SPDX-License-Identifier: Apache-2.0
@rem
@if "%DEBUG%"=="" @echo off
@rem ##########################################################################
@rem
@rem Gradle startup script for Windows
@rem
@rem ##########################################################################
@rem Set local scope for the variables with windows NT shell
if "%OS%"=="Windows_NT" setlocal
set DIRNAME=%~dp0
if "%DIRNAME%"=="" set DIRNAME=.
@rem This is normally unused
set APP_BASE_NAME=%~n0
set APP_HOME=%DIRNAME%
@rem Resolve any "." and ".." in APP_HOME to make it shorter.
for %%i in ("%APP_HOME%") do set APP_HOME=%%~fi
@rem Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
set DEFAULT_JVM_OPTS="-Xmx64m" "-Xms64m"
@rem Find java.exe
if defined JAVA_HOME goto findJavaFromJavaHome
set JAVA_EXE=java.exe
%JAVA_EXE% -version >NUL 2>&1
if %ERRORLEVEL% equ 0 goto execute
echo. 1>&2
echo ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH. 1>&2
echo. 1>&2
echo Please set the JAVA_HOME variable in your environment to match the 1>&2
echo location of your Java installation. 1>&2
goto fail
:findJavaFromJavaHome
set JAVA_HOME=%JAVA_HOME:"=%
set JAVA_EXE=%JAVA_HOME%/bin/java.exe
if exist "%JAVA_EXE%" goto execute
echo. 1>&2
echo ERROR: JAVA_HOME is set to an invalid directory: %JAVA_HOME% 1>&2
echo. 1>&2
echo Please set the JAVA_HOME variable in your environment to match the 1>&2
echo location of your Java installation. 1>&2
goto fail
:execute
@rem Setup the command line
set CLASSPATH=%APP_HOME%\gradle\wrapper\gradle-wrapper.jar
@rem Execute Gradle
"%JAVA_EXE%" %DEFAULT_JVM_OPTS% %JAVA_OPTS% %GRADLE_OPTS% "-Dorg.gradle.appname=%APP_BASE_NAME%" -classpath "%CLASSPATH%" org.gradle.wrapper.GradleWrapperMain %*
:end
@rem End local scope for the variables with windows NT shell
if %ERRORLEVEL% equ 0 goto mainEnd
:fail
rem Set variable GRADLE_EXIT_CONSOLE if you need the _script_ return code instead of
rem the _cmd.exe /c_ return code!
set EXIT_CODE=%ERRORLEVEL%
if %EXIT_CODE% equ 0 set EXIT_CODE=1
if not ""=="%GRADLE_EXIT_CONSOLE%" exit %EXIT_CODE%
exit /b %EXIT_CODE%
:mainEnd
if "%OS%"=="Windows_NT" endlocal
:omega
+5
View File
@@ -0,0 +1,5 @@
include ':app'
include ':capacitor-cordova-android-plugins'
project(':capacitor-cordova-android-plugins').projectDir = new File('./capacitor-cordova-android-plugins/')
apply from: 'capacitor.settings.gradle'
+16
View File
@@ -0,0 +1,16 @@
ext {
minSdkVersion = 23
compileSdkVersion = 35
targetSdkVersion = 35
androidxActivityVersion = '1.9.2'
androidxAppCompatVersion = '1.7.0'
androidxCoordinatorLayoutVersion = '1.2.0'
androidxCoreVersion = '1.15.0'
androidxFragmentVersion = '1.8.4'
coreSplashScreenVersion = '1.0.1'
androidxWebkitVersion = '1.12.1'
junitVersion = '4.13.2'
androidxJunitVersion = '1.2.1'
androidxEspressoCoreVersion = '3.6.1'
cordovaAndroidVersion = '10.1.1'
}
+26
View File
@@ -0,0 +1,26 @@
import type { CapacitorConfig } from '@capacitor/cli';
// WikiDeck mobile shell. Build the web UI with the node URL baked in, then sync:
// VITE_WIKIDECK_NETWORK=testnet VITE_WIKIDECK_GENESIS_TIME=1792411200 \
// VITE_NODE_URL=http://192.168.1.32:3101 npm run build && npx cap sync android
// (see android/build-debug.sh)
const config: CapacitorConfig = {
appId: 'dev.eom.wikideck',
appName: 'WikiDeck',
webDir: 'dist',
server: {
androidScheme: 'https',
// The private testnet node is plain HTTP on the LAN.
cleartext: true,
},
android: {
// Android 15 (targetSdk 35) draws edge-to-edge; keep the WebView clear of the system bars.
adjustMarginsForEdgeToEdge: 'auto',
},
plugins: {
// Route fetch through native HTTP: avoids CORS and the node's browser Origin check.
CapacitorHttp: { enabled: true },
},
};
export default config;
+39
View File
@@ -0,0 +1,39 @@
# Gated candidate profile, NOT an announcement of a production-ready or launched network.
# Review docs/mainnet-launch.md before attempting activation.
services:
mainnet:
build:
context: .
target: runtime
args:
WIKIDECK_NETWORK: mainnet
WIKIDECK_GENESIS_TIME: "${WIKIDECK_GENESIS_TIME:-1793491200}"
ports:
- "127.0.0.1:${WIKIDECK_MAINNET_PORT:-3201}:3001"
environment:
WIKIDECK_NETWORK: mainnet
WIKIDECK_GENESIS_TIME: "${WIKIDECK_GENESIS_TIME:-1793491200}"
PUBLIC_ORIGIN: "${WIKIDECK_MAINNET_ORIGIN:-}"
WIKIDECK_PEERS: "${WIKIDECK_MAINNET_PEERS:-}"
WIKIDECK_MAINNET_LAUNCH: "${WIKIDECK_MAINNET_LAUNCH:-}"
WIKIDECK_MAINNET_RISK_ACK: "${WIKIDECK_MAINNET_RISK_ACK:-}"
WIKIDECK_ALLOW_PRIVATE_PEERS: "0"
# Empty by default; set only the actual edge-to-container source IP/CIDR after topology review.
WIKIDECK_TRUSTED_PROXIES: "${WIKIDECK_TRUSTED_PROXIES:-}"
# Diagnostics are deliberately disabled here. Never publish an operations port.
volumes:
- mainnet-candidate-data:/app/data
init: true
restart: "no"
read_only: true
tmpfs:
- /tmp:size=32m,mode=1777,noexec,nosuid
cap_drop: [ALL]
security_opt: [no-new-privileges:true]
pids_limit: 128
mem_limit: 1g
cpus: 1.0
stop_grace_period: 30s
volumes:
mainnet-candidate-data:
+17 -3
View File
@@ -1,10 +1,16 @@
# Three independent validators, separate volumes, no miner auto-starts.
x-node-environment: &node-environment
WIKIDECK_NETWORK: testnet
services:
node-a:
image: wikideck:testnet
build: { context: ., target: runtime }
build:
context: .
target: runtime
args: { WIKIDECK_NETWORK: testnet, WIKIDECK_GENESIS_TIME: "1792411200" }
ports: ["127.0.0.1:3101:3001"]
environment:
<<: *node-environment
PUBLIC_ORIGIN: http://127.0.0.1:3101
WIKIDECK_PEERS: http://node-b:3001,http://node-c:3001
WIKIDECK_ALLOW_PRIVATE_PEERS: "1"
@@ -20,9 +26,13 @@ services:
stop_grace_period: 30s
node-b:
image: wikideck:testnet
build: { context: ., target: runtime }
build:
context: .
target: runtime
args: { WIKIDECK_NETWORK: testnet, WIKIDECK_GENESIS_TIME: "1792411200" }
ports: ["127.0.0.1:3102:3001"]
environment:
<<: *node-environment
PUBLIC_ORIGIN: http://127.0.0.1:3102
WIKIDECK_PEERS: http://node-a:3001,http://node-c:3001
WIKIDECK_ALLOW_PRIVATE_PEERS: "1"
@@ -38,9 +48,13 @@ services:
stop_grace_period: 30s
node-c:
image: wikideck:testnet
build: { context: ., target: runtime }
build:
context: .
target: runtime
args: { WIKIDECK_NETWORK: testnet, WIKIDECK_GENESIS_TIME: "1792411200" }
ports: ["127.0.0.1:3103:3001"]
environment:
<<: *node-environment
PUBLIC_ORIGIN: http://127.0.0.1:3103
WIKIDECK_PEERS: http://node-a:3001,http://node-b:3001
WIKIDECK_ALLOW_PRIVATE_PEERS: "1"
+16 -8
View File
@@ -3,17 +3,25 @@ services:
build:
context: .
target: runtime
args:
WIKIDECK_NETWORK: mainnet
WIKIDECK_GENESIS_TIME: "${WIKIDECK_GENESIS_TIME:-1793491200}"
ports:
# Keep the prototype private; listen on all interfaces only inside the container.
- "127.0.0.1:${WIKIDECK_PORT:-3001}:3001"
# Loopback only; terminate reviewed HTTPS at the operator's edge.
- "127.0.0.1:${WIKIDECK_MAINNET_PORT:-3201}:3001"
environment:
PUBLIC_ORIGIN: "${PUBLIC_ORIGIN:-http://127.0.0.1:${WIKIDECK_PORT:-3001}}"
WIKIDECK_PEERS: "${WIKIDECK_PEERS:-}"
WIKIDECK_ALLOW_PRIVATE_PEERS: "${WIKIDECK_ALLOW_PRIVATE_PEERS:-0}"
WIKIDECK_NETWORK: mainnet
WIKIDECK_GENESIS_TIME: "${WIKIDECK_GENESIS_TIME:-1793491200}"
PUBLIC_ORIGIN: "${WIKIDECK_MAINNET_ORIGIN:-}"
WIKIDECK_PEERS: "${WIKIDECK_MAINNET_PEERS:-}"
WIKIDECK_MAINNET_LAUNCH: "${WIKIDECK_MAINNET_LAUNCH:-}"
WIKIDECK_MAINNET_RISK_ACK: "${WIKIDECK_MAINNET_RISK_ACK:-}"
WIKIDECK_ALLOW_PRIVATE_PEERS: "0"
WIKIDECK_TRUSTED_PROXIES: "${WIKIDECK_TRUSTED_PROXIES:-}"
volumes:
- wikideck-data:/app/data
- wikideck-mainnet-data:/app/data
init: true
restart: unless-stopped
restart: "no"
read_only: true
tmpfs:
- /tmp:size=32m,mode=1777,noexec,nosuid
@@ -27,4 +35,4 @@ services:
stop_grace_period: 30s
volumes:
wikideck-data:
wikideck-mainnet-data:
+45
View File
@@ -0,0 +1,45 @@
# Kubernetes mainnet preparation
This directory intentionally contains no apply-ready manifests yet. The code is mainnet-first and supports a genesis frozen into the browser bundle and runtime image, but the current launch gate requires deployment facts that were not supplied. Placeholder origins, approval IDs, TLS secrets, or storage settings would not produce a working or safe deployment.
## Required topology for this implementation
- Build and publish one immutable image with `WIKIDECK_NETWORK=mainnet` and one chosen `WIKIDECK_GENESIS_TIME`. Give every peer and browser bundle that same chain identity. A runtime-only genesis override is rejected by the image manifest check.
- Give each node a stable, unique public HTTPS origin. TLS must terminate at an operator-managed ingress/reverse proxy; the application listener itself is HTTP on port 3001.
- Configure at least two other distinct public HTTPS peer origins per node for public peering. Private IPs/DNS and HTTP are rejected by default. An operator-controlled private candidate cluster may explicitly set `WIKIDECK_ALLOW_PRIVATE_PEERS=1` and use distinct per-node ClusterIP DNS origins. This opt-in does not change the exact chain-ID/risk/genesis gates or the HTTPS browser-origin requirement.
- Route each public peer hostname deterministically to exactly one node. A shared load-balanced hostname is not a stable peer identity. Separate per-peer Deployments/Services/Ingress routes are the simplest current shape; each node must exclude its own origin from its peer list.
- Give every node an exclusive persistent volume mounted at `/app/data`. Never share a ledger volume between pods. Agree the storage class, capacity, backup/restore procedure, resource limits, failure domains, and upgrade/reorg policy before enabling the launch approval.
- Configure the exact manifest `CHAIN_ID` as `WIKIDECK_MAINNET_LAUNCH`, and explicitly set `WIKIDECK_MAINNET_RISK_ACK=operator-accepts-unaudited-risk`. The latter accepts unaudited residual risk; it is not an audit or readiness check.
## Pod hardening baseline
Use a dedicated namespace and least-privilege service account (or disable token automount if no Kubernetes API access is needed), non-root execution matching the image's `node` user, read-only root filesystem, dropped Linux capabilities, `allowPrivilegeEscalation: false`, `seccompProfile: RuntimeDefault`, and a bounded writable `/tmp` volume. Use resource requests and limits based on measured RandomX/replay/sync behavior, not guesses. Readiness/liveness should use `/healthz`; allow a startup grace period for ledger replay and native engine initialization. Restrict ingress to the reviewed edge and egress to DNS, the configured public peers, and required Wikimedia endpoints. Add NetworkPolicy, TLS, image provenance/signature enforcement, monitoring, and tested encrypted backups through the cluster's established platform controls.
Do not make the public origin or configured peers imply that peers are independent or that the network is secure. This software has no external audit, public bootstrap service, peer discovery, or demonstrated production capacity.
## Private candidate cluster
Deploy the public node and two internal peers as three separate single-replica
role instances, each with its own Service and PVC. Configure each peer list with
the other two Services, for example `http://wikideck-peer-a.wikideck.svc.cluster.local:80`.
Allow same-namespace validator pods on target TCP 3001 through both ingress and
egress NetworkPolicy; Service port 80 is translated to that target port. Only the
public node needs ingress; all nodes may use the public node's HTTPS browser
origin. Keep API credentials and wallet keys out of validator pods.
Read `--network-manifest` from the exact published image to obtain the candidate
chain ID without opening storage or starting a listener. Its
`privatePeersSupported: true` capability distinguishes updated images from older
images that reject this topology. Do not override genesis at runtime. The default
image genesis is now fixed at November 1, 2026, 00:00 UTC (`1793491200`), so no
January wait or custom build argument is needed. Rebuild and publish the updated
image; an older published image still contains its previous genesis. Existing
ledgers and wallets must retain their original chain identity.
Private peer RPC is unauthenticated and, with HTTP, unencrypted. Use a trusted
cluster boundary and scoped policies; add mesh transport protection if needed.
This topology does not demonstrate independent peers/miners or production readiness.
## Values needed before manifests can be finalized
Choose the genesis timestamp and image registry/digest; provide the namespace, ingress class and TLS/certificate mechanism, public DNS name per node, peer topology, storage class/size, resource budget, network-policy egress model, backup destination/retention, and desired availability/recovery behavior. Then calculate and publish the exact manifest and chain approval. Until those values and the peer topology are available, this checklist is the safe preparation artifact rather than a misleading runnable YAML template.
+41
View File
@@ -0,0 +1,41 @@
# Example for nginx on the SAME host as loopback-bound node :3201.
# Not installed by this repository. Replace host/cert paths and validate real topology.
# Node: WIKIDECK_TRUSTED_PROXIES=127.0.0.1,::1 (never use this for a container bridge without adjustment).
limit_req_zone $binary_remote_addr zone=wikideck_ip:10m rate=5r/s;
limit_conn_zone $binary_remote_addr zone=wikideck_conn:10m;
server {
listen 80;
server_name wikideck.example.invalid;
return 308 https://$host$request_uri;
}
server {
listen 443 ssl;
server_name wikideck.example.invalid;
ssl_certificate /etc/ssl/wikideck/fullchain.pem;
ssl_certificate_key /etc/ssl/wikideck/private.key;
ssl_protocols TLSv1.2 TLSv1.3;
server_tokens off;
client_max_body_size 68k;
client_body_timeout 15s;
client_header_timeout 10s;
keepalive_timeout 10s;
send_timeout 30s;
limit_conn wikideck_conn 16;
limit_req_status 429;
# Do not proxy private operations listener. Its port must never be publicly published.
location / {
limit_req zone=wikideck_ip burst=20 nodelay;
proxy_pass http://127.0.0.1:3201;
proxy_http_version 1.1;
proxy_set_header Host $host;
# Overwrite, do NOT append an untrusted client-supplied X-Forwarded-For chain.
proxy_set_header X-Forwarded-For $remote_addr;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_set_header Forwarded "";
proxy_set_header Connection "";
proxy_connect_timeout 5s;
proxy_send_timeout 15s;
proxy_read_timeout 30s;
}
}
+22
View File
@@ -0,0 +1,22 @@
# Night 1 private testnet (sim-testnet)
A 3-node **testnet-only** fixture on poweredge-r350. It is reachable on the LAN only, and uses the Docker project `wikideck-night`.
| node | URL |
|---|---|
| node-a | http://192.168.1.32:3101 |
| node-b | http://192.168.1.32:3102 |
| node-c | http://192.168.1.32:3103 |
- `compose.lan.yaml` overrides only the published ports (bound to `192.168.1.32`, never `0.0.0.0`) and `PUBLIC_ORIGIN`. It changes nothing in consensus or the image.
- `mine.ts` is a multi-worker native RandomX miner. It runs inside the `wikideck:testnet` image, which ships the native engine, so the host needs no cmake.
- It mines to 4 disposable wallets (alice, bob, carol, dave) and writes chain-bound encrypted browser backups to `$STATE_DIR/wallets/`.
- It never changes consensus code. Speed comes from the protocol's own testnet parameters: the initial target accepts 1/16 of hashes, and each 10-block retarget at most quadruples the difficulty.
- Each template costs several Wikipedia API calls, and Wikipedia rate-limits our egress IP. The miner therefore paces templates (`MIN_GAP`, default 45 s) and backs off exponentially on failure.
- `compose.lan.yaml` also sets, per node:
- `WIKIDECK_USER_AGENT`, a Wikimedia-policy UA with a project URL (#19). Without it, Wikipedia returned 429 to our egress and templates failed.
- the game-mode env from #22: A = classic and trending, B = draft, C = all three.
- `autodeploy.sh` follows the integration branch. Run it from a dedicated detached worktree: `git worktree add --detach ../deploy-src origin/night/2026-11-16 && cd ../deploy-src && deploy/night/autodeploy.sh`. It polls the Gitea API every 3 min and redeploys whenever the head moves.
- `testnet.sh` has `up`, `mine N [workers]`, `status` and `down`. Run it from the repo root on a checkout of the integration branch. `up` rebuilds the image from that checkout; the ledgers persist in the named volumes.
Never point this at mainnet or the live `wikideck` namespace.
+40
View File
@@ -0,0 +1,40 @@
#!/bin/sh
# sim-testnet: keep the wikideck-night testnet on the integration branch head.
# Run from the root of a *dedicated* checkout (e.g. a detached git worktree): it checks out
# each new head of $BRANCH and runs `deploy/night/testnet.sh rolling` (ledgers persist in the volumes):
# nodes restart one at a time, and deploys wait (up to FORCE_AFTER) while matches or lobbies are live,
# because lobbies are in memory and a restart ends games in progress.
# Head detection uses the public Gitea API, which also works where SSH auth isn't available.
set -u
BRANCH=${BRANCH:-night/2026-11-16}
API=${API:-https://gitea.eom.dev/api/v1/repos/claude/wikideck}
INTERVAL=${INTERVAL:-180}
LOG=${LOG:-/tmp/wikideck-night-autodeploy.log}
FORCE_AFTER=${FORCE_AFTER:-900} # defer while matches/lobbies are live, but at most this many seconds
deferred_since=
enc=$(printf %s "$BRANCH" | sed 's#/#%2F#g')
while :; do
new=$(node -e "fetch('$API/branches/$enc',{signal:AbortSignal.timeout(15000)}).then(r=>r.json()).then(b=>console.log(b.commit.id)).catch(()=>{})")
cur=$(git rev-parse HEAD)
if [ -z "$new" ]; then echo "WARN branch API failed"
elif [ "$new" != "$cur" ]; then
if ! git fetch -q origin "+refs/heads/$BRANCH:refs/remotes/origin/$BRANCH" 2>"$LOG"; then
echo "WARN fetch failed: $(tail -1 "$LOG")"
else
busy=$(deploy/night/testnet.sh busy 2>/dev/null || echo 0)
now=$(date +%s)
if [ "${busy:-0}" -gt 0 ] && { [ -z "$deferred_since" ] || [ $((now - deferred_since)) -lt "$FORCE_AFTER" ]; }; then
[ -z "$deferred_since" ] && { deferred_since=$now; echo "DEFER ${new%"${new#???????}"}: $busy live matches/lobbies"; }
sleep 60; continue
fi
[ -n "$deferred_since" ] && [ "${busy:-0}" -gt 0 ] && echo "FORCE deploy after ${FORCE_AFTER}s ($busy still live)"
deferred_since=
git checkout -q --detach "$new"
echo "DEPLOY start $(git log --oneline -1 | cut -c1-90)"
if deploy/night/testnet.sh rolling >"$LOG" 2>&1; then
sleep 20; echo "DEPLOY ok $(git rev-parse --short HEAD): $(deploy/night/testnet.sh status 2>&1 | tr '\n' ' ')"
else echo "DEPLOY FAILED $(git rev-parse --short HEAD): $(tail -3 "$LOG" | tr '\n' ' ')"; fi
fi
fi
sleep "$INTERVAL"
done
+28
View File
@@ -0,0 +1,28 @@
# sim-testnet override: publish the fixture on r350's LAN address only (not 0.0.0.0).
# WIKIDECK_USER_AGENT identifies the testnet to Wikimedia (UA policy; no personal contact info).
# Game-mode env (modes team, #6) is inert until night/modes/hosting merges.
services:
node-a:
ports: !override ["192.168.1.32:3101:3001"]
environment:
PUBLIC_ORIGIN: http://192.168.1.32:3101
WIKIDECK_USER_AGENT: &ua WikiDeck-testnet/0.1 (https://gitea.eom.dev/claude/wikideck; private testnet)
WIKIDECK_MODES: classic,trending
WIKIDECK_GAME_OPERATOR: Testnet A
WIKIDECK_GAME_PEERS: http://192.168.1.32:3102,http://192.168.1.32:3103
node-b:
ports: !override ["192.168.1.32:3102:3001"]
environment:
PUBLIC_ORIGIN: http://192.168.1.32:3102
WIKIDECK_USER_AGENT: *ua
WIKIDECK_MODES: draft
WIKIDECK_GAME_OPERATOR: Testnet B
WIKIDECK_GAME_PEERS: http://192.168.1.32:3101,http://192.168.1.32:3103
node-c:
ports: !override ["192.168.1.32:3103:3001"]
environment:
PUBLIC_ORIGIN: http://192.168.1.32:3103
WIKIDECK_USER_AGENT: *ua
WIKIDECK_MODES: classic,trending,draft
WIKIDECK_GAME_OPERATOR: Testnet C
WIKIDECK_GAME_PEERS: http://192.168.1.32:3101,http://192.168.1.32:3102
+100
View File
@@ -0,0 +1,100 @@
// sim-testnet: multi-worker native RandomX miner for the PRIVATE wikideck-night testnet.
// Runs inside the wikideck:testnet image (which ships the native engine). Testnet only.
// Usage: node --experimental-strip-types /night/mine.ts <blocks> <workers>
import { mkdirSync, readFileSync, writeFileSync, existsSync, appendFileSync } from 'node:fs';
import { randomInt, webcrypto } from 'node:crypto';
import { RandomX } from '/app/server/testnet/randomx.ts';
import { CHAIN_ID, sha256, miningBlob, mintId, blockSchema, address } from '/app/server/testnet/consensus.ts';
import { blockClaimMessage, NETWORK, type MiningTemplate, type TestnetBlock } from '/app/shared/testnet.ts';
import { encryptWallet } from '/app/shared/wallet-backup.ts';
if (!NETWORK.includes('testnet')) throw new Error('testnet only');
const blocks = Number(process.argv[2] ?? 45), workers = Number(process.argv[3] ?? 4);
if (!(workers >= 1 && workers <= 8)) throw new Error('1..8 workers');
const NODES = ['http://node-a:3001', 'http://node-b:3001', 'http://node-c:3001'];
const WALLET_NAMES = ['alice', 'bob', 'carol', 'dave'];
// PUBLIC, TESTNET-ONLY password for the disposable wallets' browser backups (posted on #4). Never reuse.
const PASSWORD = process.env.TESTNET_WALLET_PASSWORD ?? 'wikideck-night-testnet';
const STATE = process.env.STATE_DIR ?? '/state';
const dir = `${STATE}/wallets`;
mkdirSync(dir, { recursive: true });
const subtle = webcrypto.subtle;
type W = { name: string; address: string; publicKey: string; privateKey: CryptoKey };
async function loadOrCreate(name: string): Promise<W> {
const keyFile = `${dir}/${name}.jwk.json`;
let jwk: JsonWebKey;
if (existsSync(keyFile)) jwk = JSON.parse(readFileSync(keyFile, 'utf8'));
else {
const kp = await subtle.generateKey({ name: 'ECDSA', namedCurve: 'P-256' }, true, ['sign', 'verify']) as CryptoKeyPair;
jwk = await subtle.exportKey('jwk', kp.privateKey);
writeFileSync(keyFile, JSON.stringify(jwk), { mode: 0o600 });
}
const privateKey = await subtle.importKey('jwk', jwk, { name: 'ECDSA', namedCurve: 'P-256' }, true, ['sign']);
const pubJwk = { kty: jwk.kty, crv: jwk.crv, x: jwk.x, y: jwk.y };
const pub = await subtle.importKey('jwk', pubJwk, { name: 'ECDSA', namedCurve: 'P-256' }, true, ['verify']);
const publicKey = Buffer.from(await subtle.exportKey('spki', pub)).toString('base64');
const w = { name, address: address(publicKey), publicKey, privateKey };
const backupFile = `${dir}/${name}.backup.json`;
if (!existsSync(backupFile)) writeFileSync(backupFile, JSON.stringify(await encryptWallet(privateKey, publicKey, PASSWORD), null, 2));
return w;
}
const wallets = await Promise.all(WALLET_NAMES.map(loadOrCreate));
writeFileSync(`${dir}/addresses.json`, JSON.stringify(Object.fromEntries(wallets.map(w => [w.name, w.address])), null, 2));
console.log(`chain ${CHAIN_ID}; wallets: ${wallets.map(w => `${w.name}=${w.address.slice(0, 12)}`).join(' ')}`);
async function api(node: string, path: string, body?: unknown) {
const r = await fetch(node + path, { method: body ? 'POST' : 'GET', redirect: 'error', signal: AbortSignal.timeout(60000),
headers: body ? { 'Content-Type': 'application/json' } : {}, body: body ? JSON.stringify(body) : undefined });
const data = await r.json();
if (!r.ok) throw new Error(data.error ?? `HTTP ${r.status}`);
return data;
}
const engines = Array.from({ length: workers }, () => new RandomX());
const close = () => { for (const e of engines) e.close(); };
process.once('SIGTERM', () => { close(); process.exit(0); });
let accepted = 0, round = 0, fails = 0;
// Each template costs several Wikipedia API calls, and Wikipedia rate-limits this egress IP:
// pace templates (>= MIN_GAP s apart) and back off exponentially on failure.
const MIN_GAP = Number(process.env.MIN_GAP ?? 45), sleep = (s: number) => new Promise(r => setTimeout(r, s * 1000));
let lastTemplate = 0;
try {
while (accepted < blocks) {
const wait = lastTemplate + MIN_GAP * 1000 - Date.now(); if (wait > 0) await sleep(wait / 1000);
lastTemplate = Date.now();
const node = NODES[round % NODES.length], w = wallets[accepted % wallets.length]; round++;
let template: MiningTemplate;
try { template = await api(node, '/api/mining/template', { owner: w.address, publicKey: w.publicKey }); fails = 0; }
catch (e) { fails++; const b = Math.min(60 * 2 ** (fails - 1), 600); console.error(`template ${node}: ${(e as Error).message} (fail ${fails}, backoff ${b}s)`); await sleep(b); continue; }
const h = template.header; if (h.chainId !== CHAIN_ID) throw new Error('wrong chain');
const started = Date.now(); let attempts = 0; let found: { hash: string; nonce: number } | null = null;
let base = randomInt(0, 2 ** 40); const BATCH = 64;
while (!found) {
const results = await Promise.all(engines.map((e, i) => {
const hdr = { ...h, nonce: base + i * BATCH };
return e.mine(h.key, miningBlob(hdr), h.target, 32, BATCH);
}));
attempts += BATCH * workers; base += BATCH * workers;
found = results.find(Boolean) ?? null;
if (!found && Date.now() / 1000 - h.createdAt > 600) break; // stale template
}
if (!found) continue;
h.nonce = found.nonce;
const hash = sha256(miningBlob(h));
const signature = Buffer.from(await subtle.sign({ name: 'ECDSA', hash: 'SHA-256' }, w.privateKey,
new TextEncoder().encode(blockClaimMessage(hash)))).toString('base64');
const block: TestnetBlock = { ...h, hash, tokenId: mintId(h), powHash: found.hash, signature, transactions: template.transactions };
blockSchema.parse(block);
try {
const res = await api(node, '/api/blocks', block);
if (!res.accepted) { console.error(`not accepted: ${JSON.stringify(res).slice(0, 200)}`); continue; }
accepted++;
const secs = (Date.now() - started) / 1000;
const line = `#${h.height} ${w.name} "${h.article.title}" token=${block.tokenId} via=${node} ${attempts}H ${(attempts / secs).toFixed(0)}H/s`;
console.log(`[${accepted}/${blocks}] ${line}`);
appendFileSync(`${STATE}/mined.log`, `${new Date().toISOString()} ${line}\n`);
} catch (e) { console.error(`submit: ${(e as Error).message}`); }
}
} finally { close(); }
console.log('done');
+49
View File
@@ -0,0 +1,49 @@
#!/bin/sh
# sim-testnet: private 3-node testnet fixture for WikiDeck Night 1 (poweredge-r350, LAN only).
# Usage (from the repo root):
# deploy/night/testnet.sh up build the image from this checkout and (re)start the 3 nodes
# deploy/night/testnet.sh rolling build, then restart node-a, b, c one at a time (waits for /healthz)
# deploy/night/testnet.sh busy print live matches + open lobbies across the 3 nodes
# deploy/night/testnet.sh mine N [W] mine N real RandomX blocks with W workers (default 4, max 8)
# deploy/night/testnet.sh status tip/height of each node
# deploy/night/testnet.sh down stop the nodes (volumes, i.e. the ledgers, are kept)
# STATE_DIR (default ~/.local/share/wikideck-night) holds the disposable testnet wallets and mined.log.
# It is outside the repo on purpose: never commit wallets.
set -eu
P=wikideck-night
LAN_IP=${LAN_IP:-192.168.1.32}
STATE_DIR=${STATE_DIR:-$HOME/.local/share/wikideck-night}
DC="docker compose -p $P -f compose.testnet.yaml -f deploy/night/compose.lan.yaml"
case "${1:-}" in
up) $DC build node-a && $DC up -d --no-build ;;
down) $DC down ;;
rolling)
# Never take the whole testnet dark: one node at a time, each must pass /healthz before the next.
$DC build node-a
for n in a b c; do
port=$(( 3100 + $(printf %d "'$n") - 96 ))
$DC up -d --no-build --no-deps node-$n
i=0
until node -e "fetch('http://$LAN_IP:$port/healthz',{signal:AbortSignal.timeout(3000)}).then(r=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))"; do
i=$((i+1)); [ $i -ge 60 ] && { echo "node-$n not healthy after 120 s"; exit 1; }; sleep 2
done
echo "node-$n healthy"
done ;;
busy)
# Sum of liveMatches + openLobbies over all hosted modes on all nodes (unreachable node counts 0).
node -e "Promise.all([3101,3102,3103].map(p=>fetch('http://$LAN_IP:'+p+'/api/game/modes',{signal:AbortSignal.timeout(5000)}).then(r=>r.json()).then(m=>m.modes.reduce((t,x)=>t+(x.stats?.liveMatches??0)+(x.stats?.openLobbies??0),0)).catch(()=>0))).then(v=>console.log(v.reduce((a,b)=>a+b,0)))" ;;
mine)
mkdir -p "$STATE_DIR"
W=${3:-4}
docker rm -f $P-miner >/dev/null 2>&1 || true
docker run -d --name $P-miner --network ${P}_default --cpus "$W" --memory 1600m \
--user "$(id -u):$(id -g)" -e WIKIDECK_NETWORK=testnet -e WIKIDECK_GENESIS_TIME=1792411200 \
-e MIN_GAP="${MIN_GAP:-45}" -v "$STATE_DIR":/state -v "$PWD/deploy/night":/night:ro \
wikideck:testnet node --experimental-strip-types /night/mine.ts "$2" "$W"
echo "follow with: docker logs -f $P-miner" ;;
status)
for port in 3101 3102 3103; do
docker run --rm --network host wikideck:testnet node -e "fetch('http://$LAN_IP:$port/p2p/status').then(r=>r.json()).then(s=>console.log($port,'blocks',s.size,'tip',s.tip.slice(0,16))).catch(e=>console.log($port,'DOWN',e.message))"
done ;;
*) sed -n 2,11p "$0"; exit 1 ;;
esac
+17
View File
@@ -0,0 +1,17 @@
# Canonical ledger indexes
The ledger maintains SQLite indexes for canonical NFT ownership/sequence, mint ordering, block identity, and transaction inclusion. The UI uses these indexes for wallet decks, NFT metadata, and current-chain pagination; they do not change consensus rules or replace proof validation.
## Persistence and reorganization behavior
Accepted blocks, pending transactions, index connect/disconnect updates, undo records, and canonical event history commit in one SQLite transaction. Losing side branches and mempool submissions do not create canonical ownership events. On a reorganization, disconnects apply tip-first and connects ancestor-first. A failed write leaves both the stored ledger and in-memory tip unchanged.
On initial open, stored blocks are replayed and validated before the local index is built or checked. Later opens verify indexed block identity, ownership, transaction inclusion, and event continuity against that validated chain. Inconsistent state fails closed; no automatic reset or history pruning occurs. These local indexes and event records are not consensus data or a public proof that a node observed an honest network.
Wallet-deck results are paged and can be pinned to a canonical tip; stale tips return HTTP 409. Canonical block pages are bounded. Direct trades continue to be validated as complete atomic transactions against the canonical ownership state. No market journal, reconciliation worker, payment observer, or NFT sale-lock signer is part of this design.
## Validation and capacity limits
The regression suite covers indexed ownership and sequence against replay for transfers, atomic multi-card trades and forks; append/disconnect/reconnect ordering; side-branch and mempool exclusion; durable restart; failed writes; and stale-tip pagination.
The [canonical-index benchmark](../scripts/benchmark-canonical-index.ts), available as `npm run benchmark:index -- 10000`, creates and removes its own temporary SQL fixture. Its synthetic results are a baseline only—not production acceptance limits. Real PoW validation, startup replay, full DAG memory, concurrent HTTP load, and the intended deployment hardware must be measured independently. Indexes improve lookups but do not solve the retained-DAG and full-state costs described in the [consensus reference](testnet-consensus.md) and [production plan](card-only-production-plan.md).
+113
View File
@@ -0,0 +1,113 @@
# Card-only production readiness plan
**Status:** proposed readiness plan, updated 2026-10-28. This is not launch approval, security certification, or permission to expose a public node.
## Product and release boundary
Wikideck's current product scope is **RandomX mining, collecting cards, and direct player-to-player card transfers/trades**. This release will not implement or offer a marketplace, auctions, sales, custody, XMR deposits/withdrawals, or real-money checkout. Direct card trades are signed and applied atomically by the chain; they are not paid orders.
The nearest reasonable release target is a **publicly hosted, non-monetary testnet beta**. It must be labeled experimental and valueless. It is not a mainnet launch, a promise of ownership permanence, an investment product, or an independently secured network. Mining consumes compute, electricity, and time; no reward rate or profitability is promised.
Mainnet or valuable-card use is a separate, higher bar. It remains **NO-GO** until consensus, wallet, network, capacity, deployment, and operator risk-acceptance gates below have evidence and named sign-off. The operator has chosen not to commission external audits; this is a conscious acceptance of additional residual risk, not evidence that the protocol or implementation is secure.
## Current starting point
- Mining, local encrypted wallets, collections, transfers, and atomic card-for-card trades are implemented for the experimental testnet.
- Marketplace, custody, sale-settlement, and Monero payment source, UI, APIs, deployment profiles, active tests, and operator documentation have been removed from the supported repository. Direct on-chain card transfers/trades remain core functionality.
- Existing runtime volumes and historical backup packages may contain local financial journals or external database bindings. Removing source code does not delete or resolve those records; preserve them separately and do not claim new ledger-only backups include them.
- The current network has no public bootnodes or automatic peer discovery. Peers must be configured by operators; containers on one host do not provide independent network security.
- The inherited initial proof-of-work target accepts one in sixteen hashes, and the ten-block retarget is not production-calibrated. Nodes retain the DAG in memory, rebuild state roots across the NFT set, and revalidate stored proofs at startup.
- Browser-wallet encryption is implemented, but same-origin code can abuse an unlocked wallet; old plaintext copies and cross-browser recovery remain concerns.
See the [mainnet launch limitations](mainnet-launch.md), [consensus reference](testnet-consensus.md), [wallet limitations](wallet-security.md), and [security review](security-review.md). Existing passing tests and local rehearsals are useful evidence, not production certification.
## Ordered release gates
### Gate 0 — freeze the card-only scope · implemented
- Publish the scope above in the product and operator docs. Market/custody/payment source, UI, APIs, dependencies, tests, deployment files, and operator tasks are removed from the supported repository.
- Preserve existing node and financial data during retirement. Do not remove Compose volumes, old journals, PostgreSQL schemas, or historical backups. Review/archive any existing market journal before stopping a market deployment; the card-only node will not reconcile or resolve it.
- Version the node backup contract for ledger-only backups. Refuse to create a ledger-only package if legacy financial data or a PostgreSQL custody binding remains in the selected data directory. Keep old paired backup verification/restore available as a data-preservation path; restored legacy files are inert and are not reopened by the card-only application.
- Update the production plan, operator launch checklist, security scope, support text, and release notes so they no longer imply that market, custody, or payment work is part of this release.
- Choose and document the only supported launch target: testnet beta or a separately reviewed mainnet. Do not mix the two identities or reuse testnet data/backups as mainnet state.
**Exit evidence:** scope owner approval is still required. Source/deployment removals are complete; API regression checks cover removed routes while preserving card trades; backup v2 and legacy-data refusal are covered; old paired backups remain portable. No active product docs promise sales or payments.
### Gate 1 — internal protocol and application review · complete by operator decision
**Operator decision (2026-10-28):** no external audit or independent review will be commissioned for this release. The product will proceed using internal code review, regression tests, local adversarial exercises, and explicit operator risk acceptance. This does not certify the software or establish that undiscovered critical defects are absent.
Internal source-assisted review covered consensus serialization/validation, signatures, state commitments, fork choice, timestamps/retarget, peer URL/DNS/pagination/response handling, RandomX native boundary, wallet signing/recovery documentation and tests.
Internal findings and dispositions:
- **Clock rollback could strand accepted SQLite history.** Startup replay applied the live future-skew admission bound to every persisted block. A block accepted at the allowed future edge could fail replay if the clock later moved backward. Startup now replays previously accepted SQLite history without reapplying that live-time bound. New blocks imported from legacy JSON retain a pinned import-time clock and imported-block cutoff; their future timestamps are checked on import and on later replays, so an untrusted legacy file cannot seed arbitrarily future-dated history. Regression tests cover both rollback recovery and repeated rejection of future-dated legacy data.
- **Invalid work reached full NFT-state materialization.** Block validation computed the state transition/root before checking the claimed and actual RandomX proof. Cheap commitment/signature/claimed-target checks now precede native proof verification, and the complete state transition/root is computed only after the native proof matches. A regression uses an instrumented state to assert that a wrong claimed proof is rejected before NFT-state enumeration.
- Added a deterministic 30-block versus 16-block competing-fork regression: equal cumulative work keeps the current tip, and a strictly higher-work branch then causes a 30-block reorganization while preserving both DAG branches. This is a local FakePow consensus-state test, not a multi-host security rehearsal.
Current validation evidence: the Docker test image and TypeScript/Vite build succeeded; the complete offline suite passed **95/95** with the native RandomX HTTP integration; the packaged native engine's upstream known-vector self-test passed. Tests cover signatures/transaction replay, atomic trades, commitments, retarget/key boundaries, equal-work fork choice, reorg/requeue, malformed schemas, peer DNS/manifest/pagination/response rejection, mempool and RPC bounds, wallet backup crypto, and restart/revalidation. These checks are the selected assurance approach; they are not a security certification.
**Residual risks carried forward:** no external auditor or independent reviewer will sign off. The candidate is not frozen. Multi-host eclipse/partition and clock-skew rehearsals, hostile populated-state load/reorg tests, and broad browser-wallet/XSS/recovery exercises remain open engineering work for Gates 2–5. The operator accepts proceeding without external audit; do not describe the result as audited or independently certified.
**Exit evidence:** internal findings fixed with regression coverage; current build, offline suite and native known-vector results recorded; operator accepts the unaudited-risk boundary. Gate 1 is **complete under this explicitly reduced scope**. This completion is not external assurance, and its remaining multi-host/load/wallet exercises continue under the following gates.
### Gate 2 — establish a credible network · in progress
- **Mainnet candidate inspection (2026-10-28, read-only):** `wikideck-randomx-mainnet-1`, candidate chain ID `3ef36901f49d57127ebb0bf43625c7334e9b8ef817fd31953761723b5474ef6c`, genesis `72be7cdac8e625562bc8a49a82a406f41b96a6fde285c9961dceafdbaca8cd1d`, candidate genesis time 2027-01-01 00:00 UTC. The genesis time has not arrived; this is a candidate manifest, not an announced launch. No production node was started and no approval or data volume was created.
- **Repeat benchmark (2026-10-28):** same executable and container limits, 32 hashes measured 25.11 H/s warm (p50 39.13 ms, p95 48.18 ms), 883 ms cold initialization, and 642 ms for a key/cache switch. These two short samples are not a confidence interval and do not establish normal host variability. Both are verification-path hashes, not miner throughput, dedicated-host measurements, aggregate network hashpower, or a security budget. Do not extrapolate them to choose consensus parameters.
- The first bounded sample in the same container was 23.35 H/s warm (p50 41.09 ms, p95 55.81 ms), 796 ms cold initialization, and 594 ms for a key/cache switch. Together these two observations show run-to-run variation only; they are not intended-host miner measurements.
- **Bootstrap policy for the first beta:** use a fixed, operator-curated HTTPS peer allowlist supplied as `WIKIDECK_PEERS`; do not add DHT, automatic discovery, or peers learned from untrusted network messages. Treat configured peers as trust/bootstrap contacts only; each node independently validates every block and transaction. The current protocol provides no peer identity or independent-operation proof beyond the configured URL and TLS connection.
- **Local rehearsal status:** the complete offline suite's native RandomX HTTP integration syncs two separately started node processes over loopback, gossips a mined transfer, and verifies restart/replay. This is local functional evidence only. A disposable three-node Compose attempt on 2026-10-28 could not allocate a Docker bridge subnet because the configured address pools were exhausted; Compose failed before creating services or volumes. No multi-host/failure-domain rehearsal has been performed.
- **Mainnet peer configuration remains empty/unverified:** the repository has no operator-supplied public HTTPS origins, ownership/failure-domain evidence, or exercised TLS endpoints for the two required mainnet peers. The compose profile only forwards operator-provided peers and origin; placeholders are not launch infrastructure.
- For a multi-node rehearsal, use at least three separately administered nodes across distinct failure domains and network providers; record operator, host/region/provider, address/origin, chain ID, software digest, availability window, and which node supplies each bootstrap connection. Three containers or aliases on one host do not count as independent.
- Measure expected and adversarial hashpower on intended hardware; quantify rewrite costs and concentration risk. Decide whether this experimental beta retains the current easy target or whether a future mainnet identity needs different initial difficulty, retarget, timestamp, and stall-recovery rules. Do not alter a launched chain's consensus.
- Rehearse bootstrap with an empty node, peer loss/recovery, partition/rejoin, conflicting branches, stale/malicious configured peer, and independent mining. Record the observed tips, cumulative work, sync lag, and recovery time. Peer URL count alone is not evidence that peers are online or independent.
- State the finality/reorganization limits honestly. There is no formal finality today; a confirmation count is not an absolute guarantee.
**Exit evidence:** explicit testnet-only beta decision (or separately approved mainnet candidate), documented bootstrap peer allowlist and ownership/failure domains, repeated intended-hardware hashrate data with variance, multi-host partition/reorg/stall rehearsal, a written risk/attack-cost estimate, and an operator-owned peer/upgrade/incident plan. Internal sign-off only; no external audit is planned.
### Gate 3 — demonstrate capacity and recovery limits
- Define the expected chain length, block rate, peer count, user/API load, disk budget, memory ceiling, startup time, and latency/error objectives.
- Run repeated tests on populated long-chain fixtures and the intended hardware, with concurrent mining, peer sync, wallet/deck reads, transfers/trades, reorgs, and hostile traffic. Measure tail latency, memory, CPU, disk/WAL growth, and restart/revalidation time.
- Test clean shutdown, process crash, disk pressure, corrupted/stale state, isolated restore, and replay using the exact release binary. Do not treat `/healthz` alone as readiness.
- Define and test an archival, pruning, or capacity transition before resource growth can make nodes unable to sync or restart. Do not alter consensus history silently to reclaim storage.
**Exit evidence:** reproducible benchmark report with workload and limits, enforced resource budgets, failure/reorg results, and a successful isolated restore/replay drill.
### Gate 4 — wallet, trades, and user protection
- Test create, lock, unlock, backup, restore in a fresh browser/profile, and recovery after app/browser upgrade across the supported browser matrix. Test signed transfer/trade review, stale offers, pending state, and reorg outcomes.
- Perform an internal threat review of same-origin script compromise, extensions, password entry, old plaintext-wallet migration, downloaded backup handling, and key-loss support. Decide whether the beta accepts browser-only signing; document residual risks rather than implying external wallet assurance.
- Publish experimental/no-value disclosures, mining resource risks, irreversible-transfer warnings, privacy/data-retention terms, Wikipedia text/image attribution and license handling, support contact, abuse-report path, and stop conditions.
- Select and publish an appropriate project software license before public distribution. A Wikideck license does not replace Wikipedia or dependency licenses.
**Exit evidence:** internal wallet threat review, verified recovery drill, tested trade flows, reviewed participant disclosures, public support/abuse contacts, and license decision.
### Gate 5 — operate and release safely
- Prefer a managed host unless the intended self-hosted site proves ISP permission, reachable ingress/CGNAT handling, DNS/TLS renewal, trusted-proxy correctness, no direct backend exposure, firewalling, restricted administration, patching, and power/network resilience.
- Deploy independent monitoring and alert delivery for peer/tip freshness, reorgs, native miner failures, latency/queue saturation, disk/WAL growth, and service availability. Assign a real responder and incident/rollback procedure; do not imply HA or 24/7 response without it.
- Encrypt backups off-host and rehearse cold restore. Define and meet recovery point/time objectives; protect wallet recovery separately from node data. Preserve database/WAL consistency and verify restored proofs/ownership before resuming service.
- Freeze the release artifact: reviewed base-image digest, dependency/native versions, SBOM, vulnerability and native-code scans, build provenance/signature, configuration manifest, and rollback image. Re-run the complete suite and smoke test against that exact artifact.
- Run a limited, non-monetary soak and record go/no-go, known risks, release hashes, operator, reviewer, and stop criteria. Do not convert a beta into mainnet by changing an environment variable or changing the chain identity after launch.
**Exit evidence:** intended-host deployment test, alert drill, off-host restore result with measured RPO/RTO, signed/frozen artifacts, exact-release test report, and recorded operator/reviewer approval for the beta's limited exposure.
## Release decisions
### Non-monetary public beta
May be considered only after all applicable gates pass, the node remains on the explicitly labeled experimental testnet, cards are not represented as valuable, and the operator accepts the single-host or stated availability limits. This is a production-hosted beta, not production chain security.
### Mainnet or valuable cards
Requires all gates plus measured and operator-accepted consensus/economic risks, calibrated and frozen consensus parameters, a credible peer/miner topology, reviewed wallet protections, sustained capacity and adversarial evidence, and explicit governance/upgrade and incident procedures. External audit is intentionally not a requirement under the operator's decision, so residual risk remains higher and must be explicitly accepted before any mainnet activation. Until then, retain **NO-GO**.
### Markets and real money
Out of scope for this product plan. No market/custody/payment milestone is a prerequisite to the card-only beta. Any later decision to add sales or funds requires a new product scope, threat model, legal review, settlement/custody design, and separately approved release plan; it must not be inferred from this document.
## Evidence record
For every gate, record an owner, exact release/manifest, test or rehearsal, measured result, artifact location/hash, reviewer, residual risk, and disposition. Missing evidence is **unverified**, not passed. A successful build, local container startup, peer count, test total, or environment flag is never launch approval by itself.
+62
View File
@@ -0,0 +1,62 @@
# Consensus calibration and adversarial analysis — stage 3
**Decision remains NO-GO.** This stage supplies reproducible offline tools and bounded measurements, not new launch parameters or an operational network. Consensus, initial target, genesis, and chain identity are unchanged.
## Tools and safety boundaries
- `analyze:consensus`: deterministic JSON report calling the **actual** `expectedTarget`, `medianTime` and `workFor` helpers. Six 60-record traces: steady 900-second spacing, rapid one-second spacing, slow 7,200-second spacing, a seven-day pause at a window boundary, minimum-median timestamps with normally advancing wall time, and timestamps at the allowed +120-second future edge.
- Records contain only fields read by those mathematical helpers. They are **not valid blocks**, have no PoW/signatures/state commitments, are never submitted to a ledger and do not add a fake-PoW production option. Spacing is externally imposed rather than sampled from constant-hashrate mining. The comparison of cumulative work is not proof that a miner can actually produce those traces.
- `benchmark:randomx`: requires `--consent`, uses the trusted local reference binary and accepts `--hashes` (4–256, default 32), optional `--fast` (explicit >2.3 GiB memory choice), and `--network-hps` (integer 1–10^12, default 1). Unknown/invalid arguments fail before starting native work. A 60-second wall-time deadline closes the engine; native requests also retain their own bounds. No endpoint, wallet, ledger, payout or launch option exists. CPU/energy consumption is explicit.
- Benchmark output includes executable SHA-256, platform/Node version, mode, cold hash time, repeat hash, warm sequential IPC-inclusive rate, hash latency samples and one epoch/cache switch. The declared network hashrate is a **hypothesis**, never inferred by multiplying a local worker rate by an assumed number of machines. Output targets are illustrative strings only and are never written into consensus/configuration.
These tools can be invoked via the package scripts on Node 24. The Docker test target includes the trusted Linux-native engine; offline benchmark execution can use a network-disabled disposable container. Do not benchmark a production node or enable fast mode without an explicit memory/energy budget. Keep reports with the exact image/native digest and host/container constraints.
## Verified arithmetic and interpretation
Wikideck interprets the 256-bit RandomX output as a big-endian integer and accepts hash <= target. Assuming uniform hash output, success probability is `(target + 1) / 2^256`, expected trials are `2^256 / (target + 1)`, and expected seconds at H hashes/s are expected trials divided by H. These are expectations, not schedules or upper time bounds.
The current `0fff…ffff` maximum/initial target requires **16 expected hashes**. The 15-minute setting controls the retarget goal; it does not force miners to wait 15 minutes before submitting an initially easy block. For a specified integer network rate H and goal S seconds, the analysis computes `floor(2^256 / (H*S)) - 1`, bounded by the existing allowed target range. This formula assumes stable aggregate honest work and does not choose a safe launch rate, security budget, or rounding-independent exact interval.
The report also includes `(q/(1-q))^z` for q < 0.5 (1 for q >= 0.5), under an idealized fixed-difficulty Poisson race with attacker share q and a **known current deficit** of z blocks. This is the probability of ever catching up under those assumptions, **not** a guarantee that a block is final: attacker progress while a block accumulates confirmations is not modeled. There is no network latency, eclipse, selfish mining, timestamp manipulation, adaptive difficulty, economic cost or attack detection model.
## Local native measurement (2026-10-20)
One sample, offline Linux x64 Docker, **1 CPU quota / 1 GiB memory limit**, Node v24.22.0, light mode, 32 sequential warm hashes, existing pinned RandomX v1.2.1 implementation:
| Measurement | Result |
| --- | --- |
| Executable SHA-256 | 5e1b99ea9df075d75a676a4ea1090e62d7738d03ac6316c41f59dedfffac4318 |
| Cold hash (including initialization/startup) | 7,379 ms |
| Immediate repeated same-key hash | 1,520 ms |
| Warm batch elapsed | 16,645 ms |
| Warm sequential IPC-inclusive rate | **1.9225 H/s** |
| Warm p50 / p95 individual request | 552 / 755 ms |
| One epoch/cache switch hash | 4,226 ms |
| Initial expected interval at this measured single-worker rate | **8.323 seconds** |
The host was **not a dedicated controlled benchmark machine**; co-tenancy, scheduling, container CPU throttling, virtualization and thermal conditions were not instrumented. The surprising repeat-vs-warm timings reinforce that one short sample cannot certify capacity. It is a `hash`/verification-path measurement with per-request IPC, not native miner batch throughput; it includes neither full block validation nor state-root/SQLite/network overhead. Cache initialization/switch measurements are single samples, not statistically characterized distributions. Fast mode was not measured. Node `process.memoryUsage` would omit the native subprocess, so no misleading whole-engine RSS claim is made.
For an **explicitly hypothetical** aggregate 100 H/s network and 900-second target, the tool produced illustrative target `0000ba69dbdd3ac13d7c4a74c8111cb7aecee4bd24e8d5b85d9222dc8bfdff5b` (90,000 expected hashes). This was **not applied** and is not a recommendation. Expected rewrite cost, miner diversity, rate variance, launch surge/decline and adversarial hashrate remain unmeasured.
## Deterministic scenario findings
1. **Steady/slow at maximum target:** stable 900-second spacing retains the initial target; 8× slower spacing cannot ease above maximum. There is no wall-clock emergency easing during a stall.
2. **Rapid trace:** difficulty does not change before the ten-block boundary. After ten one-second-spaced records the target is quartered by the elapsed-time floor (`floor(8100/4) = 2025 seconds`). Further imposed rapid windows keep increasing required work; this does not mean the same hashrate could sustain those arrival times.
3. **Seven-day boundary pause:** the pause is placed before the first block of the third ten-block window. The endpoint difference within that window excludes the inter-window gap. This illustrates a boundary blind spot, not a fully executed exploit: the choice of where a pause occurs matters and requires review of alternative retarget designs. The current maximum target also limits what this particular trace can reveal about recovery from already-high difficulty.
4. **Median-minimum trace:** timestamps just above the actual prior median can satisfy timestamp bounds while lagging wall time substantially. Retarget uses timestamp differences, not independently verified elapsed mining time. Analyze time-warp/oscillation attacks with quantified hashpower and viable block production before selecting rules.
5. **Constant future-edge offset:** +120 seconds is allowed by the current timestamp rule. A constant offset cancels out of interval differences; it is not by itself evidence of a difficulty exploit. Clock skew and shifting offsets still need adversarial analysis. Validator tests reject timestamps at/below median or beyond future allowance **before proof work** and exercise the exact allowed edge reaching subsequent commitment checks.
6. **Partition work comparison:** synthetic steady and rapid branches have different accumulated work, and the existing strict-work rule favors the greater-work branch. Existing ledger/native integration tests cover real validated branch/reorg handling separately. The synthetic report is not a multi-host partition, adversarial mining or economic-security rehearsal.
## Validation and remaining launch evidence
Build and full regression suite passed **92 tests, 0 failed, 0 skipped**, including real RandomX HTTP integration and six new calibration/timestamp tests. Tests assert inclusive-target arithmetic, bounded inputs, exact window/clamp behavior, reproducibility, scenario timestamp bounds, probability assumptions, consent refusal before engine access and validator timestamp rejection. There is no claim of formal consensus correctness.
Before production calibration can be approved:
- Benchmark the exact final native/image build on multiple intended independent miner/validator hosts; record CPU model, operating mode, effective quotas, competing load, cold/warm/cache-switch distributions and batch-miner throughput. Use repeated runs and documented confidence/variance, not this single sample.
- Establish credible aggregate honest hashrate, launch ramp/surge/drop ranges and attack/rewrite cost assumptions; independently review target choice and the inherited ten-block/clamped endpoint retarget, including gaps across windows and timestamp attacks.
- Add stochastic mining/partition/clock/adaptive-hashpower models and sustained multi-host rehearsals with valid reference proofs, full state commitments, mempool/reorg persistence, networking and realistic resource pressure.
- Freeze the selected rules and actual genesis date with a new identity where required. Publish the manifest/artifact hashes and record internal findings plus explicit operator risk acceptance. Do not patch an active chain's rules in place.
- Document reorganization and confirmation limitations honestly; no finite confirmation count is absolute finality on this unbootstrapped PoW chain.
See [security review](security-review.md), [launch evidence register](operator-launch.md) and the read-only preflight. The calibration blocker remains open.
+464
View File
@@ -0,0 +1,464 @@
# WikiDeck: Edit War — balance report
Owner: sim-balance (claude/wikideck#4). Harness: `shared/bots/`, `scripts/sim/` (PR #14).
## Report v1: 2026-11-17 00:45 EST
**Engine:** `night/2026-11-16` @ 482de6c/4a4941c (engine #25 and #27, keywords and Portals; card data #28).
**Bots:** view-only (`viewFor`); see Method.
**Card pool:** `docs/game/card-sample.json` (236 cards), derived per §1 with percentiles from the sample.
**Decks:** the 4 §2.1 starter decks, 4 "realistic owned" decks (12 random cards from the sample), and archetypes.
**Runs:**
- full field: 3,000 matches per ruleset (report mix, 4 workers; r350 is memory-limited tonight)
- focused starters + owned round robin: 2,000 matches per ruleset
- greedy lever sweeps: 2,000–3,000 per variant
## v1.1 re-run on the real engine hooks (00:31)
> **Seat figures are pre-#38.** Every first-player and seat number in this report was measured on an engine before #38, which removed `opponent.queuedCount` from `PlayerView`. The bots never read the queue split (`stateFromView` uses the total hand, and an invariance test covers it), so the effect should be nil. A re-run on the merged head is still the clean number, and it is open for v1.2.
**Engine:** `night/engine/v1-1` @ 4a0dbd0, with `RULESET_V1_1` (DESIGN #32): `viral {base 1, cap 3}`, `encyclopedic {sides 1}`, `rising.powerDelta -1`, `cultureKeyword: home`, `costThresholds [26,39,51,63,78]`. It is the same focused run as above: the §2.1 starters plus 4 owned decks (cost thresholds and Culture keyword per v1.1), 2,000 matches per ruleset.
| | v1 | **v1.1 (#32, exact)** | v1.1 + item 6 |
|---|---|---|---|
| starter deck WR, overall (FG / Head / Pio / Star) | 65 / 40 / 57 / 54 | **52 / 52 / 52 / 53** ✅ | 50 / 56 / 50 / 52 |
| starter pairings, min–max | 23–77% | 39–61% (only FG–Head 61%, n = 51, is outside 45–55) | 42–58% |
| owned vs starters (FG / Head / Pio / Star) | 37 / 54 / 40 / 41 | 51 / 42 / 47 / 43 | 51 / 41 / 46 / 48 |
| first player (pre-#38) | 53.4% | **51.2% [48.8–53.6]** ✅ | 51.0% [48.6–53.5] |
| decided in the first third | 4.1% | 3.3% | 4.1% |
| greedy > maxpower, lookahead-lite > greedy | 62%, 60% | 63%, 65% | 51%, 70% |
Net keyword value per card under the exact v1.1 (counterfactual trigger value plus `powerDelta`):
| keyword | v1.1 net per card |
|---|---|
| translated | **+1.64** |
| nostalgic | **+1.69** |
| viral | +1.15 (base included) |
| encyclopedic | +0.77 |
| rising | 1.71 − 1 = +0.71 |
| local | +0.69 |
My earlier flat-+1 stand-in for encyclopedic overstated it: the real hook is worth +0.77, not +1.23.
**Updated recommendation:**
1. **Ship #32 (v1.1) as it is.** Every starter's overall WR is 52–53% and first player is 51.2%. The one out-of-band pairing (FG vs Headliners at 61%) is within this run's noise.
2. **Item 6 is an optional v1.2 candidate.** `nostalgic` and `translated` at `powerDelta -1` would bring translated and nostalgic (about +1.65) in line with the others (0.7–1.15). It fixes FG vs Headliners (50%), but at n ≈ 55 per pairing it isn't clearly better overall (42–58% vs 39–61%). Decide it on a larger run.
3. **Owned vs starters is 42–51%:** roughly the variety-not-power goal, slightly weak against Headliners and Stargazers.
Raw tables:
#### `v1.1`: 2000 matches, 10 draws, 0 engine errors
- First player (turn-1 initiative), symmetric games: **51.2% [48.8–53.6] (n=1641)**
- Decided early: 3.3% of decisive games locked in during the first third, 12.1% by half-time
- Skill: greedy>maxpower 63.3% (n=147), lookahead-lite>greedy 64.6% (n=99), greedy>random 83.5% (n=103)
| starters (row vs col) | field-guide | headliners | pioneers | stargazers | vs all decks |
|---|---|---|---|---|---|
| field-guide | – | 61% (n=51) | 49% (n=53) | 56% (n=55) | 51.8% |
| headliners | 39% (n=51) | – | 44% (n=54) | 50% (n=54) | 52.1% |
| pioneers | 51% (n=53) | 56% (n=54) | – | 48% (n=58) | 52.3% |
| stargazers | 44% (n=55) | 50% (n=54) | 52% (n=58) | – | 53.2% |
| owned (random 12 from card-sample) vs | field-guide | headliners | pioneers | stargazers | all |
|---|---|---|---|---|---|
| owned-0 | 39% | 32% | 41% | 46% | 40.3% |
| owned-1 | 50% | 43% | 42% | 38% | 45.1% |
| owned-2 | 55% | 37% | 62% | 44% | 49.7% |
| owned-3 | 62% | 55% | 43% | 45% | 55.0% |
| **all owned** | **51%** (n=233) | **42%** (n=196) | **47%** (n=198) | **43%** (n=198) | |
| keyword | power per card | games decided by it | cards played |
|---|---|---|---|
| rising | +1.71 | 13.3% | 1503 |
| nostalgic | +1.69 | 13.4% | 1791 |
| translated | +1.64 | 17.8% | 4976 |
| viral | +1.15 | 7.3% | 1852 |
| encyclopedic | +0.77 | 6.2% | 2349 |
| local | +0.69 | 6.4% | 4052 |
#### `v1.1+item6`: 2000 matches, 9 draws, 0 engine errors
- First player (turn-1 initiative), symmetric games: **51.0% [48.6–53.5] (n=1640)**
- Decided early: 4.1% of decisive games locked in during the first third, 12.6% by half-time
- Skill: greedy>maxpower 51.0% (n=147), lookahead-lite>greedy 70.0% (n=100), greedy>random 82.7% (n=104)
| starters (row vs col) | field-guide | headliners | pioneers | stargazers | vs all decks |
|---|---|---|---|---|---|
| field-guide | – | 50% (n=52) | 50% (n=52) | 53% (n=55) | 49.7% |
| headliners | 50% (n=52) | – | 58% (n=55) | 48% (n=54) | 56.1% |
| pioneers | 50% (n=52) | 42% (n=55) | – | 42% (n=57) | 49.7% |
| stargazers | 47% (n=55) | 52% (n=54) | 58% (n=57) | – | 51.9% |
| owned (random 12 from card-sample) vs | field-guide | headliners | pioneers | stargazers | all |
|---|---|---|---|---|---|
| owned-0 | 48% | 41% | 50% | 39% | 47.1% |
| owned-1 | 41% | 44% | 47% | 52% | 47.2% |
| owned-2 | 57% | 36% | 38% | 51% | 48.8% |
| owned-3 | 60% | 41% | 49% | 49% | 49.4% |
| **all owned** | **51%** (n=233) | **41%** (n=197) | **46%** (n=200) | **48%** (n=198) | |
| keyword | power per card | games decided by it | cards played |
|---|---|---|---|
| rising | +1.72 | 14.2% | 1546 |
| translated | +1.69 | 17.0% | 4787 |
| nostalgic | +1.68 | 13.2% | 1709 |
| viral | +1.15 | 8.8% | 1924 |
| encyclopedic | +0.79 | 6.7% | 2403 |
| local | +0.70 | 6.7% | 4127 |
### TL;DR for the council (report v1, 00:25; see the v1.1 re-run above for the updated recommendation)
1. **Seat and initiative are fine (pre-#38):** 51.5% [49.6–53.5] for the turn-1 initiative holder (n = 2,421 symmetric games), and seat 0 51.2%. Draws are 0.4%. The tie-break (total power) decides 4.7% of games. *This full-field run started before the bot-side queue-count fix (PR #14); the effect is at most a 0.02 nudge to the heuristic. The focused starters run, after the fix, gives a borderline 53.4% [51.0–55.8] under v1 and **51.0% [48.6–53.4]** under the recommended v1.1. Initiative only matters through reveal order, and v1.1 removes most of the order-sensitive power.*
2. **Keyword values are lopsided, confirming the critic's figures within 0.2:**
| keyword | power per card |
|---|---|
| rising | +1.76 |
| translated | +1.53 |
| nostalgic | +1.45 |
| local | +0.65 |
| **viral** | **+0.19** |
| **encyclopedic** | **+0.19** |
Values are counterfactual replays. viral and encyclopedic are dead: raising viral to amount 2 / cap 3 only reached +0.24, and encyclopedic amount 3 only +0.33. **Their conditions are the problem, not their numbers.**
3. **The starter decks are not 45–55% in v1:** pairings run 23–77%. Field Guide (rising + nostalgic) wins 65% overall and Headliners (local + viral) 40%. **With v1.1 items 1–6, every pairing is 45–55%.**
4. **Owned decks lose to the starters**, because they are about 51% viral and squashed into cost 2–4. That is the wrong way round for "NFTs give variety, not power".
5. **`costThresholds` squash fame.** The owned cost mix is 3/23/37/22/10/5. `[26,39,51,63,78]` gives 13/20/22/20/15/10, which hits the target and confirms the critic and designer-a.
6. **Pure extremes are dead under every lever:** heavy (cost 5–6 only) at 15%, swarm (cost ≤ 2) at 2%. Top-heavy (curve 1/1/2/2/3/3) is **57%**, so the §8 top-heavy retarget (≥ 40%) **passes**. Swarm and heavy aren't fixable without breaking curve decks: every lever that helps one hurts the other (sweep below). Treat them as non-decks.
7. **Matches are always 6 turns.** Only 6.5% of decisive games are decided in the first third, and 19.9% by half-time, so most games stay live to the end.
8. **Skill expression is healthy:**
- greedy beats maxpower ("always play the biggest card") 59.5%. §8 wants the naive line under 55% against a smart bot; it is at 40.5%.
- lookahead-lite beats greedy 63.7%.
- greedy beats random 77.5%.
### Proposed v1.1 rule changes (concrete)
| # | Change | Evidence | Status |
|---|---|---|---|
| 1 | `costThresholds: [26, 39, 51, 63, 78]` (designer-a's quantile rule) | owned cost mix 3/23/37/22/10/5 → 13/20/22/20/15/10 | **adopt** |
| 2 | `rising.powerDelta: -1` | rising +1.76 per card, the top keyword | **adopt** |
| 3 | viral: "+1, plus +1 per earlier reveal this turn, max +3" (critic) | viral +0.19; buffing amount/cap alone gives +0.24 | **adopt.** It needs the base +1. |
| 4 | encyclopedic: "+2 if *you* have no other cards here" (critic) | +0.19; amount 3 alone gives +0.33 | **adopt.** It needs the condition change. |
| 5 | Culture takes its home-stat keyword (critic) | owned decks drop from 51% viral to 18% | see the v1.1 results |
| 6 | `nostalgic.powerDelta -1`, `translated.powerDelta -1` | +1.45 / +1.53 per card, close to rising. Without them Field Guide beats Headliners 60% under 1–5 (approximated hooks). | **optional, v1.2**: see the v1.1 re-run above |
| 7 | No change for heavy or swarm | no lever fixes both | – |
| 8 | No change for seat or initiative | 48–52% target met | – |
### Starter round robin and owned vs starters (focused run: §2.1 starters plus 4 owned decks, 2,000 matches per ruleset)
| | v1 | v1.1 critic package (1–5) | **v1.1 critic + item 6** |
|---|---|---|---|
| starter pairings, min–max | **23–77%** | 40–60% | **45–55%** ✅ |
| starter deck WR (FG / Head / Pio / Star) | 65 / 40 / 57 / 54 | 50 / 48 / 54 / 54 | 51 / 55 / 52 / 52 |
| owned vs starters (FG / Head / Pio / Star) | 37 / 54 / 40 / 41 | 52 / 45 / 46 / 43 | 50 / 40 / 46 / 47 |
| first player (turn-1 initiative) | 53.4% [51.0–55.8] | 52.5% [50.1–54.9] | **51.0% [48.6–53.4]** ✅ |
| lookahead-lite > greedy | 60% | 70% | 66% |
Each starter pairing has about 50–58 symmetric games (±13 pp at 95%). Each owned-vs-starter cell has about 200 games (±7 pp).
**Net keyword value under v1.1** (counterfactual trigger value per card, plus `powerDelta`), from the 3,000-match critic-package full field:
| keyword | v1.1 net per card |
|---|---|
| rising | 1.74 − 1 = **+0.74** |
| viral | 0.16 + 1 = **+1.16** |
| encyclopedic | ≈ 0.23 + 1 = **+1.23** (approximated) |
| translated | +1.59, or **+0.59** with item 6 |
| nostalgic | +1.58, or **+0.58** with item 6 |
| local | **+0.66** |
So every keyword lands within 0.6–1.2. This matches the critic's numbers (viral +1.15, encyclopedic +0.9, rising +0.9) to within 0.35.
**Recommendation: ship items 1–5 plus item 6** (`nostalgic.powerDelta -1`, `translated.powerDelta -1`). Without item 6, Field Guide still beats Headliners 60% and Headliners sits at 40–41% against every starter. With it, every pairing is inside 45–55. Owned decks then win 40–50% against the starters: the variety-not-power goal, slightly on the weak side. If v1.2 wants owned decks a bit stronger, the next lever is `local.amount 2` (Headliners and owned decks are about 25% local).
Raw tables:
#### `v1` (DEFAULT_RULESET), focused: 2000 matches, 9 draws, 0 engine errors
- First player (turn-1 initiative), symmetric games: **53.4% [51.0–55.8] (n=1641)**
- Decided early: 4.1% of decisive games locked in during the first third, 15.5% by half-time
- Skill: greedy>maxpower 61.6% (n=146), lookahead-lite>greedy 60.0% (n=100), greedy>random 81.7% (n=104)
| starters (row vs col) | field-guide | headliners | pioneers | stargazers | vs all decks |
|---|---|---|---|---|---|
| field-guide | – | 77% (n=52) | 57% (n=53) | 69% (n=54) | 64.8% |
| headliners | 23% (n=52) | – | 36% (n=56) | 35% (n=54) | 39.6% |
| pioneers | 43% (n=53) | 64% (n=56) | – | 52% (n=58) | 56.9% |
| stargazers | 31% (n=54) | 65% (n=54) | 48% (n=58) | – | 54.3% |
| owned (random 12 from card-sample) vs | field-guide | headliners | pioneers | stargazers | all |
|---|---|---|---|---|---|
| owned-0 | 42% | 48% | 35% | 37% | 40.0% |
| owned-1 | 26% | 62% | 40% | 45% | 47.4% |
| owned-2 | 45% | 57% | 44% | 43% | 52.7% |
| owned-3 | 35% | 47% | 40% | 37% | 42.3% |
| **all owned** | **37%** (n=233) | **54%** (n=197) | **40%** (n=200) | **41%** (n=197) | |
| keyword | power per card | games decided by it | cards played |
|---|---|---|---|
| rising | +1.75 | 12.0% | 1822 |
| translated | +1.55 | 18.6% | 2263 |
| nostalgic | +1.54 | 10.3% | 2197 |
| local | +0.71 | 6.2% | 4300 |
| viral | +0.22 | 3.3% | 6034 |
| encyclopedic | +0.17 | 0.8% | 1042 |
#### `v1.1-critic`: 2000 matches, 4 draws, 0 engine errors
- First player (turn-1 initiative), symmetric games: **52.5% [50.1–54.9] (n=1646)**
- Decided early: 3.5% of decisive games locked in during the first third, 11.4% by half-time
- Skill: greedy>maxpower 59.6% (n=146), lookahead-lite>greedy 70.0% (n=100), greedy>random 80.8% (n=104)
| starters (row vs col) | field-guide | headliners | pioneers | stargazers | vs all decks |
|---|---|---|---|---|---|
| field-guide | – | 60% (n=52) | 51% (n=53) | 51% (n=55) | 50.1% |
| headliners | 40% (n=52) | – | 41% (n=56) | 41% (n=54) | 48.3% |
| pioneers | 49% (n=53) | 59% (n=56) | – | 55% (n=58) | 54.2% |
| stargazers | 49% (n=55) | 59% (n=54) | 45% (n=58) | – | 54.1% |
| owned (random 12 from card-sample) vs | field-guide | headliners | pioneers | stargazers | all |
|---|---|---|---|---|---|
| owned-0 | 46% | 43% | 46% | 49% | 43.8% |
| owned-1 | 50% | 43% | 40% | 34% | 44.1% |
| owned-2 | 52% | 47% | 56% | 51% | 52.7% |
| owned-3 | 63% | 49% | 43% | 42% | 52.4% |
| **all owned** | **52%** (n=233) | **45%** (n=198) | **46%** (n=198) | **43%** (n=199) | |
| keyword | power per card | games decided by it | cards played |
|---|---|---|---|
| rising | +1.71 | 12.8% | 1505 |
| nostalgic | +1.68 | 13.6% | 1789 |
| translated | +1.64 | 17.6% | 4491 |
| local | +0.70 | 6.3% | 4081 |
| encyclopedic | +0.21 | 1.7% | 2177 |
| viral | +0.13 | 0.9% | 2524 |
#### `v1.1-critic-nerf`: 2000 matches, 14 draws, 0 engine errors
- First player (turn-1 initiative), symmetric games: **51.0% [48.6–53.4] (n=1636)**
- Decided early: 4.0% of decisive games locked in during the first third, 12.7% by half-time
- Skill: greedy>maxpower 58.2% (n=146), lookahead-lite>greedy 66.0% (n=100), greedy>random 79.8% (n=104)
| starters (row vs col) | field-guide | headliners | pioneers | stargazers | vs all decks |
|---|---|---|---|---|---|
| field-guide | – | 49% (n=51) | 50% (n=52) | 55% (n=55) | 50.6% |
| headliners | 51% (n=51) | – | 50% (n=54) | 46% (n=54) | 55.3% |
| pioneers | 50% (n=52) | 50% (n=54) | – | 47% (n=58) | 51.8% |
| stargazers | 45% (n=55) | 54% (n=54) | 53% (n=58) | – | 51.9% |
| owned (random 12 from card-sample) vs | field-guide | headliners | pioneers | stargazers | all |
|---|---|---|---|---|---|
| owned-0 | 51% | 43% | 52% | 46% | 46.0% |
| owned-1 | 40% | 35% | 47% | 33% | 43.1% |
| owned-2 | 52% | 42% | 37% | 56% | 51.1% |
| owned-3 | 58% | 38% | 47% | 55% | 50.1% |
| **all owned** | **50%** (n=231) | **40%** (n=197) | **46%** (n=199) | **47%** (n=197) | |
| keyword | power per card | games decided by it | cards played |
|---|---|---|---|
| rising | +1.72 | 15.6% | 1534 |
| translated | +1.69 | 16.3% | 4315 |
| nostalgic | +1.66 | 13.3% | 1686 |
| local | +0.71 | 6.4% | 4113 |
| encyclopedic | +0.21 | 1.8% | 2221 |
| viral | +0.13 | 0.7% | 2571 |
#### Critic package, full field (3,000 matches)
#### `v1.1-critic`: 3000 matches, 14 draws, 0 engine errors
- First player (turn-1 initiative), symmetric games: **49.7% [47.7–51.7] (n=2418)**
- Decided early: 8.0% of decisive games locked in during the first third, 17.9% by half-time
- Skill: greedy>maxpower 54.6% (n=238), lookahead-lite>greedy 62.7% (n=161), greedy>random 84.0% (n=169)
| starters (row vs col) | field-guide | headliners | pioneers | stargazers | vs all decks |
|---|---|---|---|---|---|
| field-guide | – | 38% (n=8) | 50% (n=8) | 29% (n=7) | 60.6% |
| headliners | 62% (n=8) | – | 36% (n=11) | 56% (n=9) | 46.8% |
| pioneers | 50% (n=8) | 64% (n=11) | – | 50% (n=12) | 59.7% |
| stargazers | 71% (n=7) | 44% (n=9) | 50% (n=12) | – | 64.8% |
| owned (random 12 from card-sample) vs | field-guide | headliners | pioneers | stargazers | all |
|---|---|---|---|---|---|
| owned-0 | 33% | 64% | 36% | 50% | 55.3% |
| owned-1 | 50% | 50% | 33% | 43% | 54.9% |
| owned-2 | 55% | 58% | 27% | 42% | 52.7% |
| owned-3 | 11% | 42% | 9% | 50% | 52.8% |
| **all owned** | **38%** (n=37) | **54%** (n=41) | **26%** (n=39) | **46%** (n=37) | |
| archetype | WR vs other decks |
|---|---|
| curve | 58.6% [51.7–65.2] (n=203) |
| top-heavy | 52.2% [45.4–58.9] (n=205) |
| heavy | 23.4% [18.1–29.7] (n=205) |
| swarm | 0.9% [0.2–3.3] (n=220) |
| mono-Culture | 55.4% [48.9–61.6] (n=233) |
| mono-People | 47.5% [41.0–54.1] (n=221) |
| mono-Nature | 40.3% [34.0–46.9] (n=216) |
| keyword | power per card | games decided by it | cards played |
|---|---|---|---|
| rising | +1.74 | 21.1% | 3184 |
| translated | +1.59 | 15.8% | 3450 |
| nostalgic | +1.58 | 19.7% | 2871 |
| local | +0.66 | 7.5% | 4696 |
| encyclopedic | +0.23 | 2.0% | 4548 |
| viral | +0.16 | 1.5% | 4733 |
### v1 full-field detail
#### `v1` (DEFAULT_RULESET): 3000 matches, 13 draws, 0 engine errors
- First player (turn-1 initiative), symmetric games: **51.5% [49.6–53.5] (n=2421)**
- Decided early: 6.5% of decisive games locked in during the first third, 19.9% by half-time
- Skill: greedy>maxpower 59.5% (n=237), lookahead-lite>greedy 63.7% (n=160), greedy>random 77.5% (n=169)
| starters (row vs col) | field-guide | headliners | pioneers | stargazers | vs all decks |
|---|---|---|---|---|---|
| field-guide | – | 70% (n=10) | 57% (n=7) | 83% (n=6) | 67.4% |
| headliners | 30% (n=10) | – | 33% (n=12) | 56% (n=9) | 51.7% |
| pioneers | 43% (n=7) | 67% (n=12) | – | 64% (n=14) | 67.0% |
| stargazers | 17% (n=6) | 44% (n=9) | 36% (n=14) | – | 53.3% |
| owned (random 12 from card-sample) vs | field-guide | headliners | pioneers | stargazers | all |
|---|---|---|---|---|---|
| owned-0 | 29% | 40% | 15% | 58% | 51.0% |
| owned-1 | 33% | 33% | 67% | 56% | 51.6% |
| owned-2 | 27% | 62% | 62% | 25% | 59.7% |
| owned-3 | 43% | 50% | 56% | 70% | 57.7% |
| **all owned** | **31%** (n=42) | **48%** (n=48) | **44%** (n=36) | **51%** (n=43) | |
| archetype | WR vs other decks |
|---|---|
| curve | 50.0% [43.5–56.5] (n=222) |
| top-heavy | 56.8% [50.2–63.1] (n=222) |
| heavy | 15.3% [11.0–20.9] (n=203) |
| swarm | 2.0% [0.9–4.6] (n=250) |
| mono-Culture | 42.2% [35.7–48.9] (n=211) |
| mono-People | 51.7% [44.9–58.4] (n=207) |
| mono-Nature | 56.0% [49.6–62.3] (n=232) |
| keyword | power per card | games decided by it | cards played |
|---|---|---|---|
| rising | +1.76 | 15.6% | 4027 |
| translated | +1.53 | 17.5% | 1266 |
| nostalgic | +1.45 | 16.4% | 3610 |
| local | +0.65 | 6.1% | 6781 |
| viral | +0.19 | 2.3% | 8030 |
| encyclopedic | +0.19 | 1.4% | 1002 |
*Starter-vs-starter cells here have small n; the focused round robin above has the real numbers. Keyword table: counterfactual trigger value per card (excluding any powerDelta).*
### Lever sweeps (greedy mirrors, real pool)
Before keywords were implemented (engine core @ c938407), I swept these levers. Each row is 3,000 greedy-mirror games. Columns are win rate vs the other decks.
| variant | heavy | swarm | curve | first player |
|---|---|---|---|---|
| v1 | 11.9% | 0.9% | 64.5% | 50.5% |
| `energyCarryMax 2` | 17.8% | 0.9% | 59.3% | 51.0% |
| `energyByTurn [2..7]` | **42.4%** | 0.0% | 41.3% | 49.7% |
| `powerByCost [0,2,3,4,6,8,10]` | 8.1% | 13.4% | 65.6% | 49.3% |
| `powerByCost [0,2,4,5,7,8,10]` | 0.3% | **36.1%** | 62.1% | 49.7% |
| `laneCapacity 4` | 10.3% | 0.6% | 63.6% | 50.7% |
| `onTopicBonus [0,2,2,2,2,3,3]` | 10.2% | 6.8% | 62.2% | 49.7% |
**Reading:** energy 2→7 rescues heavy but kills swarm and curve, and flatter power rescues swarm but kills heavy. Neither helps the normal decks. Keep v1's curve.
### Caveats
- Bots aren't people. Greedy doesn't plan energy carry and queues one card at a time; lookahead-lite is shallow.
- Hidden cards in determinizations are sampled from the loaner pool, not from the opponent's real decklist.
- Some critic-package items need engine hooks that don't exist yet, so this run approximates them:
| item | modelled as | exact? |
|---|---|---|
| viral base +1 | `powerDelta +1` | yes |
| encyclopedic "your side only" | `powerDelta +1`, about the critic's measured +0.9 | no: an approximation until engine-b adds `encyclopedic {sides 1}` |
`powerDelta > 0` is a sim stand-in: the engine's derivation clamps `powerDelta` to [−3, 0].
- Owned decks are 4 random 12-card draws, not curated. They have no curve fill, so some are lopsided.
## Method
**Engine.** Every match runs on the real rules engine (`shared/game`, public `index.ts` API only), through
`shared/bots/game-adapter.ts`. Nothing in the sim re-implements rules. The engine commit is named in each
results section.
**Bots** (`shared/bots/index.ts`, each a pure `(state, player, rng) => action`):
- `random`: a uniformly random legal play or commit.
- `maxpower`: always queues the highest printed-power card it can afford, in a random legal lane. This is the
DESIGN §8 "always play the highest-power card" line.
- `greedy`: one-ply. It takes the action whose result scores best on a lane heuristic: a soft lead per Topic
plus total power. The heuristic sees the bot's own queued plays and the opponent's *revealed* cards only.
- `lookahead-lite` / `lookahead`: flat Monte Carlo over the top 3–4 greedy candidates, with 4–8 rollouts
each. Before each rollout, the opponent's queued plays are taken back and its hand and deck reshuffled
(determinization), so the bot never reads hidden information.
**Seats.** Plans are simultaneous: seat 0 queues and commits, then seat 1. Neither bot can see the other's
queue, so the order leaks nothing. "First player" means the seat with **initiative on turn 1**.
**Schedule.** Per ruleset, matches are drawn from a fixed mix of bot pairings with random seats:
| pairing | share | used for |
|---|---|---|
| greedy vs greedy | 55% | balance (decks, cards, keywords, first player) |
| lookahead-lite vs lookahead-lite | 12% | balance with a stronger player |
| random vs random | 5% | baseline |
| greedy vs lookahead-lite, maxpower vs lookahead-lite, maxpower vs greedy, random vs greedy | 28% | skill expression |
Deck, card and keyword statistics use only **bot-symmetric** games (the same bot on both seats). Deck pairs are
drawn uniformly, mirrors included; the per-deck win rate excludes mirrors. `--mix fast` (greedy only) is used for
quick lever sweeps.
**Decks.** These are built from a card pool by `defaultDecks()`:
- 6 seeded starter decks (`starterDeck`)
- `curve` (cost counts 2/3/3/2/1/1)
- `top-heavy` (1/1/2/2/3/3)
- `heavy` (cost 5–6 first: the §8 "all-famous" line)
- `swarm` (cost ≤ 2 first)
- `mono-<faction>` curve decks for factions with ≥ 6 cards
- `kw-<keyword>` curve decks for keywords on ≥ 3 cards
The pool is the engine's preview pool, or council-data's `card-sample.json` derived per DESIGN §1
(percentile → fame → cost, home stat, faction keyword). Loaner power always follows the ruleset's
`powerByCost`.
**Metrics.**
- Win rates with Wilson 95% intervals. **↑/↓** marks an interval that excludes 50%.
- First-player WR: initiative holder on turn 1, symmetric games.
- Match length in turns.
- "Decided early": the turn from which the eventual winner led on the heuristic until the end, as a fraction of
the match.
- Per-card WR: the win rate of the side that played the card.
- Per-stat correlations.
- **Keyword value:** a counterfactual. Each finished match is replayed with the same actions and the keyword
removed from one side. We report the side's lost total power per keyword card, and how often removing the
keyword made the result worse for that side ("games decided by it"). DESIGN §10.1 asks for
`keywordParams.<id>.powerDelta` when a keyword is worth more than about 1 power.
**Reproduce.**
```
SIM_POOL=docs/game/card-sample.json SIM_RULESETS=scripts/sim/variants/v1.json \
npx tsx scripts/sim/run.ts --engine game --matches 5000 --workers 10 --out sim-out
python3 scripts/sim/summary.py sim-out
```
**Limits.**
- Bots are not humans. Greedy undervalues saving energy and bluffing.
- Lookahead rollouts are short.
- Deck archetypes are built automatically, not tuned by hand.
- The 100-card sample is skewed: Culture 53, People 30, Nature 9, History 5, Science 3, Space 0.
+149
View File
@@ -0,0 +1,149 @@
# Card data: real stat distributions (council-data)
Status: **final for tonight**, 236 fully enriched cards, generated 2026-11-16 23:20 EST. Sampling stopped early on night-lead's request because Wikimedia was rate-limiting (429) our shared egress IP. 330 cards were picked; the first 236 finished enrichment.
**Method:** `scripts/night/card_sample.py` draws random ns0 articles in batches of 20 and keeps those that pass the `server/wiki.ts` filter: a thumbnail from upload/thumb.wikimedia.org and an intro extract of at least 40 characters. About 44% of random pages pass. Each card is enriched with the same API parameters as `server/enrichment.ts`:
- six completed months, Jun–Oct 2026; attention uses October
- monthly views summed from the daily REST pageviews
On top of that it adds Wikidata P31 and a 30-day drift (views 10-17..11-15 vs 09-17..10-16).
**Politeness:** the sampler runs serially at 1 request a second or slower, honours Retry-After, and caches every response to disk. `scripts/night/card_analyze.py` produces everything below, plus the `percentiles` (99 cut points each, for DESIGN §1.1) in `card-sample.json`.
## Files
- `card-sample.json`: `cards[]` with the `PageEnrichment` shape, plus `category` (legacy `category()`), `instanceOf` (P31), `faction` (the P31 rules below), `tiers` and `drift30`. Top level: `percentiles`, `tierBreakpoints`, `factionRules` and `note`.
- `starter-cards.json`: the loaner **pool**, 20 real frozen snapshots (a `MintStatsSnapshot` in `snapshot`). It also holds `starterDeckPageIds`: the 48 DESIGN §2.1 starter titles with their real pageIds and `cardId`s.
## Headline findings (236 cards)
1. **Raw scores are compressed.**
| score | median | p10–p90 | max |
|---|---|---|---|
| attention | 32 | 20–53 | 77 |
| depth | 63 | 54–73 | 87 |
| reach | 20 | 0–45 | 70 |
No honest card reaches 100. Use `percentiles` (§1.1) or the 1–10 tiers, never the raw 0–100 values.
2. **Correlations (Spearman):**
- attention and depth: 0.61, so these two double-count fame
- reach is nearly independent of both (0.17 with attention, 0.02 with depth)
3. **Reach has a hard floor:** 18% of cards (42) have 0 other languages, so reach is 0. Under the §1.1 counting, all of them get pct_reach of about 8.
4. **Raw values are long-tailed:**
| value | median | p90 | max |
|---|---|---|---|
| monthly views | 82 | 1,576 | 42k |
| pageBytes | 6.1 KB | 24 KB | |
| languageCount | 3 | 22 | |
5. **Trend is mostly down:**
- 65% of cards have a negative trendPercent, 34% positive; median −9.5%
- 16% swing by more than 50% either way
- null: 1%
6. **Live drift is small in score terms.** The 30-day median |pageview drift| is 22% (p90 60%). That moves attention by only **1.7 points** (p90 6.3), so at-mint and live values differ little unless someone inflates them.
7. **The legacy `category()` split is very skewed:**
| category | share |
|---|---|
| Culture | 56% |
| People | 26% |
| Nature | 8% |
| History | 8% |
| Science | 1% |
| Space | 0.4% |
8. **Candidate factions from P31:**
| faction | share |
|---|---|
| People | 29% |
| Places | 20% |
| Nature | 8% |
| Structures | 7% |
| Works | 5% |
| Events | 5% |
| Organisations | 4% |
| Ideas & Tech | 3% |
| Sport | 3% |
| unclassified | 14% |
Suggested 6 for v2: People, Places, Nature, Works & Culture (Works + Organisations + Sport), Structures, Events & Ideas, plus a neutral fallback.
## Full tables
Sample: **236 cards**; ~44% (149/340 logged during sampling) of random pages pass the illustrated filter. Pageviews available for 236.
| stat | n | min | p10 | p25 | median | p75 | p90 | max | mean | skew |
|---|---|---|---|---|---|---|---|---|---|---|
| attention | 236 | 13 | 20 | 25 | 32 | 41 | 53 | 77 | 35 | 0.78 |
| depth | 236 | 48 | 54 | 58 | 63 | 68 | 73 | 87 | 64 | 0.42 |
| reach | 236 | 0 | 0 | 10 | 20 | 33 | 45 | 70 | 23 | 0.36 |
| pageBytes | 236 | 772 | 1847 | 3084 | 6110 | 12805 | 24412 | 168332 | 11691 | 4.61 |
| languageCount | 236 | 0 | 0 | 1 | 3 | 9 | 22 | 124 | 8 | 4.14 |
| monthlyViews | 236 | 5 | 16 | 30 | 82 | 300 | 1576 | 42025 | 812 | 9.82 |
| dailyAverage | 236 | 0.2 | 0.5 | 1.0 | 2.6 | 9.7 | 50.8 | 1355.6 | 26.2 | 9.82 |
| trendPercent | 234 | -86.6 | -49.3 | -26.3 | -9.5 | 5.3 | 31.2 | 388.9 | -3.6 | 4.11 |
| drift30Percent | 235 | -94.7 | -58.2 | -33.8 | -12.9 | 1.0 | 27.7 | 2268.7 | -4.2 | 14.07 |
**Spearman rank correlations** (cards with pageviews):
| | attention | depth | reach | trendPercent | drift30Percent |
|---|---|---|---|---|---|
| attention | 1.00 | 0.61 | 0.17 | 0.35 | 0.33 |
| depth | 0.61 | 1.00 | 0.02 | 0.15 | 0.14 |
| reach | 0.17 | 0.02 | 1.00 | 0.00 | 0.09 |
| trendPercent | 0.35 | 0.15 | 0.00 | 1.00 | 0.31 |
| drift30Percent | 0.33 | 0.14 | 0.09 | 0.31 | 1.00 |
(n=234; attention~monthlyViews, depth~pageBytes, reach~languageCount are monotonic so their rank correlations are identical.)
**trendPercent sign:** positive 79 (34%), negative 153 (65%), zero 2; null (prior month 0 or no views) 2. |trend| > 50%: 38; > 100%: 7.
**30-day drift** (views 10-17..11-15 vs 09-17..10-16): median |drift| 22%, p90 |drift| 60%. Attention score change implied: median 1.7 points, p90 6.3 points.
**attention histogram (deciles 0-9,10-19..90-100):** 0 19 74 74 36 20 12 1 0 0
**depth histogram (deciles 0-9,10-19..90-100):** 0 0 0 0 2 72 117 38 7 0
**reach histogram (deciles 0-9,10-19..90-100):** 42 60 53 43 21 15 1 1 0 0
Reach = 0 (no other languages): 42; reach 100 (>=999 langs) impossible in practice, max langs 124. Attention null: 0; attention 0: 0.
**Sum of three stats:** min 73, p10 89, p25 102, median 116, p75 135, p90 156, max 222
**Legacy `category()` (server/wiki.ts):** Culture 133, People 62, Nature 19, History 18, Science 3, Space 1
**Top Wikidata P31 classes:** human 64, taxon 16, village 7, commune of France 6, river 5, human settlement 4, musical group 3, building 3, gene 3, village of Poland 2, civil parish 2, protest 2, legislative election 2, castle 2, events in a specific year or time period 2, township of Pennsylvania 2, film 2, company 2, island 2, rural municipality of Austria 1, municipality of Austria 1, municipality of Brazil 1, television film 1, nor'easter 1, rail company (Japan) 1, municipality 1, association football match 1, final 1, local government area of Victoria 1, plant disease 1, basketball team season 1, mosque 1, functional programming language 1, purely functional programming language 1, dependently typed programming language 1, programming language 1, free software 1, proof assistant 1, municipality of Croatia 1, municipality of Bangladesh 1
No Wikidata item: 0; no P31: 13
**Most frequent category words (cards containing):** stubs 81, births 55, people 50, from 50, century 46, living 33, geography 30, american 29, deaths 28, united 27, county 27, states 23, established 21, establishments 20, alumni 18, places 17, university 16, buildings 16, state 15, national 14, structures 14, male 13, players 12, district 11, villages 11, sportsmen 11, members 11, described 10, football 10, sports 10, city 10, film 10, british 9, companies 9, populated 9, sportspeople 9, biography 9, york 9, women 8, taxa 8, germany 8, politicians 8, television 7, writers 7, french 7, genera 7, based 7, communes 7, language 7, expatriate 7, named 7, olympics 7, completed 7, historic 7, register 7, faculty 6, pennsylvania 6, civil 6, history 6, california 6
## Rank-normalised tiers (1–10)
Tier = 1 + number of breakpoints the score is **>=**. Breakpoints are the sample deciles p10..p90 of the 0–100 score. Ties make some tiers thin (scores are integers).
| stat | p10 | p20 | p30 | p40 | p50 | p60 | p70 | p80 | p90 |
|---|---|---|---|---|---|---|---|---|---|
| attention | 20 | 24 | 27 | 30 | 32 | 35 | 39 | 44 | 53 |
| depth | 54 | 57 | 59 | 61 | 63 | 65 | 67 | 69 | 73 |
| reach | 0 | 10 | 10 | 16 | 20 | 26 | 30 | 36 | 45 |
Resulting tier counts — attention: None:0 1:19 2:27 3:24 4:23 5:23 6:17 7:28 8:25 9:22 10:28
; depth: 1:14 2:23 3:26 4:21 5:25 6:26 7:22 8:21 9:30 10:28
; reach: 1:0 2:42 3:0 4:30 5:30 6:32 7:21 8:27 9:27 10:27
## Candidate factions (Wikidata P31 rules)
| faction | n | share | median attention | median depth | median reach | example |
|---|---|---|---|---|---|---|
| People | 68 | 29% | 31 | 65 | 22 | Salahuddin of Aceh |
| Places | 48 | 20% | 28 | 61 | 32 | St. Georgen am Längsee |
| Other | 34 | 14% | 40 | 65 | 16 | Village (Pennsylvania) |
| Nature | 20 | 8% | 30 | 56 | 29 | Dichromorpha viridis |
| Structures | 17 | 7% | 26 | 57 | 10 | Pathrail Mosque |
| Works | 12 | 5% | 36 | 63 | 13 | The Best of Men |
| Events | 12 | 5% | 40 | 72 | 13 | April 2021 nor'easter |
| Organisations | 10 | 4% | 48 | 68 | 10 | Hiroshima Electric Railway |
| Ideas & Tech | 8 | 3% | 30 | 63 | 16 | Agda (programming language) |
| Sport | 7 | 3% | 27 | 70 | 16 | 1912 Argentine Primera División final |
Unclassified P31 labels: typeface 1, annexation 1, position 1, family 1, historic site 1, automobile manufacturer 1, historical society 1, stock character 1, pigeon breed 1, shopping center 1, Wikimedia list article 1, clinical sign 1, tram route 1, rubber-tyred tram 1, oil refinery 1, waterfall 1, orisha 1, water deity 1, Loa 1, arch 1, award 1, sex scandal 1, military base 1, artillery battery 1
+673
View File
@@ -0,0 +1,673 @@
# WikiDeck: Edit War — game design v1 (draft)
**Status:** v1 DRAFT, freeze target 23:30 EST 2026-11-16.
**Owners:** council-designer-a owns sections 0–9 (research, card derivation, rules). council-designer-b
owns sections 10–12 (keywords, peer-hosted modes, worked example). council-critic reviews.
Every number in **`code font`** is a rules-config key (see §9). Modes can override any of them.
A two-player lane game for 3–6 minute matches. Each player has a 12-card deck and there are 3
**Topics** (lanes) with 3 slots per side. There are 6 turns, both players plan at the same time, and
energy ramps from 1 to 6, with 1 unspent energy carried over. Reveals interleave one card at a time.
Win 2 of the 3 Topics. Each Topic is judged on one Wikipedia stat (**Attention**, **Depth** or
**Reach**), and a card is stronger in a Topic that matches its best stat. Fame (how high a card's
stats are overall) sets its **cost**, not its efficiency, so famous cards are sidegrades. They are
not upgrades.
---
## 0. Research summary (what we took, what we left)
| Game | Why it's fun | Taken | Left |
|---|---|---|---|
| Marvel Snap | 3-minute matches; simultaneous turns; 3 lanes, so the decision is *where*, not just *what* | lanes, about 6 turns, simultaneous planning, small deck | stakes betting (modes only); we deliberately diverge in lane size (3), energy carry-over, interleaved alternating reveals, stats visible from turn 1 with hidden Portals, and on-topic tie-breaks |
| Hearthstone | mana curve; readable cards; tempo | energy = turn number, curve-shaped costs | 30-card decks and 15-minute matches; hero powers |
| Magic: The Gathering | factions (colours) with distinct identities; deckbuilding depth | 6 factions with their own keyword flavour | land screw (resource variance); the stack; long text |
| Balatro | huge "moment" when a combo goes off; simple base rules plus modifiers | lane modifiers (Portals) as the run-to-run variety | single-player scoring |
| Slay the Spire | readable intents; every card is a choice | open information about lane state and power | roguelike runs (could be a mode) |
| Gwent | rows; best 2 of 3; passing and bluffing | best 2 of 3 Topics; hidden commitments | multi-round card advantage (too long) |
| Pokémon TCG | type advantage | on-topic bonus (card stat matches lane stat) | energy attachment; coin flips |
| Top Trumps | instantly understood: "my number beats yours" | lanes judged on one stat | picking the stat is solved play (always pick your best) |
| Dominion | deck building as play | — (too long for 3–6 min) | — |
**What is ours, not borrowed:**
- (a) each Topic is judged on one Wikipedia stat (the Top Trumps twist)
- (b) cost comes from an article's fame instead of being printed
- (c) a trend arrow driven by live pageviews
- (d) a total-power and consensus tie-break instead of stakes betting
- smaller choices: 3 slots per lane, 1 carried-over energy, interleaved reveals, and stats visible
from turn 1 with hidden Portals
The core loop is a lanes game in the style of Snap. The Top Trumps twist is that each lane asks for
one Wikipedia stat. That makes every stat matter and gives each turn its main decision: play the card
on-topic for the bonus, or put it where the fight is.
---
## 1. Card derivation (card data → game card)
The engine never reads raw 0–100 stats as power. Every game card is derived **deterministically** from a
`CardInput` that the match host freezes at match start:
```
CardInput = { cardId, pageId, title, category, rarity,
snapshot: MintStatsSnapshot, // at mint, immutable
live: { trendPercent: number | null, dailyAverage: number | null }, // frozen at match start
check: 'verified' | 'unverified' | 'unchecked' | 'loaner', // set by the host (§1.10)
pinned?: { cost, home, faction } } // loaners only (§2.1): replaces §1.1–1.5
```
### 1.1 Percentiles (rank-normalisation)
For each stat `s` in {attention, depth, reach}, `pct_s` = the percentile (0–100, integer) of the card's
snapshot score in the reference table `percentiles.<s>`. That table holds 99 ascending cut points from
council-data's `docs/game/card-sample.json`.
`pct_s` = the number of cut points strictly below the score, plus half the number equal to it, rounded
down. A `null` attention gives `pct_attention = null`.
### 1.2 Fame → Cost
`fame` = the integer mean (rounded half up) of the non-null percentiles.
`cost` = 1 + the number of thresholds in `costThresholds` (5 ascending values, default
`[26, 39, 51, 63, 78]` in v1.1) that are ≤ `fame`. So cost runs 1–6.
**Thresholds are data, not guesses (v1.1).** `fame` is a *mean* of up to 3 percentiles, so it clusters
around 50 even though each percentile is uniform. The v1.0 values `[15, 35, 55, 75, 90]` treated fame as
uniform, so 82% of the pool landed on cost 2–4 (3/23/37/22/10/5 %). `costThresholds` is therefore set to
the fame values at the **15/35/55/75/90 % points of the reference sample** (the same `card-sample.json`
whose `percentiles` table §1.1 uses). That's the integer set that brings the sample's cost spread closest
to the target 15/20/20/20/15/10 %. The v1.1 defaults come from the 236-card sample (fame 8–98) and
give 13/20/22/20/15/10 %. Whenever council-data refreshes the sample, the percentile table and
`costThresholds` are regenerated together and ship as one change. Pinned loaners (§2.1) are unaffected.
### 1.3 Power
`power` = `powerByCost[cost]`, default `[_, 1, 2, 4, 6, 8, 10]` (cost 1 → 1 … cost 6 → 10). The
critic's sim showed that heavy decks were too weak with `[…5,7,9]`.
Power depends **only on cost**. Two cards of the same cost have the same base power, whatever their
fame. Inflating stats (forged snapshots, template shopping) only moves a card up the curve. It never
makes a card more efficient.
### 1.4 Home stat
`home` = the stat with the highest `pct_s`. Ties are broken in this order: depth, reach, attention.
Depth and reach are the hardest to inflate. A `null` stat can never be home.
When a card is in a Topic whose stat equals its `home`, it gets **+`onTopicBonus[cost]`** power
(default `[0, 1, 1, 2, 2, 3, 3]`, i.e. ⌈cost/2⌉).
On-topic efficiency is then about 2 power per energy at every cost (2.0 / 1.5 / 2.0 / 1.75 / 2.0 / 2.0).
Off-topic, cheap cards are weaker (1.0) than expensive ones (1.5). Cheap cards want their Topic, and
big cards are flexible.
### 1.5 Faction
`faction` = the card's `category` (Space, Nature, People, Science, History, Culture), from
`server/wiki.ts`. Culture is about half the pool, so faction keywords are written so that a
mono-faction deck gains no consistency edge (§10: no v1 keyword scales with faction count).
### 1.6 Trend (live, small, capped)
`trend` = +1 if `live.trendPercent ≥ trendUp` (default 25), −1 if `≤ trendDown` (default **−200**,
which means there is **no down arrow in v1**: 69% of cards trend negative, so a ▼ would mostly land as
a quiet penalty on famous owned cards; modes can turn it back on), otherwise 0. A `null` value gives 0. Trend also needs `live.dailyAverage ≥ trendMinDaily` (default
**20**), because percentages on tiny articles are noise; otherwise trend is 0. The host computes it **once at match start**, and it is frozen for
the match. It shows as an arrow (▲), not a number. It adds to power in every Topic. *Why no ▼ by default:* 69% of the sample has a
negative `trendPercent`, and only busy (mostly owned, famous) cards pass `trendMinDaily`, so a ▼ would quietly penalise collectors.
Modes may turn it back on with e.g. `trendDown: -25`.
### 1.7 Rarity
The hash-derived `rarity` (`shared/protocol.ts:38`) is **cosmetic in v1** (frame colour only). Mining luck must not
become power or keyword choice.
### 1.8 Keyword
Each card has at most **one** keyword: cost-1 cards are vanilla, and the others get their faction's keyword (§10.1).
### 1.9 Card face (phone)
The card face shows **cost** (top-left), **power** (big, centre; the bonus shows as a "+N" badge on the home icon when the card is on-topic), a **home-stat icon**
(📣 attention / 📚 depth / 🌐 reach), an optional trend arrow, and one keyword word. That is two
numbers, plus the live lane power when the card is on the board.
### 1.10 Plausibility (anti-forgery, off-chain)
When `verifyStats` is true (the default for hosted matches), the match host checks each owned card's
snapshot before the match:
`pageBytes` against that revision's byte size, `languageCount` ≤ the live count + 5, and views for the
snapshot month within ±10 % of the pageviews API.
A card that fails is played with its live values. The **host replaces `snapshot`** with a snapshot
built from the live values (`deriveStats` on the live bytes, languages and views) **before** it builds
the `CardInput`. The match log marks the card `unverified`.
The engine does no I/O and only ever sees one version of each card.
- **Caching:** results are cached per `(pageId, revisionId, snapshotMonth)`. They are immutable, so a
cached pass or fail never expires. Live values for a failed card are cached for 24 h.
- **Card can't be verified** (the pageId or revision doesn't exist, or the title or category doesn't
match): the card is **rejected at deck validation**, and the player is asked to swap in a loaner.
- **API failure:** if Wikipedia can't be reached and nothing is cached, the card plays from its snapshot
with `check: 'unchecked'`. The engine then caps its cost at `uncheckedCostCap` (default **4**), and
power follows the capped cost. Matches are never blocked.
- The engine's only use of `check` is that cost cap. Loaner cards are `check: 'loaner'`.
---
## 2. Decks
- **Deck size:** exactly `deckSize` cards (default **12**).
- **Copies:** at most `maxCopiesPerPage` cards with the same `pageId` (default **2**). Owned and loaner
copies count **together**. Each copy keeps its own snapshot, so the two copies may differ in cost.
- **Deck validation** (engine): the size is right, the copy limit holds, every card's `check` is
valid, and there are no more loaners than `maxLoaners`. An invalid deck can't start a match.
- **Owned cards:** any card the player's wallet owns at match start. Ownership is read off-chain. The
match never moves, locks or burns a card.
- **Loaner cards:** off-chain preview cards with a fixed `CardInput`, shipped with the game data
(`docs/game/starter-cards.json`, picked by council-data: about 24 pageIds across all 6 categories,
with frozen, hand-checked snapshot stats, including the 4 cards in `shared/catalog.ts`).
Loaners follow the same curve (§2.1 pins their cost, home and faction), so they are no weaker or stronger than owned cards.
- **Starter decks:** `starterDecks` (default: 4 fixed 12-card loaner decks). Each one has two leading
factions and an average cost of 3.0–3.3, and is **curve-shaped**: at least 3 cards of cost 4 or more,
and at most 4 of cost 1. A new player picks one and can play at once.
- **Mixed decks:** an owned deck may fill any slots with loaners, up to `maxLoaners` (default 12, which
means no limit). Owning NFTs gives **variety**, not power: owned and loaner cards follow the same
curve.
### 2.1 Starter decks v1 *(owner: council-designer-b)*
These are the default `starterDecks`: four 12-card loaner decks.
- **Every deck has the same curve:** costs `1,1,2,2,3,3,3,4,4,5,5,6`. That's an average of 3.25, with 5 cards of cost 4 or more and 2 of cost 1.
- **Every deck has the same shape:** 6 cards from each of two factions, and 4 cards for each home stat. So every deck can play on-topic in every lane.
- **Loaner values are pinned.** Cost, home and faction are hand-set (not derived from `starter-cards.json`), so a loaner carries
`pinned: { cost, home, faction }`. For `check: 'loaner'` the engine takes these instead of §1.1–1.5.
Everything after them still applies: `powerByCost`, `onTopicBonus`, `vanillaBonus`, the §10.1 keyword choice and `keywordParams`.
A mode that changes those numbers changes loaners exactly as it changes owned cards.
- **Pinned cost isn't fame.** Cost was set by hand to fit the curve, not from the article's fame. Rarity is `Common` (frame only) and trend is 0.
- **Loaner ids:** `cardId` = `loaner:<kebab-title>`, and `pageId` = the article's real pageId (v1 pins the 4 `shared/catalog.ts` ones: Moon 19331, Periodic table 23053, Octopus 22780,
Albert Einstein 736; the rest are filled in from one batched API call in v1.1, `null` until then). The copy limit counts per `pageId`
together with owned copies (§2); a loaner with `pageId: null` counts per `cardId`.
- **Relation to `docs/game/starter-cards.json`** (council-data, real frozen snapshots of random articles): that file is the
**loaner pool** for the sim and for "fill a slot with a loaner" in the deck builder. It is **not** the default starter decks.
The decks below pin their values on purpose, for two reasons. Famous, recognisable articles would all derive to cost 5–6.
And the random sample has no Space cards (CARD_DATA.md finding 6).
- **v1.1 note (factions):** CARD_DATA.md finding 6 shows the legacy `category()` split as Culture 53%, People 30%, Nature 9%,
History 5%, Science 3%, Space 0%. In v1, owned decks were therefore about 50% viral and 30% local. Viral was the weakest keyword, so a typical owned deck
may be *weaker* than the starter decks. That's the wrong way round for "NFTs give variety, not power". v1 ships as is.
For v1.1, sim-balance runs a **realistic owned deck** (12 cards drawn from `card-sample.json` with its real category mix)
against each starter. **v1.1 fix:** viral gets a base of +1, and Culture cards take a keyword by home stat (§10.1), so owned decks
now have all six keywords in the mix. **BALANCE.md v1.1:** starters 52–53% overall, and owned vs starters 42–51%
(weakest vs Headliners and Stargazers; v1.2 candidates: `nostalgic`/`translated` `powerDelta -1`, `local.amount 2`). The P31-based factions council-data proposes (People, Places, Nature, Works & Culture,
Structures, Events & Ideas) are the v2 fix, and would need the host to fetch P31 (it isn't in `PageEnrichment`).
Keywords follow from §10.1: cost 1 → vanilla, otherwise the faction keyword (shown in brackets).
**A. Stargazers: Space (translated) + Science (encyclopedic)**
| Cost | Card | Faction | Home |
|---|---|---|---|
| 1 | Moon | Space | reach |
| 1 | Electron | Science | depth |
| 2 | Mars | Space | attention |
| 2 | Periodic table | Science | depth |
| 3 | Saturn | Space | reach |
| 3 | Photosynthesis | Science | attention |
| 3 | Black hole | Space | depth |
| 4 | Quantum mechanics | Science | reach |
| 4 | Milky Way | Space | attention |
| 5 | DNA | Science | attention |
| 5 | Sun | Space | reach |
| 6 | Theory of relativity | Science | depth |
**B. Field Guide: Nature (rising) + History (nostalgic)**
| Cost | Card | Faction | Home |
|---|---|---|---|
| 1 | Octopus | Nature | depth |
| 1 | Stonehenge | History | attention |
| 2 | Axolotl | Nature | attention |
| 2 | Rosetta Stone | History | reach |
| 3 | Coral reef | Nature | depth |
| 3 | Roman Empire | History | reach |
| 3 | Honey bee | Nature | attention |
| 4 | Ancient Egypt | History | depth |
| 4 | Amazon rainforest | Nature | reach |
| 5 | Blue whale | Nature | attention |
| 5 | Great Wall of China | History | reach |
| 6 | World War II | History | depth |
**C. Headliners: People (local) + Culture (by home stat: attention → viral, depth → encyclopedic, reach → translated)**
| Cost | Card | Faction | Home |
|---|---|---|---|
| 1 | Haiku | Culture | attention |
| 1 | Ada Lovelace | People | depth |
| 2 | Jazz | Culture | attention |
| 2 | Marie Curie | People | reach |
| 3 | Mona Lisa | Culture | reach |
| 3 | Frida Kahlo | People | attention |
| 3 | Chess | Culture | depth |
| 4 | Leonardo da Vinci | People | reach |
| 4 | The Beatles | Culture | attention |
| 5 | William Shakespeare | People | depth |
| 5 | Olympic Games | Culture | reach |
| 6 | Wolfgang Amadeus Mozart | People | depth |
**D. Pioneers: People (local) + Space (translated)**
| Cost | Card | Faction | Home |
|---|---|---|---|
| 1 | Sputnik 1 | Space | reach |
| 1 | Nikola Tesla | People | attention |
| 2 | Hubble Space Telescope | Space | depth |
| 2 | Isaac Newton | People | reach |
| 3 | Neil Armstrong | People | attention |
| 3 | Jupiter | Space | reach |
| 3 | Galileo Galilei | People | depth |
| 4 | Albert Einstein | People | attention |
| 4 | International Space Station | Space | depth |
| 5 | Apollo 11 | Space | attention |
| 5 | Charles Darwin | People | reach |
| 6 | Solar System | Space | depth |
- **Coverage:** all 6 factions appear, and all 4 `shared/catalog.ts` cards are included (Moon, Periodic table, Octopus, Albert Einstein).
- **Tutorial:** Field Guide vs Stargazers, seed 1, against the easy bot.
- **sim-balance:** please run the 4 decks round robin first. The target is every pairing at 45–55%. If a pairing misses, retune by
swapping home stats within a deck, never by changing the curve.
---
## 3. Board
- 3 **Topics** (lanes), numbered 1–3 from left to right (top to bottom in portrait).
- Each Topic has a **stat** (attention, depth or reach) and an optional **Portal** modifier (§10.3).
Topics have **no faction**. No v1 keyword counts factions: faction only picks the keyword.
- Each player has at most `laneCapacity` cards per Topic (default **3**). That gives 9 slots per side,
which caps cheap-card flooding and fits 3 cards across a 360 px phone.
- **Topic pool:** `topicPool` is a list of Topics, each `{ id, name, stat, portal | null }`. The default
pool has 9 Topics: 3 per stat, one plain and two with Portals. `pool_s` is the sub-list with stat
`s`, in file order.
- **Topic draw (exact):**
- If `distinctStats` is true (the default): for each stat in the fixed order
`[attention, depth, reach]`, pick `pool_s[floor(next() × len(pool_s))]`. That uses 3 `next()`
calls. Then Fisher–Yates the 3 picks into lanes 1–3 (i = 2, 1; j = floor(next() × (i+1))), which
uses 2 calls.
- If `distinctStats` is false: Fisher–Yates the whole pool (i from n−1 down to 1) and take the first
`topicCount` entries as lanes 1–3.
- **Stats visible, Portals hidden:** every Topic's **stat** is face-up from turn 1. Its **Portal**
modifier is face-down until the start of turn `portalReveal[i]` (default `[2, 3, 4]`). You always
know where your card is on-topic, but not what twist a lane will get.
---
## 4. Match setup (deterministic)
1. The input is `{ seed: uint32, rules, players: [p0, p1] }`, and each player's input is
`{ id, deck: CardInput[12] }`. Seat order is input order.
2. RNG: **mulberry32(seed)**. Every random choice draws `next()` in this fixed order:
1. shuffle p0's deck (Fisher–Yates, i from n−1 down to 1, j = floor(next() × (i+1)))
2. shuffle p1's deck the same way
3. draw and shuffle the Topics
4. the initiative coin: p0 has initiative on odd turns if next() < 0.5, otherwise p1
5. later random events (keywords only), in resolution order
3. Each player draws `openingHand` cards (default **3**). There are no mulligans in v1
(`mulligan: false`).
4. **Seed fairness (hosted PvP):** commit–reveal. Each party *i* (host, p0, p1) sends
`sha256(r_i)` for a random 32-byte `r_i`, then reveals `r_i` once all commits are in.
`seed` = the first 4 bytes (big-endian uint32) of `sha256(r_host ‖ r_p0 ‖ r_p1)`.
The host still relays the hidden plans, so hosted PvP trusts the host not to leak them.
Local and bot matches use any seed.
---
## 5. Turn structure
The match lasts `turns` turns (default **6**, hard cap). Each turn has these phases:
1. **Start**
- Reveal any Portal scheduled for turn *t* (resolve its "on reveal").
- Each player's energy = `energyByTurn[t-1]` + carry + `bonusEnergy`. Carry is
min(`energyCarryMax`, unspent energy from last turn), default max **1**. Saving one energy on
turn 2 makes a turn-3 four-drop possible: a real tempo choice.
- Each player draws 1 card if their deck isn't empty and their hand holds fewer than `handLimit`
(default **7**) cards. Otherwise they don't draw. This includes turn 1, so the turn-1 hand is
3 + 1 = 4 cards.
- Order within Start: Portal reveal, then energy, then draw. The Underdog check (step 2) comes after
all three.
2. **Underdog check** (mode option, `underdog: false` by default, because it can be sandbagged). When
it's on, from turn 2, a player who leads in 0 Topics while the opponent leads in ≥ 2 gets
**+1 energy** this turn.
3. **Plan** (simultaneous, hidden)
- Each player queues an ordered list of plays `(handIndex, topic)`.
- Total cost must be ≤ their energy, and each Topic needs room on that player's side.
- Plays can be undone freely until **Commit**. Passing is an empty commit.
- **Hidden means nothing leaks.** While you plan, the opponent's hand count and energy don't change
visibly. Only the bare fact "opponent is ready" (`committed`) is shown. No queue count, no spent energy.
- The turn timer is `turnSeconds` (default **25**). On timeout, whatever is queued is committed.
4. **Reveal (interleaved)**
- Plays reveal **one at a time, alternating**: the initiative player's 1st, the other player's
1st, the initiative player's 2nd, and so on. When one player runs out, the rest of the other's
reveal in order.
- Each card enters its Topic and its "on reveal" effect resolves at once.
- So the order in which you queue cards matters: lead with the card you want resolved first.
- Energy is reserved when a card is queued. A queued play that is no longer legal at reveal (the
lane is full, or a Portal rejects it) **fizzles**: the card stays in hand and its energy is lost.
5. **End**
- Recompute all Topic powers (§6).
- End-of-turn effects resolve, the initiative player's first.
- Initiative strictly **alternates** each turn (set by the coin in §4).
- Plans are locked before anything reveals, so initiative gives away no information. It only
decides the order of on-reveal effects. §10 includes keywords where that order matters.
A typical turn is: look at the lanes, tap a card, tap a Topic (×1–3), tap Commit. That is under 10
seconds, done with one thumb, with no dragging.
---
## 6. Power and Topic control
- A card's **lane power** = `power` + (`onTopicBonus[cost]` if its home stat is the Topic's stat) + `trend`
+ keyword and Portal modifiers. It is never below 0.
- A player's **Topic power** = the sum of their cards' lane power in that Topic.
- A player **leads** a Topic if their Topic power is strictly greater than the opponent's.
- **Consensus tie-break:** if the powers are equal and > 0, the player with more **on-topic** cards
there leads. If that is also equal, nobody leads. 0–0 has no leader.
---
## 7. Win conditions and tie-breaks
**Scoring** happens after the last turn's End phase (§5.5): every end-of-turn effect, including
the last turn's, and every "at scoring" Portal resolves **before** Topics are judged. Then:
1. The player who leads **more Topics** wins (2–1, 2–0, 1–0 and 3–0 all count).
2. If they lead the same number of Topics: the higher **total power across all Topics** wins.
3. If that is also equal: the match is a **draw**.
**Concede:** a player can concede at any time, and the opponent wins at once. A player who misses
`afkTurns` timeouts in a row (default **2**) with nothing queued concedes.
---
## 8. Degenerate lines and why they fail (for the critic)
| Line | Why it isn't dominant |
|---|---|
| All-famous (all max-stat) deck | It becomes an all-cost-5/6 deck with four dead early turns. Cheap cards are more efficient **per energy**, expensive cards **per card**, and only 9 cards are seen per match (3 + 6 draws). So a pure cost-5/6 deck is **weak by design**: **BALANCE.md v1: 15%**, and no lever (power curve, energy ramp) fixes it without breaking curve or swarm decks. There is no hard deck rule; the deck builder warns below 4 cards of cost ≤ 2, and every starter is curve-shaped (§2.1). **Target met:** a **top-heavy curve** (half the deck at cost 4 or more) wins **57%** (target ≥ 40%). `energyByTurn [2..7]` was rejected for v1.1: it lifts heavy decks but kills swarm (20%). |
| All-cheap flood (cost 1–2) | Weak by design: only 9 slots and 9 cards seen. **BALANCE.md v1: 2%** (critic's sim 7–11%). Every lever that revives it breaks curve decks (sweep in BALANCE.md). viral (base +1, v1.1) is its consolation, not its fix. |
| All one faction | No v1 keyword scales with faction count. `factionSynergyCap` is reserved for modes. Culture, the largest pool, splits over three keywords by home stat (§10.1), so it has no single keyword to stack. |
| Owning NFTs buys power (pay-to-win) | Power depends only on cost (§1.3), and cost follows a fixed curve. **BALANCE.md v1.1 re-run (real `RULESET_V1_1`):** each starter deck averages **52–53%** overall. A realistic owned deck from `card-sample.json` wins **42–51%** against the starters (n ≈ 55 per pairing, ±13 pp). So owning cards buys variety, not power. v1.2 levers if owned decks should be a little stronger: `local.amount 2`, and `nostalgic`/`translated` `powerDelta -1` (the two strongest keywords, about +1.65 net per card). |
| "Always play the highest-power card" | It ignores the on-topic bonus, lane capacity and the hidden Portals. **BALANCE.md v1:** the naive max-power line wins **40%** against greedy, and lookahead-lite beats greedy **60–70%**, so skill shows (target < 55%). Critic's sim: 18–21% against a smart bot. |
| Stall or pass | Passing leaves lanes to any 1-power card. Two passing players draw. |
| Mirror the AI | Hidden simultaneous commits: you only see the opponent's cards after you commit. |
| Seat or initiative | Planning is simultaneous and initiative strictly alternates (each seat has it for 3 of the 6 turns). **BALANCE.md v1.1 re-run (real `RULESET_V1_1`): first player 51.2% [48.8–53.6]** (target 48–52%; v1 was 53.4%). |
| Forged or shopped stats | They only raise cost (§1.3), the host check falls back to live stats (§1.10), and trend is capped at ±1. |
| Inflated live pageviews | Worth at most +1, frozen at match start, and only counts for articles with ≥ 20 views a day. |
| Swinginess | No effect may change a lane by more than 6 power or move cards between lanes in v1 (a keyword-design rule for §10). |
---
## 9. Rules config (all defaults)
The engine takes a `RulesConfig` object. Peer-hosted modes override it (§11). Defaults:
```json
{
"version": "1.1",
"deckSize": 12, "maxCopiesPerPage": 2, "maxLoaners": 12,
"openingHand": 3, "handLimit": 7, "mulligan": false,
"turns": 6, "turnSeconds": 25, "afkTurns": 2,
"energyByTurn": [1, 2, 3, 4, 5, 6], "energyCarryMax": 1, "bonusEnergy": 0, "underdog": false,
"topicCount": 3, "distinctStats": true, "portalReveal": [2, 3, 4], "laneCapacity": 3,
"topicPool": [ /* inline Topic objects {id, name, stat, portal}; default = the 9 in §10.3 */ ],
"keywordParams": { /* per-keyword numbers, keyed by keyword id (§10) */ },
"portalParams": { /* per-Portal numbers, keyed by Topic id (§10.3) */ },
"costThresholds": [26, 39, 51, 63, 78],
"powerByCost": [0, 1, 2, 4, 6, 8, 10],
"onTopicBonus": [0, 1, 1, 2, 2, 3, 3], "vanillaBonus": 1,
"keywordsEnabled": ["viral","rising","local","encyclopedic","nostalgic","translated"], "cultureKeyword": "home", "ownedCards": true,
"trendUp": 25, "trendDown": -200, "trendMinDaily": 20, "uncheckedCostCap": 4,
"factionSynergyCap": 3,
"verifyStats": true,
"percentiles": "card-sample-2026-11-16",
"starterDecks": "default"
}
```
**What modes may vary:** every key above, plus `topicPool` contents and the starter-deck lists.
**Fixed for all modes:** matches are off-chain and never move or burn NFTs; the RNG and its draw order
(§4); the shape of the derivation (§1: percentile → fame → cost → power); at most one keyword per card;
the phases of a turn (§5).
A mode must keep `turns × turnSeconds ≤ 240` so that matches stay within 3–6 minutes.
---
## 10. Keywords *(owner: council-designer-b)*
**The v1 keyword set is these six, one per faction.** The ids are the ones engine-b already uses on `night/engine/keywords`.
Topics carry only a **stat**, never a faction. Every keyword below that mentions a Topic keys off stats or board state.
### 10.1 Which keyword a card gets
First rule that matches:
1. `cost == 1` → **no keyword** (vanilla). The card gets +`vanillaBonus` power (default **1**), so a cost-1 card is 2.
2. `"edition"` is in `keywordsEnabled` (off by default) and the deck holds another card with the same `pageId` → **edition**.
3. If the faction is **Culture** and `cultureKeyword` is `"home"` (v1.1 default) → the keyword for its **home stat**:
attention → **viral**, depth → **encyclopedic**, reach → **translated**. (`"faction"` gives v1 behaviour: every Culture card is viral.)
4. Otherwise → the card's **faction keyword** from 10.2.
In steps 3–4, if the chosen keyword isn't in `keywordsEnabled`, the card is vanilla, *without* `vanillaBonus`.
*Why step 3:* Culture is about half of all owned cards. Without it, owned decks are half viral and see only 2 of the 6 keywords.
Rarity plays no part here (it's cosmetic, §1.7). Keywords don't lower power, and v1 has no per-keyword power deltas.
If sim shows that a keyword is worth more than about 1 power on average, sim-balance sets `keywordParams.<id>.powerDelta` (default 0, range −3 to 0).
### 10.2 Keyword list (v1)
| Faction | id (card word) | Trigger | Reminder text | Cap | Order-sensitive? |
|---|---|---|---|---|---|
| Culture | **viral** | On reveal | +1 power, plus +1 for each other card you revealed earlier this turn. | +3 | yes: your queue order |
| Nature | **rising** | End of turn | +1 power. | +2 total | no |
| People | **local** | Ongoing | +1 power while on-topic (the Topic's stat is its home stat). | +1 | no |
| Science | **encyclopedic** | On reveal | +2 power if you have no other cards here. | +2 | **yes: interleaving** |
| History | **nostalgic** | On reveal | +2 power if you are losing here. | +2 | **yes: interleaving** |
| Space | **translated** | On reveal | +1 power to your most recently revealed card in each *other* Topic. | +2 | yes |
Why this set: every keyword is one line and one trigger, and only adds power. The largest swing is +2 on one card,
so no lane moves by more than 3 from a keyword (viral at its cap). Nothing scales with faction count, so a mono-Culture deck gets no
consistency edge (`factionSynergyCap` is unused). Viral rewards wide turns, and Culture cards are spread over three keywords (§10.1 step 3).
Three keywords (encyclopedic, nostalgic, translated) react to the interleaved reveal (§5.4), so initiative and queue order matter.
Precise rules (for engine-b):
- **On reveal** resolves the moment the card enters its Topic, and sees every card revealed before it (both players,
including earlier reveals this turn). The bonuses it grants are permanent counters.
- **Ongoing** is a condition, recalculated whenever power is read.
- **End of turn** resolves in §5.5 order: the initiative player's cards first, then Topic 1→3, then reveal order.
- **viral:** gain = min(`base` + `amount` × N, `cap`), where N = the owner's cards revealed earlier in *this* turn's Reveal phase, in any Topic.
Cards that went back to hand don't count. v1.1: base 1, amount 1, cap 3 (v1 was base 0, cap 2).
- **rising:** stops after it has gained +2 in total.
- **local:** "on-topic" is exactly the §1.4/§6 condition. It's simply an extra +1 on top of `onTopicBonus`.
- **encyclopedic:** with `sides: 1` (v1.1 default), "no other cards here" means **your** side of this Topic is empty, apart from this card,
when it reveals; the opponent's cards don't matter. `sides: 2` is v1 (both sides empty).
- **nostalgic:** "losing here" means the opponent's Topic power here is strictly greater than yours, counted just *before* this card is added.
The consensus tie-break doesn't count: a tie is not losing.
- **translated:** "most recently revealed" means the highest global reveal sequence number among your cards in that Topic. Topics where you have no card are skipped.
- No v1 keyword uses the RNG. §4's "later random events" slot stays empty in v1.
**Already in code, disabled by default (`keywordsEnabled` excludes them; for modes only):**
- **edition:** +2 for each other copy of this `pageId` on your side.
- **obscure:** the opponent sees it face down until the end of the match.
- **wildcard:** its home stat becomes the Topic's stat on reveal.
These are engine-b's existing implementations. The faction-based versions of `local` and `wildcard`, and the token version of `viral`, are
**replaced** by the definitions above.
**`keywordParams` defaults** (keyed by keyword id; modes may change them within the §11 ranges):
```json
{ "viral": {"base": 1, "amount": 1, "cap": 3, "powerDelta": 0}, "rising": {"amount": 1, "cap": 2, "powerDelta": -1},
"local": {"amount": 1, "powerDelta": 0}, "encyclopedic": {"amount": 2, "sides": 1, "powerDelta": 0},
"nostalgic": {"amount": 2, "powerDelta": 0}, "translated": {"amount": 1, "powerDelta": 0},
"edition": {"amount": 2, "powerDelta": 0}, "obscure": {"powerDelta": 0}, "wildcard": {"powerDelta": 0} }
```
End-of-turn effects on the last turn resolve **before** scoring (§7).
**v1.2 candidates** (not in v1.1; decide on a larger run, per BALANCE.md's v1.1 re-run): the keyword values are still uneven:
translated +1.64 and nostalgic +1.69 per card vs local +0.69 (critic's target was ±0.3). The starter decks stay even only because each one
pairs a strong keyword with a weak one. The first levers are `nostalgic.powerDelta -1` and `translated.powerDelta -1`, plus `local.amount 2`
if owned decks stay under 45% against Headliners. Both are config only.
### 10.3 Portals (Topic modifiers) and the default Topic pool
Each Topic has `{ id, name, stat, portal | null }`. The default `topicPool` has 9 Topics, one plain and two with
Portals for each stat. A Portal applies from the moment its Topic is revealed (§3), except where it says otherwise.
| Topic | Stat | Portal rule |
|---|---|---|
| Current Events | attention | – |
| Main Page | attention | At scoring, each card here gets +1 power. |
| Trending | attention | Trend arrows (§1.6) count double here (±2). |
| Archive | depth | – |
| Featured Article | depth | At end of the match, the player with the single highest-power card here gets +2 Topic power here. If the highest card power here is tied (between the two players), nobody gets the bonus. |
| Stub | depth | Lane capacity here is 2 per player. |
| Translations | reach | – |
| Interwiki | reach | +1 Topic power here for a player who has at least one card in all 3 Topics. |
| Disambiguation | reach | Every card here counts as on-topic, whatever its home stat. |
As data: `portal = { kind: "scoringCardBonus", amount: 1 } | { kind: "trendMultiplier", factor: 2 } |
{ kind: "highestCardBonus", amount: 2 } | { kind: "laneCapacity", value: 2 } | { kind: "spreadBonus", amount: 1 } |
{ kind: "allOnTopic" }`. Modes may add Topics with these kinds, using new numbers, but not new kinds (engine code).
**`portalParams` defaults** (keyed by Topic id): `{ "main-page": {"amount": 1}, "trending": {"factor": 2},
"featured-article": {"amount": 2}, "stub": {"value": 2}, "interwiki": {"amount": 1} }`. Disambiguation has no numbers.
Topic ids are the names in kebab-case: `current-events, main-page, trending, archive, featured-article, stub,
translations, interwiki, disambiguation`. The UI shows each Portal's one-line rule **on the lane itself**, not behind a tap.
---
## 11. Peer-hosted modes *(owner: council-designer-b)*
A **mode** is data, published by a node operator and run by the match host:
```json
{
"id": "blitz", // [a-z0-9-]{3,32}, unique per host
"name": "Blitz", "description": "Four quick turns.", // ≤ 40 and ≤ 280 chars, plain text
"rules": { "turns": 4, "energyByTurn": [2, 3, 4, 5], "portalReveal": [1, 1, 2], "deckSize": 10 },
"topicPool": [ /* optional: Topic objects (10.3); default if omitted */ ],
"starterDecks": [ /* optional: lists of loaner pageIds */ ]
}
```
- `rules` is a **partial** `RulesConfig` (§9) merged over the defaults. Unknown keys are rejected.
- **Validation** (modes owns the code; these are the limits). If any check fails, the mode is refused:
| Key | Allowed |
|---|---|
| `turns` | 3–8, and `turns × turnSeconds ≤ 240` |
| `energyByTurn` | length = `turns`, each 0–10 |
| `bonusEnergy` | 0–3 |
| `portalReveal` | length = `topicCount`, each 1..`turns`, non-decreasing |
| `topicCount` | 1–5; `distinctStats` only if ≤ 3 |
| `laneCapacity` | 1–6 |
| `deckSize` | 6–20; `maxCopiesPerPage` 1–4; `maxLoaners` 0..`deckSize` |
| `openingHand` | 0–6; `handLimit` 1–10 |
| `turnSeconds` | 10–60; `afkTurns` 1–5 |
| `costThresholds` | 5 strictly ascending integers in 0–100 |
| `powerByCost` | 7 integers in 0–20, non-decreasing from index 1 |
| `onTopicBonus` (7 entries), `vanillaBonus` | 0–5 each |
| `trendUp` / `trendDown` | 0–200 / −200–0 |
| `factionSynergyCap` | 0–6 |
| `keywordsEnabled` | any subset of the 9 implemented ids (the 6 v1 defaults + edition, obscure, wildcard); a card whose keyword is disabled is vanilla, *without* `vanillaBonus` |
| `keywordParams`, `portalParams` | only the keys listed in §10 / §10.3; amounts and caps 0–6, `viral.base` 0–3, `encyclopedic.sides` 1–2, `powerDelta` −3–0 |
| `cultureKeyword` | `"home"` or `"faction"` |
| `verifyStats` | the host may only turn this off for modes that use only loaners (`maxLoaners == deckSize` and owned cards banned) |
| `topicPool` | 3–20 Topics, Portal kinds only from 10.3 |
- **Identity:** `modeHash` = SHA-256 of the canonical JSON (sorted keys, no whitespace) of the *merged* rules
plus the topicPool. A match record holds `{ modeHash, seed, both decks' CardInputs, both players' commit lists }`,
and that is enough to replay it exactly.
- **What modes can't change** is the "Fixed for all modes" list in §9. Modes also can't add keywords or Portal kinds, or
stake, move or burn NFTs.
Example modes (good defaults for operators to publish):
| Mode | Overrides | Feel |
|---|---|---|
| **Blitz** | `turns 4`, `energyByTurn [2,3,4,5]`, `portalReveal [1,1,2]`, `deckSize 10`, `turnSeconds 20` | about 2 minutes |
| **Deep Cuts** | `topicPool`: only the 3 depth Topics, `distinctStats false` | rewards long, obscure articles |
| **Loaner League** | `maxLoaners 12` and owned cards banned (`"ownedCards": false`, a new boolean key, default true) | fully level, for tournaments |
| **Big Board** | `topicCount 5`, `distinctStats false`, `portalReveal [1,2,3,4,5]`, `laneCapacity 3` | win 3 of 5; **tablet/desktop only** (5 × 3 slots won't fit a 360 px phone) |
| **No Keywords** | `keywordsEnabled []`, `onTopicBonus [0,3,3,3,3,3,3]` | Top Trumps–style tutorial mode |
---
## 12. Worked example match *(owner: council-designer-b)*
**Ruleset for this example: v1.0** (the golden test pins it): `cultureKeyword "faction"`, `viral {base 0, cap 2}`,
`encyclopedic {sides 2}`, `rising.powerDelta 0`; everything else as §9. *(The v1.1 numbers are in §12.1.)*
These are the §1–9 rules as of this PR:
- `powerByCost [0,1,2,4,6,8,10]`, `onTopicBonus [0,1,1,2,2,3,3]`, `vanillaBonus 1`
- `laneCapacity 3`, `energyCarryMax 1`, interleaved reveals, alternating initiative, `portalReveal [2,3,4]`
- `underdog false`, no trend arrows
The derived cards are given directly (the §1 output), so the test needs no percentile table.
Topics: **T1 Archive (depth, plain)**, **T2 Main Page (attention; Portal revealed turn 3; at scoring, each card here gets +1)**,
**T3 Translations (reach, plain; revealed turn 4)**. Ana (p0) has initiative on odd turns and Bo (p1) on even turns.
| Player | Card | Faction | Cost | Base | Home | Keyword |
|---|---|---|---|---|---|---|
| Ana | Octopus | Nature | 1 | 2 | depth | – (vanilla) |
| Ana | Moon | Space | 2 | 2 | reach | translated |
| Ana | Coral reef | Nature | 2 | 2 | depth | rising |
| Ana | Albert Einstein | People | 4 | 6 | attention | local |
| Ana | Ancient Rome | History | 5 | 8 | depth | nostalgic |
| Ana | Library of Alexandria | History | 4 | 6 | depth | nostalgic |
| Ana | Periodic table | Science | 3 | 4 | depth | encyclopedic |
| Bo | Haiku | Culture | 1 | 2 | attention | – (vanilla) |
| Bo | Mars | Space | 2 | 2 | reach | translated |
| Bo | Jazz | Culture | 2 | 2 | attention | viral |
| Bo | Pop music | Culture | 4 | 6 | attention | viral |
| Bo | Jupiter | Space | 5 | 8 | reach | translated |
| Bo | Fern | Nature | 3 | 4 | depth | rising |
| Bo | Mona Lisa | Culture | 3 | 4 | attention | viral |
Topic power is written as **Ana–Bo** after End of turn. Energy is "available / spent".
| Turn | Init | Energy Ana / Bo | Reveals, in interleaved order | T1 | T2 | T3 |
|---|---|---|---|---|---|---|
| 1 | Ana | 1/1, 1/1 | Ana Octopus → T1: 2 +1 on-topic = **3**. Bo Haiku → T2: 2+1 = **3** | 3–0 | 0–3 | 0–0 |
| 2 | Bo | 2/2, 2/0 | Bo passes (to carry 1). Ana Moon → T3: 2+1 = 3. translated: Octopus +1 → 4 | 4–0 | 0–3 | 3–0 |
| 3 | Ana | 3/2, **4**/4 | Main Page Portal revealed. Ana Coral reef → T1: 2+1 = 3. Bo Mars → T3: 2+1 = 3, translated: Haiku +1 → 4. Bo Jazz → T2: 2+1 = 3, viral (Mars earlier) +1 → 4. End: rising Coral reef → 4 | 8–0 | 0–8 | 3–3 (tie; on-topic 1–1, so no leader) |
| 4 | Bo | **5**/4, 4/4 | Bo Pop music → T3: 6 (off-topic), viral +0. Ana Einstein → T2: 6+2 on-topic, local +1 = **9**. End: rising Coral reef → 5 (cap reached) | 9–0 | 9–8 | 3–9 |
| 5 | Ana | **6**/5, 5/5 | Ana Ancient Rome → T3: 8 (off-topic). nostalgic: losing 3–9, so +2 → **10**. Bo Jupiter → T1: 8 (off-topic). translated: Jazz → 5 (T2), Pop music → 7 (T3) | 9–8 | 9–9: **Bo leads** (consensus: on-topic 1 vs 2) | 13–10 |
| 6 (last) | Bo | **7**/7, 6/6 | Bo Fern → T1: 4+2 = 6. Ana Library → T1: 6+2 = 8. nostalgic: losing 9–14, so +2 → **10**. Bo Mona Lisa → T2: 4+2 = 6, viral (Fern earlier) +1 → 7. Ana Periodic table → T2: 4 (off-topic), encyclopedic: no (cards here). End: rising Fern → 7. Scoring: Main Page +1 per card in T2 | **19–15** | **15–19** | **13–10** |
Final card values:
- **T1:**
- Ana: Octopus 4 + Coral reef 5 + Library 10 = 19.
- Bo: Jupiter 8 + Fern 7 = 15.
- **T2:** each card +1 from Main Page.
- Ana: Einstein 10 + Periodic table 5 = 15.
- Bo: Haiku 5 + Jazz 6 + Mona Lisa 8 = 19.
- **T3:**
- Ana: Moon 3 + Ancient Rome 10 = 13.
- Bo: Mars 3 + Pop music 7 = 10.
**Ana wins 2–1.** Three moments show the v1 mechanics:
- **Carry-over:** Bo passed on turn 2 to afford two plays on turn 3 (4 energy).
- **Consensus:** T2 at 9–9 after turn 5 went to Bo, who had two on-topic attention cards to Ana's one.
- **Interleaving decides the match:** on turn 6 Bo had initiative, so Fern revealed before Library. Ana was then losing T1 (9 vs 14) and nostalgic fired (+2).
If Ana had revealed first, she'd have been ahead 9–8, nostalgic would not have fired, and T1 would have been 17–15. Ana still wins that lane, but by 2 instead of 4.
**Golden test for engine-b:**
- **Inject:** the Topic order, the opening hands and the draw order directly, with each player's commit list per turn as in the table. The queue order within a turn is as listed.
- **Expected final Topic powers:** `T1 19–15, T2 15–19, T3 13–10`. Winner: p0, by 2 Topics to 1.
- **Also assert:**
- after turn 5, T2 is a 9–9 tie with Bo leading by consensus
- Ana's carried energy is 1 at the start of turns 4, 5 and 6
- Bo's carried energy is 1 at the start of turn 3
### 12.1 The same match under the v1.1 ruleset
Same cards, Topics, hands and commit lists; ruleset = §9 defaults (v1.1: `viral {base 1, cap 3}`, `encyclopedic {sides 1}`,
`rising.powerDelta -1`, `cultureKeyword "home"`). The example's Culture cards (Jazz, Pop music, Mona Lisa) are attention-home, so they stay viral.
Energy and carry are unchanged.
What changes:
- **rising −1:** Coral reef is 1 base and ends at 1+1+2 = **4**; Fern is 3 base and ends at 3+2+1 = **6**.
- **viral base 1:** Jazz +2 (Mars earlier), Pop music +1 (base only), Mona Lisa +2 (Fern earlier).
- **nostalgic still fires** for Library (Ana is losing T1 8–13 just before it). **encyclopedic still doesn't** (Einstein is already on Ana's side of T2).
| After turn | T1 | T2 | T3 |
|---|---|---|---|
| 1 | 3–0 | 0–3 | 0–0 |
| 2 | 4–0 | 0–3 | 3–0 |
| 3 | 7–0 | 0–9 | 3–3 |
| 4 | 8–0 | 9–9 | 3–10 |
| 5 | 8–8: **Ana leads** (consensus: on-topic 2 vs 0) | 9–10 | 13–11 |
| 6, before scoring | 18–14 | 13–18 | 13–11 |
| **final** (Main Page +1 per card in T2) | **18–14** | **15–21** | **13–11** |
**Ana still wins 2–1.** The moments in v1.1:
- **Carry-over:** unchanged.
- **Consensus:** moves to **T1 after turn 5** (8–8). Ana has two on-topic depth cards there (Octopus, Coral reef) and Bo has only the off-topic Jupiter.
- **Interleaving:** unchanged in kind. If Ana had revealed Library before Fern, T1 would have been 8–8 at that moment, a tie is not losing,
so nostalgic would not fire and T1 would end 16–14.
**v1.1 golden test:** expected `lanePower [[18,14],[15,21],[13,11]]`, winner p0. Also assert the T1 8–8 consensus lead for p0 after turn 5.
(Cross-checked by council-critic with the #27 engine, `tower-plus:scratchpad/critic/golden11.ts`, and by hand by designer-b.)
+279
View File
@@ -0,0 +1,279 @@
# WikiDeck Duel: design v1
**Status:** v1.3, Night 2 (2026-11-17): v1 plus night-review's nits on PR #66, duel-cards' implementation choices from PR #70, and the balance v1 rules `secondPlayerCitations` 1 and `turnCap` 40 (#64). Owner: duel-design (issue #60).
**Base:** `docs/night2/BRIEF.md` (Eric's locked decisions). Where this file and BRIEF disagree, BRIEF wins. Reviews used: designer (rules), data §2–4 (derivation), critic S7 (tone).
**Consumers:** duel-core (`shared/duel/` engine and `types.ts`), duel-cards (`classify/derive/effects/text.ts`), sim-balance (bots, `DUEL_BALANCE.md`), ui-duel / ui-pvp (faces, recap), host-pvp (rulesets).
Every number in this file is a **ruleset key or a derivation table value**, not code. sim-balance tunes them in data-only PRs.
---
## 1. The game in one paragraph
Two players each bring **40 cards they own on-chain**. Each card is a Wikipedia article turned into a Magic-style card by fixed rules over real data. You start at **20 Notability**. Every turn your **Citations** (mana) maximum rises by one. You deploy Figures, attack with them, block with yours, cast Events, adopt Doctrines onto Figures. The first player to 0 Notability, or unable to draw, loses. The match is written up afterwards as an encyclopedic "Battle of X (2026)" page.
---
## 2. Rules
### 2.1 Decks and eligibility (BRIEF decisions 1–2)
| Key | Default | Rule |
|---|---|---|
| `deckSize` | **40** | Exactly this many cards. Every card is an NFT the player owns (server-verified by `ownership.ts`). No Basics, loaners, proxies or tokens in decks. |
| `maxCopiesPerPage` | **4** | Different mints of the same `pageId` may appear up to 4 times (each mint once: it is unique). |
| colour/Portal rules | none | Any mix of Portals. |
| Eligibility | owns ≥ `deckSize` | Below it, the UI shows "23 / 40 cards" and how to mine. No practice with borrowed cards. Bot matches need eligibility too. |
Testnet end-to-end mode files may lower `deckSize`. Production stays at 40.
### 2.2 Setup
- **Notability** (life): `startingLife` **20**. At 0 you lose.
- **Empty library:** drawing from an empty library loses (`libraryEmpty: 'lose'`).
- **Opening hand:** `openingHand` **7**. One mulligan (`mulligans: 1`): shuffle, draw 7, put 1 on the bottom.
- **First player** (seeded coin flip) skips their first draw.
- **Second player** gets +1 Citation maximum on their first turn only (`secondPlayerCitations` **1**). Balance v1: this moved the first-player win rate from 60.1% to 54.4% in 1,500 heuristic-mirror matches (engine 4623930, #64). Neither an extra opening card (washed out by the mulligan and the hand limit) nor `firstPlayerDraws` worked.
### 2.3 Turn structure: refresh → draw → main → combat → end
1. **Refresh.** Citation maximum +1 (to `maxCitations` **10**), plus Territory (§2.5). Citations refill to the maximum. Your permanents untap. "Pending review" ends for your Figures. Start-of-turn triggers (Stub, Institutions) resolve in the order the permanents entered.
2. **Draw** 1 card.
3. **Main.** Play cards in any order while you can pay their cost in Citations: Figures, Locations (one per turn: `locationsPerTurn` 1), Institutions, Doctrines (attach to a Figure), Events (resolve immediately).
4. **Combat** (§2.4).
5. **End.** Damage on permanents clears. "Until end of turn" effects end. Discard down to `handLimit` **7** (the player chooses; bots discard highest cost first).
**Zones.** Library and hand are hidden (a hand's size is public). The battlefield and **Deleted** (the graveyard) are **public**: a discarded card is revealed and named once it reaches Deleted.
There is **no second main phase, no stack, no priority and nothing at instant speed.** Events are sorcery-speed. Triggers resolve immediately, in a fixed order: the active player's first, in the order the cards entered.
### 2.4 Combat
- **Attack:** the active player chooses any untapped Figures without Pending review and taps them. They attack the opponent only, never permanents.
- **Block:** the defender assigns **at most one blocker per attacker**, and each blocker blocks one attacker. Untapped Figures and Garrison Locations may block. Evasion (Featured) restricts who may block: **a Garrison Location can't block a Featured attacker** (only Featured, Translated or Multilingual Figures can).
- **Damage:** simultaneous. A blocked pair deal their power to each other. An unblocked attacker deals its power to the defending player's Notability.
- A permanent with damage ≥ toughness is **destroyed** (goes to the Deleted pile, the graveyard). Destroy triggers (Legacy) fire after all combat damage.
- No trample, no first strike, no deathtouch, no lifelink in v1.
### 2.5 Card types (BRIEF; classifier in data §1, built by duel-cards)
| Type | MTG analogue | Rules |
|---|---|---|
| **Figure** | creature | Power/toughness. Enters with **Pending review** (summoning sickness): can't attack until your next turn. |
| **Location** | land-ish permanent | Cost `ceil(derivedCost/2)` (1–3). **Territory:** +1 Citation maximum while you control it; at most **+2 in total** from Territory (`territoryCap` 2). **Garrison 0/T:** can block like a 0-power Figure (T = 1 + floor(pct_depth / 25), so 1–4). Can't attack. Damage clears at end of turn as usual; a Location destroyed in a block goes to Deleted. One per turn. |
| **Event** | sorcery | Resolves at once, then goes to Deleted. |
| **Institution** | artifact | A permanent with a static or triggered ability. |
| **Doctrine** | Aura | **Adopt:** attach to target Figure (yours or the opponent's for negative Doctrines). If the Figure leaves play, the Doctrine goes to Deleted. |
Classification order (duel-cards owns it): P31 → P279 anchor classes → P31 label regex (`^human$`, fixing the `\bhuman\b` bug) → categories → description → **Doctrine**. The derived profile is **frozen** the first time a card is seen.
### 2.6 Clock and length
- `turnSeconds` **60** per turn plus `bankSeconds` **300** per player. The turn timer runs first; when it expires, the player's bank drains; when the bank is also empty, the turn passes (`timeout` action). The bank does not refill. Three timeouts in a row concede.
- **`turnCap` 40** (20 turns each, balance v1). At the cap the higher Notability wins, then the higher total power in play, then the second player. In sims it decides about 4% of games and removes every game past 20 turns each, without moving the median.
- **Sim targets** (agreed with sim-balance, #64): median 8–12 turns each; first-player win rate 48–54%; < 5% of games past 20 turns each. **Balance v1 result** (`secondPlayerCitations` 1 + `turnCap` 40; 3,000 heuristic-mirror matches, engine 4623930, duel-cards 067de83; run by ui-duel for sim-balance, #64): first player **54.7%** [53–57], turns each 6 / 8 / 12, 0% past 20, about 8 minutes median.
### 2.7 Determinism
All randomness (shuffles, coin flip, template variants) comes from the match seed via `shared/game/rng.ts`. The same seed and actions always replay to the same state, log and recap.
---
## 3. Derivation summary (duel-cards implements; data §2)
- **Percentiles** `pct_a`, `pct_d`, `pct_r` from the sample cut points (DESIGN.md §1.1 counting). **fame** = mean of the three, rounded.
- **Cost** 1–6 by `costThresholds` [26, 39, 51, 63, 78].
- **Figure body budget** P+T = [3, 4, 6, 8, 10, 12] for cost 1–6 (MTG vanilla curve). Subtract **1 per taxed keyword** (§4). At cost 1, a second taxed keyword is skipped, so a body never drops below 1/1. Then `power = clamp(round(budget × pct_a / (pct_a + pct_d)), 1, budget − 1)`, toughness = the rest. Famous cards are bigger and pricier, never more efficient.
- **Magnitude** for non-Figures: `k = ceil(cost / 2)` (1–3).
- **Portal** (faction): Arts, Earth, History, Science, Society, or Neutral, scored by regex hits (designer §3.2); ties go to the type default (Figure → Society, Location → Earth, Event → History, Institution → Arts, Doctrine → Science). No deckbuilding restriction.
- **Flavour text:** first sentence of on-chain `article.extract`; if absent (the sample has none), the Wikidata `description`; if that is empty too, none.
---
## 4. Keywords (10)
Eight card keywords, each triggered only by real data and applied **to Figures only**, at most **2 per card** in the priority order below. Plus two rules keywords. Thresholds are measured on the 236-card sample (86 Figures); sim-balance may move them.
| # | Keyword | Trigger (real data) | Rules text | Body tax | Rate among Figures |
|---|---|---|---|---|---|
| 1 | **Protected** | category `Living people` | Can't be the target of Events or Doctrines an opponent controls. | 1 | 38% |
| 2 | **Featured** (flying) | pct_attention ≥ 85 | Can't be blocked except by Featured, Translated or Multilingual Figures. | 1 | 15% |
| 3 | **Legacy** | a `<year> deaths` category (or P570 in the frozen profile) | When this is destroyed, draw a card. | 1 | 33% |
| 4 | **Multilingual** | pct_reach ≥ 90 | Can block Featured Figures. Can't be blocked by Local Figures. | 1 | ~5% |
| 5 | **Translated** (reach) | 75 ≤ pct_reach < 90 | Can block Featured Figures. | 0 | 19% |
| 6 | **Well-sourced** | pct_depth ≥ 85 | Prevent the first 1 damage that would be dealt to this each turn. | 1 | ~13% before cap |
| 7 | **Stub** | pageBytes < 3000 or a `…stubs` category | At the start of your turn, put a +1/+1 counter on this, to a maximum of 2. | 0 | 33% |
| 8 | **Local** | raw reach = 0 (no other language editions) | Gets +1/+1 while you control a Compatriot Location (same P17). | 0 | 7% before cap |
| 9 | **Connected** | the card pair's frozen claims and links (§6) | +1/+1 while you control another card Connected to it; +2/+2 if Linked. | – | target 15–30% of pairs |
| 10 | **Trending** (haste) | Trending mode only: frozen live ▲ (trend ≥ +25% and daily ≥ 20) | Can attack the turn it enters. Gets +1/+0. | – | ~2% of cards |
Rule words that are not keywords: **Pending review** (summoning sickness), **Territory**, **Garrison**, **Adopt**, **Deleted** (the graveyard).
Notes:
- Protected deliberately sits first: living people are the most targeted, and the rule that matters most for them is mechanical, not a tone rule. It doesn't stop combat.
- Multilingual and Translated are two bands of the same reach axis, so a card never has both.
- A card's face shows at most 2 keyword words; the reminder text is behind a tap.
---
## 5. Effect templates (12, MTG register)
Subtype regexes run over the frozen P31/P279 labels, then the description, then categories. First match wins, in the order listed. `k = ceil(cost/2)`.
| # | Type / subtype | Trigger regex (labels, description) | Rules text |
|---|---|---|---|
| T1 | Figure | (all Figures) | Vanilla body + keywords. No other text. |
| T2 | Location | (all Locations) | Territory. Garrison 0/T. |
| T3 | Event: conflict | `battle\|war\|siege\|annexation\|protest\|riot\|uprising\|massacre\|coup` | Deal k+1 damage to target Figure or Location. |
| T4 | Event: disaster | `storm\|nor.easter\|cyclone\|hurricane\|earthquake\|flood\|tornado\|wildfire\|eruption\|famine\|epidemic` | Deal k−1 damage (minimum 1) to each Figure and each Location. |
| T5 | Event: election | `election\|referendum\|straw poll` | Create k 1/1 Society Figure tokens named Elected Member. |
| T6 | Event: sport | `match\|final\|season\|tournament\|championship\|grand prix\|race` | Target Figure you control gets +k/+0 until end of turn, then fights target Figure you don't control. |
| T7 | Event: default (period, ceremony, scandal, other) | anything else | Draw k cards. |
| T8 | Institution: organisation | `company\|organi[sz]ation\|party\|agency\|studio\|military\|university\|diocese\|council\|band\|musical group\|association\|business\|school\|college\|institute\|manufacturer` | Cost ≥ 4: "At the start of your turn, if you control fewer than 3 tokens, create a 1/1 Figure token named Staff Member." Cost ≤ 3: "When this enters, create a 1/1 Figure token named Staff Member." |
| T9 | Institution: creative work | `film\|album\|song\|book\|novel\|series\|video game\|painting\|software\|typeface\|newspaper\|magazine` | When this enters, draw a card. Figures you control that are Connected to this get +1/+1. |
| T10 | Institution: team, award, vehicle, other | anything else | Figures you control with power 2 or less get +1/+0. |
| T11 | Doctrine: Adopt | (all non-Pathology Doctrines) | Attach to target Figure you control. It gets +k/+k. Rider by subtype, checked in the order pathology (→ T12) > language/software > mathematics, so "functional programming language" is not maths: mathematics → "and has Well-sourced"; language/software → "Whenever it deals combat damage to a player, draw a card"; ideology/religion → "Other Figures you control of its Portal get +1/+0"; otherwise none. |
| T12 | Doctrine: Pathology | `disease\|clinical sign\|pathogen\|syndrome\|disorder\|toxin\|poison` | Attach to target Figure. It gets −k/−k. |
Elected Member tokens are Society; Staff Member tokens have no Portal ("1/1 Figure token"). Tokens are not cards: they never appear in decks, are created only by T5/T8, and are removed when they leave play. Token names are fixed generic strings, never real people.
Rules-text style:
- Second person, present tense, MTG terse register: "Deal 3 damage to target Figure."
- Card names appear in rules text only as "this" (never the article title), so text stays the same length on a 360 px face.
- No adjectives, no puns, no exclamation marks.
---
## 6. Connected (data §4 tiers)
Computed by the host at match start from frozen card profiles. No network calls during a match; the edge list is written into the match record.
| Tier | Test | Bonus | Cut order |
|---|---|---|---|
| **Linked** | either article links to the other (cached outlinks), or one card's claims cite the other's Q-id | +2/+2 | keep |
| **Compatriot** | share a P17 (country) or P27 (citizenship) value | +1/+1 | keep |
| **Contemporary** | same century from P569/P570/P571/P580/P585, **only between different types** | +1/+1 | cut 4th (BRIEF) |
| **Kin** | same P31 class, excluding Q5 human and Q16521 taxon | +1/+1 | cut 4th (BRIEF) |
- A Figure gets the **single best** bonus among its Connected cards you control (no stacking): +2/+2 if any is Linked, else +1/+1.
- A Connected creative-work Institution (T9) gives its +1/+1 on top.
- **Target:** 15–30% of random card pairs connected (text proxies measured 21.5% for country OR century). If sims show more, drop the noisiest keys first (P17 = United States, then France); if fewer, add P131.
- The card shows why: "Connected — *Pathrail Mosque* (country: Bangladesh)".
- **Rivalry** (log only): when two opposing cards are Connected, the log may note it once (§8.3). No mechanical effect.
---
## 7. Modes (on the duel engine)
| Mode | Decks | Notes |
|---|---|---|
| **Classic** | constructed from your own 40+ owned cards | Ranked (Elo, keyed to the eom.dev account). Invite links and quick match. |
| **Bot** | your own 40 | Only for eligible players. Not ranked. |
| **Draft** | both players eligible | Pool drawn from both players' collections (e.g. 30 cards from each, shuffled into one pool). Snake-pick until each has `deckSize`. Drafted cards are used for that match only; nothing moves on-chain. Ownership is credited in the recap ("from the collection of bob"). |
| **Trending** | any | Live pageview trend from the daily cache, frozen at match start: ▲ (≥ +25%, daily ≥ 20) gives Trending (haste, +1/+0); ▼ (≤ −50%, daily ≥ 20) gives −1/−0. Cut 3rd. |
| **House rules** | peer-hosted | Data-only, versioned schema. Keys: `deckSize`, `maxCopiesPerPage`, `startingLife`, `openingHand`, `mulligans`, `maxCitations`, `territoryCap`, `handLimit`, `turnSeconds`, `bankSeconds`, `turnCap`, `secondPlayerCitations`, `connectedTiers`, `keywordThresholds`. Night 1 mode files are rejected cleanly with a message. |
---
## 8. Tone and story text
### 8.1 Tone rules (BRIEF decision 5, critic S7)
1. **Lean in.** Living people, tragedies, wars, disasters, crimes, diseases and dictators are ordinary cards with full mechanics. No `solemn` flags, no special cases, no softened verbs. "Casualties" is fine. A protester shot in 1970 is a 6/3 with Legacy like anything else.
2. **Deadpan.** The text takes itself completely seriously. It **never winks** and **never comments on the matchup** ("in a surprising twist", "unlikely", "ironically" are banned). The comedy lives in the nouns: "*Heart rot* was adopted by *Tom Ewell*."
3. **Encyclopedic texture.** Past tense, third person, passive where Wikipedia would use it, dates and turn numbers, and `[n]` citations linking each card's real article at its revision.
4. **No adjectives about cards or players** (no "brutal", "heroic", "humiliating", "bloodiest"). Mechanical verbs only: deployed, attacked, blocked, was destroyed, was adopted by, took N damage, was cited.
5. **Never invent facts about real people.** All generated text is a deterministic template filled **only** with real card data (title, description, dates, categories, revision id) and real match data (turn, players, numbers). No LLM prose. No made-up quotes, crimes, acts, motives or feelings. The template verbs describe **game events of the card**, and every narrative paragraph is framed as a WikiDeck engagement; the footer says so.
6. **Allowed real data in text:** title, Wikidata `description`, birth/death years parsed from the description or frozen claims, the first extract sentence (flavour), Portal, type. Nothing else about a person.
7. **Dull guard:** 3–4 synonymous variants per event, chosen by the match seed (§8.2).
| DO | DON'T |
|---|---|
| "Turn 6. *Zbyszek Godlewski* (6/3) attacked. *Pallas's cat* (5/5) blocked. *Zbyszek Godlewski* was destroyed.[3][4]" | "Godlewski got mauled by a cat lol" |
| "The *April 2021 nor'easter* dealt 1 damage to each Figure. *Dichromorpha viridis* was destroyed.[5][6]" | "Mother Nature wiped out the poor grasshopper!" |
| "*Heart rot* was adopted by *Tom Ewell*, which became a 2/2.[7][8]" | "Ewell caught a tree disease, how embarrassing" (adjective + invented claim) |
| "The **Battle of Pathrail Mosque (2026)** was a WikiDeck engagement fought on 17 November 2026 between the forces of alice and bob." | "the bloodiest battle in WikiDeck history" |
### 8.2 Match-log templates (`shared/duel/story/log.ts`)
- One template group per engine `LogEvent` kind, each with **3–4 variants**. Variant = `rng(seed, turn, eventIndex) mod n`, using the match seed, so `replay` regenerates the log exactly.
- First mention of a card in a match adds its description in parentheses: "*Tom Ewell* (American actor, 1909–1994)". Later mentions use the title only.
- Each first mention of a card assigns it the next reference number `[n]`; later mentions reuse it.
- `logFor(seat)` hides the opponent's hand: their draws read "bob drew a card." and mulligans are not named. Discards are named, because Deleted is public (§2.3).
| Event | Variants (`{card}` italic, `{player}` the seat name) |
|---|---|
| deploy (Figure/Location/Institution) | "{player} deployed {card}." · "{card} was deployed by {player}." · "{player} brought {card} into play." · "{card} entered play under {player}." |
| event cast | "{player} cited {card}." · "{card} was cited by {player}." · "{player} invoked {card}." |
| adopt | "{doctrine} was adopted by {figure}, which became a {p}/{t}." · "{figure} adopted {doctrine} and became a {p}/{t}." · "{player} attached {doctrine} to {figure} ({p}/{t})." |
| attack | "{card} ({p}/{t}) attacked." · "{player} sent {card} ({p}/{t}) into combat." · "{card} ({p}/{t}) advanced on {opponent}." |
| block | "{blocker} ({p}/{t}) blocked {attacker}." · "{attacker} was blocked by {blocker} ({p}/{t})." · "{blocker} ({p}/{t}) intercepted {attacker}." |
| player damage | "{opponent} lost {n} Notability ({life} remaining)." · "{card} dealt {n} damage to {opponent}, leaving {life} Notability." · "{opponent}'s Notability fell to {life}." |
| destroyed | "{card} was destroyed." · "{card} was moved to Deleted." · "{card} was removed from play." · "{card} did not survive the turn." |
| effect damage | "{source} dealt {n} damage to {target}." · "{target} took {n} damage from {source}." · "{source} inflicted {n} damage on {target}." |
| token created | "{source} produced {n} {token}." · "{n} {token} were created by {source}." · "{source} created {n} {token}." |
| draw (own) | "{player} drew {card}." · "{card} was drawn." · "{player} added {card} to their hand." |
| draw (hidden) | "{player} drew a card." · "{player} drew {n} cards." |
| Connected | "{a} and {b} were Connected ({reason})." · "{a} was Connected to {b} through {reason}." · "Records link {a} and {b} ({reason})." |
| discard | "{player} discarded {card}." · "{card} was discarded by {player}." · "{player} moved {card} from hand to Deleted." |
| rivalry (opposing Connected pair, once) | "{a} and {b} are both listed under {reason}." · "{a} and {b} share {reason}." · "The two sides were linked through {reason}." |
| turn start | "Turn {n}." (fixed) |
| game over | "{loser} was nominated for deletion. The result was **delete**." · "{loser} was reduced to 0 Notability. {winner} won." · "{winner} won on turn {n}." · (library) "{loser} could not draw a card and lost." · (concede) "{loser} withdrew." |
### 8.3 Battle recap (`shared/duel/story/recap.ts`)
A deterministic function `recap(matchRecord) → RecapDoc` (structured data; ui-pvp renders it). Inputs are the match record only: seed, players, decks' frozen card profiles, the event list, the date. No network, no clock.
**Title:** "{Prefix} {Name} ({year})"
- **Name:** the winner's highest-cost Location that entered play; else the loser's; else the highest-cost Event cast in the match; else the winner's highest-cost card deployed; ties broken by first played. Living people and tragedies are eligible like any card ("Battle of Zbyszek Godlewski (2026)" is valid).
- **Prefix:** "Skirmish at" (≤ 6 turns each), "Battle of" (7–12), "Siege of" (≥ 13). For a Figure name: "Battle of" becomes "{Name} campaign" (no prefix).
- **Year:** the match date's year (2026).
**Infobox:**
| Field | Value |
|---|---|
| Date | the match date, "17 November 2026" |
| Location | the Name card if it's a Location, else "WikiDeck host {hostName}" |
| Result | "Decisive {winner} victory" (margin ≥ 10 Notability) · "Pyrrhic {winner} victory" (winner ≤ 3) · "{winner} victory" · "{winner} victory by withdrawal" (concession/timeout) · "{winner} victory by attrition" (empty library) · "Victory by decision" (turn cap) |
| Belligerents | the two seat names (account display names) |
| Commanders | each side's highest-cost Figure deployed |
| Strength | cards deployed per side ("14 cards") |
| Casualties and losses | cards destroyed per side, count, then the titles |
**Sections:**
1. **Lead** (1 paragraph): "The {title} was a WikiDeck engagement fought on {date} between the forces of {a} and {b}. It lasted {n} turns and ended in {result-lowercase}.[1]" + one sentence on the first Figure deployed: "Contemporary accounts note the deployment of {card} in the {ordinal} turn.[n]"
2. **Background:** each side's Portal mix ("History was the largest Portal in the force of alice (eleven of twenty cards played)").
3. **Course of the battle:** the 3–5 largest Notability swings and every destroyed card with cost ≥ 4, in turn order, as log sentences (variants from §8.2, but seeded with `seed ⊕ 'recap'` so the recap doesn't repeat the log word for word).
4. **Aftermath:** final Notability, turns, rating change if ranked.
5. **See also:** Connected pairs and rivalries.
6. **References:** every card played by either side, in first-mention order: "{n}. ^ "{title}". Wikipedia. Revision {revisionId}. Minted as token {tokenId}." linking `https://en.wikipedia.org/w/index.php?title={title}&oldid={revisionId}`.
**Branding and footer:** WikiDeck masthead, never the Wikipedia/Wikimedia name or logo. Footer: "This page records a WikiDeck match. Card names refer to Wikipedia articles; the events described are game events." Shareable at `/recap/{matchId}`.
---
## 9. Worked examples (12 cards from `docs/game/card-sample.json`)
Computed with the rules above (data §2 budget, §4 keyword priority, §5 templates) by a script over the sample, and checked by hand. duel-cards' implementation (PR #70) agrees on all 12, with the two Portal corrections now applied (#9, #11). Flavour is the `description` because the sample has no extracts. Final numbers come from duel-cards' implementation; if they differ, the implementation and its golden test win and this table is updated.
| # | Card | Derivation | Card |
|---|---|---|---|
| 1 | **Marc Fesneau** (living person) | P31 human → Figure. "politician" → Society. pct 76/68/63 → fame 69 → cost 5, budget 10. Living people → Protected (−1) → 9 split 76:68 → 5/4. | **Marc Fesneau** · 5 · Figure — Society · **5/4** · *Protected.* "French politician (born 1971)." |
| 2 | **Tom Ewell** | human → Figure, Arts. 94/83/93 → fame 90 → cost 6, budget 12. Featured (−1), Legacy (−1); Multilingual dropped by the cap. 10 → 5/5. | **Tom Ewell** · 6 · Figure — Arts · **5/5** · *Featured. Legacy.* "American actor (1909–1994)." |
| 3 | **Zbyszek Godlewski** (killed at a 1970 protest) | human → Figure, History. 77/48/78 → fame 68 → cost 5. Legacy (1970 deaths, −1), Translated (pct_r 78, 0). 9 split 77:48 → 6/3. | **Zbyszek Godlewski** · 5 · Figure — History · **6/3** · *Legacy. Translated.* "Polish protester (1952–1970)." |
| 4 | **Pallas's cat** | taxon → Figure, Earth. 99/97/99 → fame 98 → cost 6. Featured (−1), Multilingual (−1); Well-sourced dropped. 10 → 5/5. | **Pallas's cat** · 6 · Figure — Earth · **5/5** · *Featured. Multilingual.* "Species of small wild cat." |
| 5 | **Dichromorpha viridis** | taxon → Figure, Earth. 58/3/53 → fame 38 → cost 2, budget 4. Stub (1,347 bytes, untaxed). Split clamps to 3/1. | **Dichromorpha viridis** · 2 · Figure — Earth · **3/1** · *Stub.* "Species of grasshopper." |
| 6 | **Monoid** (maths) | no P31; category "Algebraic structures" → Doctrine, Science. fame 96 → cost 6, k = 3. Maths rider. | **Monoid** · 6 · Doctrine — Science · Attach to target Figure you control. It gets +3/+3 and has Well-sourced. "Algebraic structure with an associative operation and an identity element." |
| 7 | **Heart rot** (disease) | P31 plant disease → Doctrine, Pathology (T12). Science. fame 68 → cost 5, k = 3. | **Heart rot** · 5 · Doctrine — Science · Attach to target Figure. It gets −3/−3. "Fungal disease of trees." |
| 8 | **April 2021 nor'easter** (disaster) | P31 nor'easter → Event, disaster (T4). Earth. fame 50 → cost 3, k = 2. | **April 2021 nor'easter** · 3 · Event — Earth · Deal 1 damage to each Figure and each Location. "Slow-moving nor'easter in April 2021." |
| 9 | **2025 Tehreek-e-Labbaik Pakistan protests** (conflict) | P31 protest → Event, conflict (T3). Society (protest/riot hits outnumber "war"). fame 64 → cost 5, k = 3. | **2025 Tehreek-e-Labbaik Pakistan protests** · 5 · Event — Society · Deal 4 damage to target Figure or Location. "Gaza war protests in Pakistan." |
| 10 | **1954 Gold Coast general election** | P31 legislative election → Event, election (T5). Society. fame 45 → cost 3, k = 2. | **1954 Gold Coast general election** · 3 · Event — Society · Create two 1/1 Society Figure tokens named Elected Member. *(no flavour: empty description)* |
| 11 | **Pathrail Mosque** | P31 mosque → Location, Society ("mosque" hits in P31, the description and two categories outnumber 15th-century + Sultanate). fame 49 → cost 3 → Location cost 2. pct_d 54 → Garrison 0/3. | **Pathrail Mosque** · 2 · Location — Society · Territory. Garrison 0/3. "Mosque in Faridpur, Bangladesh." |
| 12 | **Tatsunoko Production** | P31 animation studio → Institution, organisation (T8). Arts. fame 93 → cost 6. | **Tatsunoko Production** · 6 · Institution — Arts · At the start of your turn, if you control fewer than 3 tokens, create a 1/1 Figure token named Staff Member. "Japanese animation studio." |
Interactions these cards show:
- *Heart rot* on *Marc Fesneau* is illegal (Protected stops opponents' Doctrines). On *Tom Ewell* it makes a 2/2 Featured Legacy.
- *Monoid* on *Dichromorpha viridis* makes a 6/4 Well-sourced Stub that keeps growing to 8/6.
- *April 2021 nor'easter* destroys *Dichromorpha viridis* and both Elected Members, and leaves *Pathrail Mosque* at 0/3 with 1 damage.
- *Pathrail Mosque* and *2025 Tehreek-e-Labbaik Pakistan protests* are not Compatriots (Bangladesh vs Pakistan); *Pathrail Mosque* and *Zbyszek Godlewski* are not Contemporary (15th vs 20th century).
Sample log:
> Turn 5. alice deployed *Zbyszek Godlewski* (Polish protester, 1952–1970).[1]
> Turn 6. bob cited the *April 2021 nor'easter* (Slow-moving nor'easter in April 2021).[2] The *April 2021 nor'easter* dealt 1 damage to *Zbyszek Godlewski*.
> Turn 7. *Zbyszek Godlewski* (6/3) attacked. *Pallas's cat* (5/5) blocked *Zbyszek Godlewski*.[3] *Zbyszek Godlewski* was destroyed. alice drew a card.
---
## 10. Hand-offs
- **duel-core:** §2 rules and ruleset keys; `LogEvent` kinds needed by §8.2 (deploy, cast, adopt, attack, block, playerDamage, destroyed, effectDamage, tokenCreated, draw, connected, turnStart, gameOver with reason `life|library|concede|timeout|turnCap`). Each event carries card ids, P/T after the event, and Notability where relevant.
- **duel-cards:** §3 derivation, §4 thresholds and priority, §5 templates T1–T12, §6 tiers; the §9 table as golden tests.
- **sim-balance:** §2.6 targets; tune `keywordThresholds`, `connectedTiers`, T4's k−1, the Stub cap and the T8 token rate; report the first-player win rate and the cost-6 dominance check.
- **ui-pvp:** renders `RecapDoc` from §8.3.
## 11. Open questions for the sim (not blockers)
1. Disaster T4 at k−1 may be weak against 40-card boards; try k.
2. Stub at +2 counters on cheap bodies may be too strong (33% of Figures); try max 1.
3. **Locations (balance v1 finding):** Location-heavy decks win ~20% and Locations are the weakest cards. `territoryCap` 3, `locationsPerTurn` 2 and `garrisonStep` 15 move them ±3 points at most. Tested and **rejected for v1**: a T2 rider "When this enters, draw a card". It lifts the Location card (47 → 52%) but not Location-heavy decks (18 → 17%), pushes the first player back to 58.4%, and brings back library-out losses. Open for v1.1: a lever that helps Location-heavy decks without extra card flow, e.g. Territory that scales (+1 Citation per two Locations) or Garrison Locations that can block Featured.
+232
View File
@@ -0,0 +1,232 @@
# WikiDeck game engine API (v1)
**Owner:** engine-a (core and API) and engine-b (keywords, card derivation, tests). Issue: claude/wikideck#3.
**Rules:** `docs/game/DESIGN.md` v1 ("Edit War", as frozen in PR #17 @ cf2fceb). Where this file and DESIGN.md disagree, raise it on #3.
**Stability:** UI, simulation and modes build against this file. Any change to a name or shape below is
announced on #3 before it merges.
The engine is a pure TypeScript library in `shared/game/`:
- It does no I/O and has no clocks: no `Date.now()`, no `Math.random()`.
- It imports nothing from `server/` or `src/`.
- It runs entirely **off-chain**. A match never reads, moves, locks or burns an NFT.
- It is deterministic: the same `(ruleset, decks, seed, actions)` always gives the same state, byte for byte.
**Import only from `shared/game/index.ts`.**
```ts
import { createMatch, legalActions, applyAction, isOver, result, viewFor,
serialize, deserialize, replay, DEFAULT_RULESET, parseRuleset, validateDeck,
starterDeck, previewPool, type MatchState, type Action, type GameCard } from '../shared/game/index.ts';
```
## 1. Lifecycle
```ts
let state = createMatch(DEFAULT_RULESET, [deck0, deck1], seed); // turn 1, planning
while (!isOver(state)) {
// Each seat (in any order, any interleaving) queues plays, then commits.
const actions = legalActions(state, seat); // play / undo / commit
state = applyAction(state, chosen); // returns a NEW state
// When both seats have committed, the turn resolves inside applyAction.
}
const r = result(state); // { winner: 0 | 1 | null, reason, topicsLed, lanePower, totalPower, turns }
```
| Function | Description |
|---|---|
| `createMatch(ruleset, [deck0, deck1], seed: number \| string): MatchState` | Validates the ruleset (`parseRuleset`) and both decks (`validateDeck`), then performs DESIGN §4 setup and turn-1 start. A `seed` is a uint32. A string is hashed with FNV-1a (`seedFromString`). Throws `ZodError` for a bad ruleset and `IllegalActionError` for a bad deck. |
| `legalActions(state, seat): Action[]` | Lists every `play` (each card × each lane it can legally go into), every `undo`, and `commit`. Returns `[]` once the seat has committed or the match is over. It never lists `concede` (always legal while the match is running) or `timeout` (host only). |
| `applyAction(state, action): MatchState` | Pure: the input is never mutated. Throws `IllegalActionError` if the action is illegal. |
| `actionProblem(state, action): string \| null` | Explains why an action is illegal, without throwing. Use it for UI hints and for validating client input on a host. |
| `isOver(state)` / `result(state)` | `result` is `null` while the match is running. |
| `viewFor(state, seat): PlayerView` | The **hidden-information view**. Send only this to clients and AIs, never `MatchState`, because a `MatchState` contains both deck orders and the RNG state. |
| `serialize(state): string` / `deserialize(json): MatchState` | Lossless JSON. `MatchState` is plain data. |
| `replay(ruleset, decks, seed, actions): MatchState` | Rebuilds a match from its inputs. `state.actions` records every action applied, so a host can store `(ruleset, decks, seed, actions)` and verify a result later. |
| `validateDeck(ruleset, cards): string[]` | Returns the list of problems; empty means the deck is legal. |
| `cardPower(state, cardInPlay)`, `lanePower(state, lane)`, `laneLeader(state, lane)`, `topicsLed(state)` | Power queries (DESIGN §6). `laneLeader` and `topicsLed` apply the consensus tie-break. |
| `scoringPower(state, lane)` | Lane power plus revealed "at scoring" Portals (§7). `result().lanePower` uses this. |
| `laneCapacity(state, lane)` | Per-seat capacity of a lane. A Portal such as Stub may override it. |
| `isOnTopic(state, cardInPlay)` | §1.4/§6 on-topic test, including Disambiguation. |
| `STARTER_DECKS`, `starterDeckById(id, rules?)` | The 4 DESIGN §2.1 starter decks (`stargazers`, `field-guide`, `headliners`, `pioneers`), derived through `deriveCard` with the given rules (a `Ruleset` works). **Use these for new players.** |
| `starterDeck(seed)` | Older placeholder loaner deck (seeded, 2 per cost from `STARTER_POOL`). Kept for tests and tools; not a §2.1 starter deck. |
| `previewPool(rules?)` | The 20 frozen `docs/game/starter-cards.json` loaners (council-data), derived via §1 against the `card-sample-2026-11-16` percentile table. For loaner slots and draft modes. |
| `deriveCard(input, table, rules?)`, `deriveDeck(inputs, table, rules?)` | DESIGN §1 + §10.1: `CardInput` → `GameCard`. **Hosts only**: run on host-frozen, plausibility-checked inputs, never on client JSON. A `Ruleset` can be passed as `rules`. |
| `placeCard(state, seat, lane, card)`, `revealPortal(state, lane)` | Test helpers for building boards by hand. |
## 2. Actions
```ts
type Action =
| { type: 'play'; player: 0 | 1; cardId: string; lane: number } // queue a card from hand (hidden)
| { type: 'undo'; player: 0 | 1; cardId: string } // un-queue it (refunds energy)
| { type: 'commit'; player: 0 | 1 } // lock in this turn (empty = pass)
| { type: 'timeout'; player: 0 | 1 } // HOST ONLY: turn timer expired
| { type: 'concede'; player: 0 | 1 };
```
**Queuing a play**
- `play` reserves the card's cost from this turn's energy. It checks the energy, the lane's room (cards already in the lane plus queued plays must stay under `laneCapacity(state, lane)`), and, once the Portal is revealed, whether it accepts the card (`canPlay`).
- Your plays reveal in **your queue order**, interleaved with the opponent's (see below).
- `undo` and new plays are allowed until the seat sends `commit`.
**Resolving a turn**
- When both seats have committed, `applyAction` resolves the turn (DESIGN §5.4–5.5):
1. Carry is set: min(`energyCarryMax`, energy not reserved by queued plays). Energy spent on a play that later fizzles is still spent.
2. **Interleaved reveal:** the initiative seat's 1st play, then the other seat's 1st, then the initiative seat's 2nd, and so on. When one seat runs out, the other's remaining plays reveal in order. Each card gets the next global `seq` number, and its on-reveal effects (keyword, then Portal) run at once.
3. End-of-turn effects run: the initiative seat's cards first, then lanes left→right, then `seq` order.
4. On the last turn, scoring follows (§7). Otherwise the next turn starts:
- initiative **alternates**
- any Portal scheduled by `portalReveal` is revealed
- energy = `energyByTurn[t-1]` + carry + `bonusEnergy`
- each seat draws
- the underdog bonus applies (only if `underdog` is on)
- **A play that is illegal at reveal** (the lane filled up, or a Portal rejects the card) fizzles. The card stays in hand and its energy is lost. A `fizzle` log event records it.
**Timeouts.** `timeout` is how a host enforces `turnSeconds`. It commits whatever is queued. A timeout with an empty queue counts toward `afkTurns`. When that count is reached, the seat concedes and the result has `reason: 'afk'`. An ordinary `commit` resets the count.
**Seat ownership.** The engine does not authenticate anyone. The host must check that the seat in `action.player` is the caller's.
## 3. Cards
```ts
interface GameCard {
cardId: string; // unique within a deck: the NFT card hash, or `loaner:<pageId>:<n>`
pageId: number | null; // at most maxCopiesPerPage cards share a pageId; null (unpinned loaners) → counted per cardId
title: string;
faction: 'Space' | 'Nature' | 'People' | 'Science' | 'History' | 'Culture';
rarity: 'Common' | 'Rare' | 'Epic' | 'Legendary'; // cosmetic in v1
cost: number; // 0–6 (uncheckedCostCap already applied by derivation)
power: number; // 0–30, base power
home: 'attention' | 'depth' | 'reach';
trend: -1 | 0 | 1; // frozen at match start
keyword: string | null; // an id in KEYWORDS
source: 'owned' | 'loaner';
check?: 'verified' | 'unverified' | 'unchecked' | 'loaner'; // host plausibility result (§1.10), for logs
}
```
- **The engine takes already-derived cards and never reads raw stats.** The council asked for this, because snapshot stats are self-reported (#1, critic brief A.1).
- Derivation (DESIGN §1 and §10.1: percentile → fame → cost → power, home, trend, keyword and vanillaBonus) lives in `shared/game/derive.ts`, which engine-b owns. Hosts run it on host-frozen, plausibility-checked inputs before calling `createMatch`.
- `CardInput = { cardId, pageId | null, title, category, rarity, snapshot: { stats }, live: { trendPercent, dailyAverage }, check, pinned? }`. `pinned: { cost, home, faction }` replaces §1.1–1.5 **only** when `check === 'loaner'`. `check === 'unchecked'` caps cost at `uncheckedCostCap` (home and fame are kept).
- **Never build `GameCard`s from client JSON.** `validateDeck` only checks shapes and ranges; a client could send cost 1 / power 30. Constructed modes must re-derive on the host.
- Both seats can hold the same loaner `cardId`. The engine keys per seat, so UIs should key cards by `owner + cardId`.
- Modes may adjust `GameCard` fields such as `power` before `createMatch`. The engine does not re-derive them.
- **`check` and `pinned` are host decisions, not player input.** Build loaner inputs only from shipped data (`STARTER_DECKS`,
`previewPool()`, `LOANER_SNAPSHOTS`), and set `check` for owned cards from the host's own §1.10 check. Otherwise a client could
send `check: 'loaner', pinned: {cost: 1, …}` for any card.
- **`obscure` leaks in v1.** `viewFor` does not mask Obscure cards yet. Modes must not put `obscure` in `keywordsEnabled` until it does.
- A keyword that isn't in `ruleset.keywordsEnabled` never runs. Derivation is responsible for not giving that card `vanillaBonus`.
- Lane power = `power` + `bonus` + `trend` + `onTopicBonus[cost]` if on-topic + ongoing keyword and Portal effects, floored at 0.
## 4. Ruleset
`Ruleset` holds the DESIGN §9 keys, **flat** (nothing deeper than one level, apart from `topicPool` entries and the param tables).
- `parseRuleset(unknown)` is a **strict** zod parse: unknown keys are rejected and every number is bounded, so any ruleset that passes terminates within `turns` turns.
- Modes merge overrides onto `DEFAULT_RULESET` and then call `parseRuleset`.
Notes on the §9 keys:
- `topicPool` is inline `TopicDef[] = { id, name, stat, portal: PortalId | null }[]`. `DEFAULT_TOPIC_POOL` is the 9 Topics of §10.3. A Portal id is one of `PORTAL_IDS` (`main-page, trending, featured-article, stub, interwiki, disambiguation`).
- `keywordParams[keywordId]` holds keyword numbers; keys must be in `KEYWORD_IDS`. `portalParams[topicId]` holds Portal numbers, and falls back to `portalParams[portalId]`, so a mode can add a Topic that reuses a Portal kind with new numbers.
- `keywordsEnabled` ⊆ `KEYWORD_IDS` (the 6 v1 ids plus `edition`, `obscure` and `wildcard`).
- `mulligan` must be `false`. Mulligans are not implemented in v1.
- `percentiles`, `starterDecks`, `trendUp/Down`, `trendMinDaily`, `uncheckedCostCap`, `vanillaBonus`, `costThresholds`, `powerByCost`, `verifyStats` and `factionSynergyCap` are for derivation and the host. The core loop doesn't read them.
Validation also enforces:
- `turns × turnSeconds ≤ 240`
- `energyByTurn.length ≥ turns`
- `portalReveal.length = topicCount`
- `openingHand ≤ handLimit`
- `costThresholds` in ascending order
- with `distinctStats`, `topicCount == 3` and the pool has all 3 stats
- the ranges in DESIGN §11 are **stricter** and are enforced by modes (`shared/game/modes.ts`). The engine's bounds only guarantee safety and termination.
## 5. Determinism and the RNG
- The RNG is **mulberry32(seed: uint32)**, as in DESIGN §4. Its state is a single uint32 in `state.rng`.
- Draw order:
1. p0's deck is shuffled.
2. p1's deck is shuffled.
3. **The Topics are drawn.** With `distinctStats`: for each stat in the order `attention, depth, reach`, take `group[floor(next() × group.length)]`, where `group` is that stat's Topics in `topicPool` order. Then the picks are shuffled, and that is the lane order. Without `distinctStats`: the whole pool is shuffled and the first `topicCount` are taken.
4. The initiative coin is flipped: p0 has initiative on turn 1 (and every odd turn) if `next() < 0.5`; otherwise p1 does. After that, initiative strictly alternates.
5. Keyword and Portal hooks draw in resolution order.
- Shuffles are Fisher–Yates with `i` from n−1 down to 1 and `j = floor(next() × (i+1))`.
- `ongoing` power hooks get a throwaway RNG copy, so computing power never changes the match RNG.
## 6. State and views
The full `MatchState` (plain JSON) is for the **host only**:
```ts
interface MatchState {
schema: 'wikideck-match/1'; engineVersion: number; ruleset: Ruleset; seed: number; rng: number;
turn: number /* 1-based */; phase: 'planning' | 'over'; initiative: 0 | 1;
players: [PlayerState, PlayerState]; // deck, hand, energy, queue, committed, afk
lanes: { topic: TopicDef; portalRevealed: boolean; cards: [CardInPlay[], CardInPlay[]] }[];
revealSeq: number; log: LogEvent[]; actions: Action[]; result: MatchResult | null;
}
interface PlayerState { deck; hand; energy; queue: { cardId; lane }[]; committed; afk; carry }
interface CardInPlay { card: GameCard; owner: 0 | 1; lane: number; turn: number; seq: number; bonus: number }
```
`PlayerView`, returned by `viewFor(state, seat)`, is what a seat may see. In it, "you" is always index 0 and the opponent is index 1:
```ts
interface PlayerView {
you: 0 | 1; ruleset; turn; phase; initiative;
energy: number; energyLeft: number; hand: GameCard[]; queue: { cardId; lane }[]; committed; deckCount;
opponent: { handCount; deckCount; committed }; // §5.3: nothing about their queue (cards, count or energy) until reveal
lanes: { topic: TopicDef /* stat always; id/name/portal masked until portalRevealed */, portalRevealed, cards: [mine, theirs],
cardPower: [number[], number[]], power: [mine, theirs], leader: 'you' | 'opponent' | null }[];
result: MatchResult | null;
}
```
`state.log` is an append-only event list for animation and replays. Event types: `turnStart`, `portalRevealed`, `underdog`, `draw`, `reveal`, `fizzle`, `effect`, `concede` and `matchOver`. It contains both seats' draws, so a host must filter it per seat before sending it.
## 7. Results
```ts
interface MatchResult {
winner: 0 | 1 | null; // null = draw
reason: 'topics' | 'totalPower' | 'draw' | 'concede' | 'afk';
topicsLed: [number, number]; lanePower: [number, number][]; totalPower: [number, number]; turns: number;
}
```
Scoring runs after the last turn's End phase. Lane powers include revealed "at scoring" Portals, and each lane's leader uses the consensus tie-break (§6). Winner, per DESIGN §7:
1. The seat that leads more Topics.
2. If tied on Topics, the seat with more total power.
3. If that is also tied, a draw.
## 8. Keywords and Portals (extension contract)
`KEYWORDS` and `PORTALS` in `shared/game/keywords.ts` (engine-b) are registries of hook objects. v1 keywords (`V1_KEYWORDS`, one per faction via `FACTION_KEYWORD`): viral, rising, local, encyclopedic, nostalgic, translated; optional (`OPTIONAL_KEYWORDS`, off by default): edition, obscure, wildcard. Portals: the 6 kinds of §10.3. The header of that file is normative. Every hook gets `ctx = { state, rng, params, lane, log, power(card) }`.
**Keywords**
- `onReveal(ctx, self)` and `endOfTurn(ctx, self)` may change `bonus` on cards in `ctx.state`.
- `ongoing(ctx, self, target) → number` must be **pure**, and must not call `ctx.power`.
- Keyword hooks run only if the keyword is in `keywordsEnabled`.
**Portals** have the same hooks, plus:
- `onPortalReveal(ctx)`
- `canPlay(ctx, card, seat) → boolean`, checked when a play is queued and again at reveal
- `capacity(ctx) → number`
- `allOnTopic: true`
- `atScoring(ctx) → [delta0, delta1]`
Portal hooks act only once the Portal is revealed.
**Rules for all hooks**
- Hooks must not move or remove cards (DESIGN §10).
- Any randomness must come from `ctx.rng`. No v1 keyword uses it.
## 9. Testing
- `tests/game-core.test.ts` (engine-a) covers the core loop:
- the RNG reference value, ruleset bounds and setup
- hidden information and hidden Portals
- energy carry, interleaved reveal and alternating initiative
- on-topic power, the consensus tie-break, and Portal capacity, fizzle and at-scoring hooks
- disabled keywords, AFK and concede, and draws
- the random-play property: matches terminate, replay exactly, and serialise after every action
- `tests/game-keywords`, `game-derive`, `game-starter`, `game-property`, `game-golden` (engine-b): every keyword and Portal, derivation, starter decks, properties (termination, determinism, round-trip, no input mutation, legality, power ≥ 0, capacity, no hand leak) and the DESIGN §12 golden match (19–15 / 15–19 / 13–10, p0 wins).
+272
View File
@@ -0,0 +1,272 @@
# Hosting WikiDeck game modes
Anyone running a WikiDeck node can host their own **game modes**: their own twist on the
rules, played with WikiDeck cards on their node, with visible credit for hosting them.
Hosting a popular mode is the night-1 incentive to run a peer.
- Schema and validator: `shared/game/modes.ts`
- Node API: `server/game/` (mounted at `/api/game` by `server/testnet/http.ts`)
- Example modes: `server/game/modes/*.json` (shipped in the node image)
- Tests: `tests/modes-*.test.ts`
Everything here is **off-chain**. Hosting a mode never touches blocks, transfers, trades,
validation or NFT ownership. A match never moves or burns a card.
## 1. Modes are data, never code
A mode is a JSON document, at most 8 KiB:
```json
{
"schema": "wikideck.mode/1",
"id": "trending",
"version": 1,
"name": "Trending",
"description": "Live Wikipedia pageview trends power cards up or down.",
"tags": ["live-data"],
"rules": { "trendUp": 25, "trendDown": -25 },
"cards": { "preset": "trending", "percentPerPoint": 25, "maxBonus": 3, "penalise": true },
"decks": { "kind": "starter" },
"limits": { "maxConcurrentMatches": 32, "turnSeconds": 90, "lobbyMinutes": 15 }
}
```
| Field | Meaning |
|---|---|
| `id`, `version` | Lowercase slug and a version integer. Bump `version` whenever the rules change, so old match records and leaderboards stay attributable (`<origin>#<id>@<version>`). |
| `name`, `description`, `tags` | Plain text only: no control characters or `<` `>`. |
| `rules` | **Overrides** on top of the engine's `DEFAULT_RULESET` (see `docs/game/ENGINE_API.md`). Objects merge one level deep; arrays and numbers replace. The merged result then goes through the engine's strict `parseRuleset`, so the engine alone decides what's legal (bounded turns, Topics, deck sizes, known keywords and portals only; see DESIGN.md §9 for the keys). |
| `cards` | A card-mapping **preset** from a fixed list. `standard` leaves cards as dealt. `live` freezes each card's live trend arrow (DESIGN §1.6, using the ruleset's `trendUp`/`trendDown`) at match start. `trending` does the same and also adds a bounded power bonus from the live pageview trend. The mode only picks a preset and tunes its bounded numbers. |
| `decks` | `starter` (everyone gets a seeded starter deck), `draft` (snake draft from a host-seeded pool of preview cards) or `constructed` (bring your own; see "Not yet" below). |
| `limits` | Concurrency and lobby expiry for this mode on this host. `turnSeconds` only applies to engines without their own turn timer; with tonight's engine, the ruleset's `turnSeconds` (plus 5 s network grace) is the timer. |
The schema is strict: unknown fields are rejected, as are prototype keys
(`__proto__`, `constructor`, `prototype`) anywhere in the document, nesting deeper than two
levels inside `rules`, more than 48 rule keys, and oversized strings and arrays. A peer
**never** evaluates, imports or runs anything another peer sends. Nothing in a mode can name
a file, a URL, a script or a function.
## 2. Running your own modes
Configure modes with environment variables on your node:
| Variable | Default | Meaning |
|---|---|---|
| `WIKIDECK_MODES` | `classic` | Comma list of built-in ids (`classic`, `trending`, `draft`) and/or paths to `.json` files. `none` hosts nothing (the node still lists its peers' modes). |
| `WIKIDECK_MODES_DIR` | unset | A directory: every `*.json` in it is loaded as a mode. |
| `WIKIDECK_GAME_OPERATOR` | your origin's host | The name shown as **"Hosted by …"**. |
| `WIKIDECK_GAME_NOTE` | unset | A short note shown next to your name, e.g. "EU, best effort". |
| `WIKIDECK_GAME_PEERS` | `WIKIDECK_PEERS` | Peer origins whose modes to list. Use the peers' **public** origins, because players open the peer's own site to play there. |
| `WIKIDECK_GAME_TREND_INTERVAL_MS` | `20000` | Background trend warm-up pace for `trending` modes (minimum 15000). |
| `PUBLIC_ORIGIN` | `http://127.0.0.1:$PORT` | Your node's identity as a host. |
At most 16 modes per node. A bad mode file stops the node at startup with a readable error,
because rules are validated once with the engine's parser when the node starts.
Example for the night-1 testnet (three nodes, different modes):
```yaml
# node-a: the base rules plus Trending
WIKIDECK_MODES: classic,trending
WIKIDECK_GAME_OPERATOR: Testnet A
WIKIDECK_GAME_PEERS: http://192.168.1.32:3102,http://192.168.1.32:3103
# node-b: the draft
WIKIDECK_MODES: draft
WIKIDECK_GAME_OPERATOR: Testnet B
# node-c: everything, including a custom file
WIKIDECK_MODES: classic,trending,draft,/app/data/modes/blitz.json
```
### Writing a new mode
Start from `server/game/modes/classic.json`, change `id`, `name` and `description`, then
override rules. Some ideas within tonight's rules:
- **Blitz:** `"rules": { "turns": 4, "energyByTurn": [3, 4, 5, 6] }`
- **On-topic:** `"rules": { "onTopicBonus": [0, 2, 2, 3, 3, 4, 4], "trendUp": 10, "trendDown": -10 }`
- **Slow play:** `"rules": { "turns": 6, "turnSeconds": 40 }` (the engine caps turns × turnSeconds at 240)
- **Big decks:** `"rules": { "deckSize": 16, "maxLoaners": 16 }` with `"decks": { "kind": "draft", "poolSize": 40 }`.
Starter decks are 12 cards, so a non-draft mode with another deck size is rejected at startup.
Check that a mode parses with `npm test` (the built-in modes are tested), or load it on a
local node: it fails at startup if the engine rejects it.
## 3. The node API
Everything is under `/api/game`, behind the node's existing `/api` admission budget and
global rate limit, plus stricter per-route limits.
| Method | Path | What it does |
|---|---|---|
| GET | `/modes` | **Publish:** this node's host identity, total matches hosted, and each mode with live stats (players online, open lobbies, live matches, matches played, distinct players). |
| GET | `/network/modes` | **Discover:** this node's list plus each configured peer's list, fetched as data. |
| GET | `/hosts/leaderboard` | Hosts ranked by matches hosted (this node plus peers). |
| GET | `/modes/:id/leaderboard` | Per-mode player leaderboard (wins, losses, draws). |
| GET | `/modes/:id/lobbies` | Open lobbies waiting for an opponent. |
| POST | `/modes/:id/lobbies` | Create a lobby: `{ name, secret }` → `{ lobbyId, seat: 0, token }`. |
| POST | `/lobbies/:lobby/join` | Join: `{ name, secret }` → `{ lobbyId, seat: 1, token }`. The match (or draft) starts. |
| GET | `/lobbies/:lobby` | Status. With `Authorization: Bearer <token>`: your `PlayerView` and your `legalActions`. |
| POST | `/lobbies/:lobby/pick` | Draft pick: `{ index }`. |
| POST | `/lobbies/:lobby/actions` | Play: `{ action }`, which must equal one of your `legalActions`. |
| POST | `/lobbies/:lobby/resign` | Resign (or cancel a waiting lobby). |
| GET | `/matches/:matchId` | A finished match record: seed, both decks, the full action log and the result. |
**Players.** The browser generates a random `secret` once and keeps it in local storage. The
host only ever stores and shows `sha256("wikideck-player:" + secret)` as the player id, so a
leaderboard name can't be claimed by someone else who uses the same display name. Each seat
also gets a fresh bearer token per lobby; only its hash is kept. When eom.dev accounts land
(#7), the player id can come from the account instead.
**Playing on a peer's mode.** The browser's CSP keeps API calls same-origin, so the lobby
shows peer modes with "Hosted by …" and links to the peer's own site to play there. No node
proxies another node's matches.
### Restarts
Live lobbies, drafts and matches survive a node restart. The node saves them to
`$WIKIDECK_DATA_DIR/game/live.json` every 15 s and on SIGTERM/SIGINT, and restores them on start.
Player tokens keep working, and turn and draft timers restart from the moment the node is back.
A lobby is dropped instead of restored if its mode is no longer hosted with the same `version`
and rules, so a match never continues under different rules. A hard crash can lose up to 15 s
of moves.
## 4. Credit for hosts
Every mode shows **"Hosted by &lt;operator&gt;"** (with the node's origin), its player counts,
its own leaderboard, and the node's total matches hosted. `/hosts/leaderboard` ranks nodes by
matches hosted.
Tonight these numbers are **informational**: each node reports its own counts. See section 6
for how they could become rewards without trusting the host's word.
## 5. Security notes
### Limits
| Limit | Value |
|---|---|
| Mode document | 8 KiB; 48 rule keys; rule nesting depth 2; strings ≤ 64 chars inside rules |
| Modes per node / per peer response | 16 |
| Peer `/api/game/modes` response | 192 KiB, read with a streaming byte cap; 3 s timeout; no redirects; cached 30 s; at most one refresh in flight |
| Peers | at most 16, operator-configured origins only (HTTPS, or HTTP with `WIKIDECK_ALLOW_PRIVATE_PEERS=1`) |
| Lobbies + matches in memory | 512 per node; `maxConcurrentMatches` per mode; 3 per player |
| Match length | engine bounds (`turns` ≤ 12, turns × turnSeconds ≤ 240 s) plus a hard 2,000-action cap |
| Rate limits per IP | 240/min reads and plays, 60/min lobby create/join/resign, 30/min peer-discovery routes (on top of the node-wide 300/min `/api` limit and the RPC admission budget) |
| Request body | the node's 68 KB JSON limit |
| Wikipedia lookups (`trending`) | background only: one page per 20 s by default (`WIKIDECK_GAME_TREND_INTERVAL_MS`, minimum 15 s), at most 256 queued, never per request |
### Mode validation
A mode is rejected at node startup, never handed to the engine half-checked, if:
- it fails the mode schema (section 1), or
- the merged ruleset fails the engine's strict `parseRuleset`, or
- it fails the host-side checks in `server/game/engine-binding.ts`: `distinctStats` requires
`topicCount` 3; every `portalReveal` turn ≤ `turns`; `energyByTurn` has an entry per turn;
`turns × turnSeconds ≤ 240`; `powerByCost` has 7 entries; no keyword whose hidden information
`viewFor` doesn't mask yet (`obscure`) is enabled; or
- its deck policy can't produce legal decks (starter decks illegal under its rules, or a draft
pool smaller than 2 × `deckSize`).
### Wikipedia load (`verifyStats`)
Our Wikipedia egress is shared and rate-limited, so tonight the host makes **no Wikipedia
requests in any request or match path**:
- Hosted decks are loaners (starter or drafted), so there are no owned-card stats to verify,
and `verifyStats` never triggers a fetch.
- When owned decks arrive, verification must use the DESIGN §1.10 immutable cache keyed by
`(pageId, revisionId, snapshotMonth)`, with `check: 'unchecked'` and `uncheckedCostCap` as the
normal path when the cache is cold, not a per-card fetch.
- `trending` reads only the background-warmed trend cache (see "Not yet").
### Cheating
- **The client is never trusted.** Each action must deep-equal one of the engine's
`legalActions` for the token's seat, so you can't act for the other seat, play cards you
don't hold, overspend energy or act after the match ends.
- **Hidden information stays on the host.** Each seat only ever receives `viewFor(state, seat)`;
the opponent's hand and staged plays never leave the server. Spectators see status only.
- **Seeds are the host's.** The host picks a random seed when the lobby opens, so players can't
grind for a favourable shuffle. The engine hashes it to 32 bits: fine against players, but
not a secret worth relying on against a host.
- **Stalling:** when a seat runs out of turn time, the host applies the engine's `timeout`
action (auto-commit). After `afkTurns` idle timeouts the engine forfeits that seat. Timeouts
are in the action log, so they replay. A stalled draft pick is made automatically.
- **Sybils and win-trading:** anyone can make a new player id, and two ids controlled by the same
person can feed each other wins. Tonight's leaderboards are for fun. Accounts (#7) or
wallet-signed identities would raise the cost.
- **Card input.** The host builds every deck from shipped engine data (starter decks and the
loaner pool). It never accepts cards, `check` values or pinned stats from client JSON.
- **Card stats.** Card numbers come from the engine's card mapping. Minted stats are self-reported
by the minter (see council-critic in #3); a mode can only pick a mapping preset, not supply numbers.
### What a malicious host can and can't do
**Hosted PvP trusts the host.** The host holds both players' hidden plans (hands and queued plays)
and relays them; it decides the seed; and it would be the one to mark card stats as checked or
unchecked (DESIGN §1.10). Nothing tonight lets players verify a match *while* it's being played.
So a host is **trusted for the matches it hosts**, and only those. Play on hosts you trust, and
treat leaderboards on any other host as that host's claims.
A malicious host **can**:
- lie about its stats, its matches hosted and its leaderboards;
- rig its own matches: pick seeds, peek at hands, or ignore the rules;
- drop or stall your match, or record a false result.
A malicious host **can't**:
- touch your cards: matches are off-chain, and the host never receives a key or signature;
- run code on other nodes: its modes are inert data, validated with hard limits;
- make other nodes fetch anything except its own `/api/game/modes`, or make them fetch it
in a way that slows them down (byte cap, timeout, cache, single flight);
- publish modes under another node's origin: discovery pins the origin it actually fetched;
- see your player secret: it only receives it once, over its own origin, and stores the hash.
**Use a different secret per host** if you don't trust a host (the UI can derive one per origin).
**Audit trail.** Every finished match record publishes the seed, both decks and the full action
log, so anyone can re-run it with the engine's `replay(ruleset, decks, seed, actions)` and check
the result. A host that lies about a result can be caught; a host that peeked at hands can't.
## 6. Proposal: on-chain host rewards (not tonight)
Rewarding hosts on-chain is a **protocol change**, so it's out of scope for night 1. This is a
proposal for later, to be reviewed like any consensus change.
**Goal:** pay hosts for matches real players actually played, without trusting the host's count.
1. **Signed match results.** Both players sign the final result
(`matchId`, `<origin>#<id>@<version>`, seed hash, action-log hash, winner) with their wallet
keys. The host signs too. A result needs all three signatures, so a host can't invent matches
or results alone, and players can't deny a match they played.
2. **Replayable evidence.** The record keeps the seed, decks and action log. Anyone can replay it
with the pinned engine version to check the result. Disputes are settled by replay, not by vote.
3. **Batched claims.** Once per epoch, a host submits one transaction committing a Merkle root of
its signed results. A new `hostClaim` transaction type would carry the root and a count. This is
the protocol change: a new transaction type, its validation and its reward rule.
4. **Rewards that resist farming.** A host's reward per epoch grows with the number of **distinct
paired wallets** (not matches), each wallet counts toward at most one host per epoch, wallets
must hold at least one minted card older than N blocks, and the total is a small, capped
share of block rewards. That makes sybil farms cost real mining work.
5. **Challenges.** For a window after a claim, anyone can post a replay that contradicts a
committed result, or proof that a wallet was double-counted. A successful challenge voids
the host's claim for that epoch.
Open questions: who pays (a share of block rewards versus a fee from players), how to stop
colluding players and hosts from farming one another, and whether the engine version must
be pinned in consensus.
## Not yet
- `constructed` decks of **owned NFTs**: the policy exists, but tonight the host deals
starter decks for it. Playing with your own cards needs a wallet-signed ownership proof
for the lobby (and then the engine's `cardFromStats`). That's a follow-up once accounts and
wallet linking (#7) settle.
- `trending` never calls Wikipedia when a match starts. The node warms a cache-only trend
store in the background, one page every 20 s (configurable, never faster than 15 s), through its existing enrichment cache (6 h TTL),
starting with the preview-card pages. A card whose page isn't warm yet gets no bonus. This
keeps load on the shared, rate-limited Wikipedia egress tiny: a few dozen pages per 6 hours per node.
- Spectating and rematches.
- **Separate game RPC budget.** `/api/game` sits behind the node's global `/api` admission
budget (`WIKIDECK_RPC_REQUESTS_PER_SECOND`, 10 writes/s node-wide), so a flood of game
traffic can delay chain RPC (and the reverse). At human speed this is fine. The fix is to give
game routes their own `RpcBudget`, which means mounting them ahead of the shared budget in
`server/testnet/http.ts`: a change to that shared file, so it's proposed in #6 rather than made
tonight.
File diff suppressed because it is too large Load Diff
File diff suppressed because it is too large Load Diff
+57
View File
@@ -0,0 +1,57 @@
# Mainnet-first, gated candidate
The repository defaults to a **mainnet candidate**, not an audited or launched production network. Shipping code does not establish independent miners, economic security, or an operational public network. Market and payment features have been removed from the card-only product.
**Current production decision: NO-GO.** See the [2026-10-20 security review](security-review.md) and [operator evidence register and launch procedure](operator-launch.md). The read-only `launch:preflight` package script (or node entry-point `--launch-preflight`) reports local gates and unresolved blockers without starting a node or setting approvals. It deliberately exits nonzero while production readiness is blocked.
## What is implemented
| Property | Mainnet-first default | Explicit development fixture |
| --- | --- | --- |
| Selection | `WIKIDECK_NETWORK=mainnet` (or unset) | `WIKIDECK_NETWORK=testnet` |
| Network ID | `wikideck-randomx-mainnet-1` | `wikideck-randomx-testnet-6` |
| Protocol | Version 6 | Existing fixed local testnet version 6 |
| Genesis timestamp | Fixed launch default: 2026-11-01 00:00 UTC (`1793491200`); ready to start now | Fixed at 2026-10-19 00:00 UTC |
| Ledger directory | mainnet | testnet |
| Browser wallet | IndexedDB name derived from exact chain ID | Separate chain identity |
| New backup | v2 header authenticates exact chain ID | Same v2 format; legacy v1 is accepted only on the original testnet |
Both profiles use real RandomX validation, cumulative-work branch selection, the 15-minute target, mint maturity, signed NFT transfers, atomic card trades, and SQLite persistence. The mainnet candidate has no premine, token migration, hidden checkpoint, or fake-PoW option. Changing the network identifier and candidate genesis derives distinct chain/genesis IDs, RandomX initial keys, signature domains, and consensus manifest checks. Existing testnet values and data remain unchanged.
**Cards and backups do not migrate across chain identities.** Browser storage and backup headers are bound to the exact chain ID; peers reject mismatched manifests, and transactions are checked against the active chain. The selected profile and genesis are pinned at startup/browser build, never silently adopted from an RPC response. The original fixed-testnet v1 backups remain supported only on that testnet.
## Inspect without launching
The node entry point accepts `--network-manifest`. With `WIKIDECK_NETWORK=mainnet` set in the process environment, this prints the candidate network/chain/genesis/timestamp and exits **before** opening databases, starting a RandomX engine, listening, or contacting peers. It does not require activation credentials or approval and does not launch anything.
Browser builds select `VITE_WIKIDECK_NETWORK=mainnet`; server/miner processes select `WIKIDECK_NETWORK=mainnet`. These must match. `WIKIDECK_GENESIS_TIME` is also part of the chain identity. Docker's network/genesis **build arguments** pin the browser and immutable image manifest; a runtime variable cannot alter an already compiled UI. To choose another genesis, set it for the image build and runtime together, publish the resulting manifest, then use that immutable image and exact timestamp for every node and browser. A runtime-only genesis override fails closed. A mismatched UI rejects the node rather than signing for it.
## Explicit launch gates
[compose.yaml](../compose.yaml) and [compose.mainnet.yaml](../compose.mainnet.yaml) are mainnet-first definitions, use separate named volumes, bind only loopback for an external HTTPS reverse proxy, and do **not** auto-restart a failed candidate. They have no default peers, approval, or public origin. The default service therefore exits until the operator supplies real deployment values. The testnet Docker fixture is separate and explicit. **This code change does not establish or certify a public mainnet.**
Starting the candidate node or native miner requires all of:
1. `WIKIDECK_MAINNET_LAUNCH` equals the exact candidate chain ID printed by the manifest command.
2. `WIKIDECK_MAINNET_RISK_ACK=operator-accepts-unaudited-risk` records the operator's explicit acceptance that no external audit is planned and residual risks remain. This is not a safety certification; the manifest honestly reports `audited: false`.
3. The candidate genesis timestamp has arrived. If the operator changes the timestamp or consensus manifest before launch, the chain ID also changes and the old approval is invalidated. Freeze and publish the final manifest before launch.
4. An HTTPS `PUBLIC_ORIGIN` and at least two distinct configured peer origins. Public deployments require public HTTPS peers by default. An operator-controlled private candidate cluster may explicitly set `WIKIDECK_ALLOW_PRIVATE_PEERS=1` to permit private/HTTP peer RPC; this does not relax browser HTTPS, chain approval, risk acknowledgment, or genesis-time gates. Configuration count cannot prove the peers are online, independent, or sufficiently distributed; verify that operationally. Existing public peer DNS/origin controls still apply when the private opt-in is off, and HTTPS certificate verification is never disabled.
For Kubernetes public peering, give every node a stable, unique public HTTPS endpoint. For an explicitly private candidate, separate single-replica Deployments, Services and exclusive PVCs may use distinct ClusterIP DNS origins with the private opt-in. Each validator lists the other two internal Services, never itself. The browser-facing node needs a TLS ingress/reverse proxy; internal-only validators may share that HTTPS browser origin without their own ingress. NetworkPolicy must restrict peer RPC to the selected validator pods on target TCP port 3001. Private HTTP RPC is unauthenticated and unencrypted unless protected by an independently configured mesh; do not use it across untrusted networks. Three validators in one cluster/storage failure domain are not independent operators or proof of production readiness.
See the [Kubernetes mainnet preparation checklist](../deploy/kubernetes/README.md) for the deployment contract and the values needed before manifests can be finalized.
Unknown profile names fail closed. A direct import of the HTTP entry point is also guarded. Library-level manifest/consensus/storage inspection remains possible for tests and review; activation gates are accident prevention and operator risk acceptance, not a security boundary against someone who edits the source.
## Required before an actual production launch
- **Internal readiness review:** track code findings and regression evidence in-repository. No external audit is planned; the operator accepts the additional risk and must not describe the candidate as audited or certified.
- **Security model:** quantify hashpower, chain-rewrite cost and majority attacks. The inherited easy initial target and ten-block retarget are **not production-calibrated**. Choose reviewed launch difficulty/retarget behavior and test start-up/stall/recovery scenarios. Independent mining participation and sufficient hashpower are not provided by running multiple containers on one host.
- **Adversarial rehearsal:** sustained multi-host simulations, partitions/eclipses, competing branches, deep reorganizations, clock skew, bad peers, load/mempool saturation, corrupted state and restart recovery. Test the candidate rules with no valuable assets.
- **Operations:** independently administered public HTTPS nodes and miners across failure domains, pinned release artifacts and dependencies, firewall/reverse proxy/rate-limit policies, bounded storage and sync strategy, metrics, alerts, rotation and emergency-response procedures.
- **Recovery:** verified offline wallet recovery, consistent ledger snapshots, full replay/revalidation, documented incident and fork handling. No silent ledger reset or reuse of testnet data.
- **Launch governance:** select the actual genesis date, freeze the identity/configuration, publish binaries and configuration hashes, agree peer/miner readiness and upgrade policy, and explicitly record the operator's risk acceptance. Do not treat the flag as proof of safety or peer readiness.
- **User and legal review:** no monetary value or security is promised. NFT rights do not imply article/image copyright ownership. Publish experimental-use, privacy, support, and abuse-report terms appropriate to the release.
## Still not production-ready
This remains experimental software. Nodes retain the DAG and overlays in memory, rebuild state roots over the full NFT set, and revalidate proofs at startup. The HTTP UI trusts its connected node rather than independently validating the chain. There is no automatic peer discovery, established bootstrap infrastructure, formal finality, independent audit, or production abuse system. Mainnet selection does not remove any of these limitations. See the [card-only production plan](card-only-production-plan.md) for the current readiness gates.
+40
View File
@@ -0,0 +1,40 @@
# WikiDeck Android app (Night 1)
A Capacitor 7 shell (`dev.eom.wikideck`) around the React UI. It talks to a WikiDeck node over native HTTP.
![Android emulator screenshots](android/screens.jpg)
## Build a debug APK
You need JDK 21 (a full JDK with `javac`), the Android SDK (platform 35 and build-tools 35) and `ANDROID_HOME`.
```
android/build-debug.sh # testnet build: VITE_WIKIDECK_NETWORK=testnet, genesis 1792411200, node http://192.168.1.32:3101
adb install -r android/app/build/outputs/apk/debug/app-debug.apk
```
- **Settings:** override any of them with env vars, for example `VITE_NODE_URL=http://192.168.1.32:3102 android/build-debug.sh`.
- **Network:** without `VITE_WIKIDECK_NETWORK=testnet`, the bundle is a **mainnet** build.
- **Signing:** debug signing only. Nothing is published.
## How it works (`src/native.ts`, imported first in `main.tsx`; does nothing on the web)
- **Node URL:** relative `fetch('/api/…')` and `fetch('/auth/…')` calls go to `VITE_NODE_URL`. The default is the Night-1 testnet node. A runtime override is available through `setNodeUrl()`, stored in localStorage.
- **Native HTTP (`CapacitorHttp`):** fetches go through Android's HTTP stack. There's no CORS, and no WebView `Origin` header, so the node's POST origin check passes without any server change.
- Verified: a POST from the app returns 400 (validation), while the same POST with `Origin: https://localhost` returns 403.
- The interceptor merges headers (`content-length: 0, 384`). `native.ts` drops that header so `StateClient`'s size check works, and byte limits are still enforced while streaming.
- **Mining:** RandomX is desktop-only. `isNativeApp()` hides the mining panel, and `html.native-app` lets CSS target the app.
- **SSO:** "Log in with eom.dev" opens Dex in a Custom Tab (`setNativeBrowser(Browser)`). Dex redirects to `dev.eom.wikideck:/oauth/callback` (intent filter, `singleTask`), and `completeLogin()` has the node exchange the code.
- **Android behaviour:**
- The **back button** closes an open dialog first, then goes back in history, then minimizes the app.
- The **keyboard** uses `adjustResize`.
- **Edge-to-edge** margins are on, with a dark window behind the system bars.
## Emulator (inspiron-3670, user space)
The SDK, AVD and JDK are under `~/worker/jobs/night/mobile-android/{sdk,avd,jdk}`, and `env.sh` sets the paths.
```
emulator -avd wikideck -no-window -no-audio -gpu swiftshader_indirect # headless, KVM
adb exec-out screencap -p > shot.png
```
To inspect the WebView, forward DevTools with `adb forward tcp:9222 localabstract:webview_devtools_remote_$(adb shell pidof dev.eom.wikideck)`.
## Known gaps
- **Real SSO login:** needs an eom.dev account and wasn't done tonight. The deep-link round trip was verified up to Dex's `invalid_grant` for a dummy code.
- **Touch targets in the Play UI:** the dashed "Play here" lane zone doesn't respond to taps; only the lane plate does (reported on #5).
- **Mining copy on the home hero:** "Explore mining" still shows in the app.
+53
View File
@@ -0,0 +1,53 @@
# Night 1: module boundaries
Owner: `night-lead`. This file is the rule when two PRs touch the same path. If you need to edit a path you don't own, comment on the owner's issue (`[from <you>]`) or keep your change to the "shared hook" listed below.
## Who owns what
| Path | Owner | Issue |
|---|---|---|
| `docs/game/DESIGN.md` | council (designer-a, designer-b, critic) | #1 |
| `docs/game/CARD_DATA.md`, `docs/game/card-sample.json` | council-data | #2 |
| `shared/game/` (types, RNG, state machine, rules, keywords, `Ruleset`, `starterDeck`) | engine-a (core, API), engine-b (keywords, effects) | #3 |
| `docs/game/ENGINE_API.md` | engine-a | #3 |
| `shared/game/modes.ts` | modes | #6 |
| `server/game*` (e.g. `server/game/`, `server/game-*.ts`) | modes | #6 |
| `docs/game/HOSTING.md` | modes | #6 |
| `src/game/` (screens, lobby, deck builder, match, tutorial) | ui-a | #5 |
| `src/game/ai/` (browser bots, easy and normal) | ui-b | #5 |
| beta-fix files outside `src/game/` (landing page, `Card.tsx`, etc.) | ui-b, one PR per upstream issue | #5 |
| `shared/bots/`, `scripts/sim/` (bots and the simulation runner) | sim-balance | #4 |
| `docs/game/BALANCE.md` | sim-balance | #4 |
| `compose.testnet.yaml` overrides, `deploy/night/` (testnet config, per-node mode files) | sim-testnet | #4 |
| `server/account*` (e.g. `server/account/`), `src/auth/` | sso | #7 |
| `android/`, `capacitor.config.ts` | mobile-android | #8 |
| `ios/`, `docs/night/IOS.md` | mobile-ios, macos-fetch | #9 |
| `docs/night/` (except `IOS.md`), this file, `REPORT.md` | night-lead | #10 |
## Tests
`scripts/run-tests.mjs` only picks up **top-level** `tests/*.test.ts`. Name your files with your team as a prefix so they never collide:
`tests/game-*.test.ts` (engine), `tests/modes-*.test.ts`, `tests/ui-*.test.ts`, `tests/bots-*.test.ts`, `tests/account-*.test.ts`.
## Shared hooks (the only places two teams touch the same file)
1. **`server/testnet/http.ts`**: modes and sso each export an Express `Router` from their own files and add **exactly one** `app.use(...)` mount line just **above** the `/api` 404 handler (`app.use(['/api','/p2p'],(_req,res)=>{res.status(404)...`). Nothing else in this file. If SSO needs a CSP change (`connect-src` for `https://dex.eom.dev`), keep it to the `helmet` directive and say so in the PR.
2. **`src/App.tsx`**: ui-a owns it. Others get one import and one element each (sso: a login button or provider; mobile: hide the mining controls behind a `isNativeApp()` check from `src/game/` or `src/auth/`, never by deleting code).
3. **`package.json` and `package-lock.json`**: additive changes only (new dependencies or scripts). Don't reformat or bump existing versions. Lockfile conflicts are resolved by night-lead regenerating the lockfile.
4. **`shared/enrichment.ts`**: read-only for everyone tonight. Map card data into game stats inside `shared/game/`.
## Off limits (consensus and protocol; PRs touching these are rejected)
`server/chain.ts`, `shared/protocol.ts`, `shared/trade-offer.ts`, `shared/chain-identity.ts`, `shared/network.ts`, `server/testnet/consensus.ts`, `server/testnet/ledger.ts`, `server/testnet/canonical-index.ts`, `server/testnet/randomx.ts`, `server/testnet/miner.ts`, `server/testnet/peers.ts`, `native/`.
If you think you need a change there, write it up in your issue instead.
## Dependency direction
`shared/game/` imports nothing from `server/` or `src/` and does no I/O. `shared/game/modes.ts`, `shared/bots/`, `server/game*`, `src/game/` import from `shared/game/` through the API in `ENGINE_API.md`, not from its internals.
## Merging
- Branch `night/<team>/<topic>`, PR into `night/2026-11-16`, squash-merged by night-lead after night-review approves and `npm test` and `npm run build` pass.
- Small PRs merge faster. Rebase on the integration branch before asking for review if it moved under you.
- After 01:10 (code freeze), only fixes merge.
+99
View File
@@ -0,0 +1,99 @@
# WikiDeck Night 1: team handbook
Read this whole file before doing anything. It's the contract between about 17 agents on 4 machines.
**Coordinator:** the Claude session on tower-plus (Eric's interactive session).
**Lead/integrator:** `night-lead` on tower-plus.
**Window:** 2026-11-16 22:45 until **01:30 EST, hard stop**.
**Milestone:** https://gitea.eom.dev/claude/wikideck/milestone/63
## Goal
Turn WikiDeck into a game:
1. A **fun, robust, balance-tested card game**, playable in the browser against a private testnet.
2. **eom.dev accounts:** log in through Dex SSO, and link a self-custodial wallet to your account.
3. **Peer-hosted game modes:** node operators publish their own modes as an incentive to run a peer.
4. **An Android app**, built and running in an emulator, and an **iOS app** (macOS VM; Xcode sign-in waits for Eric).
## Hard rules (no exceptions)
- **Where to push:** only to `git@gitea.eom.dev:claude/wikideck.git`. Never push to `WikiDeck/wikideck` (upstream) or to `main` on this fork.
- **Branches:** every change goes on a branch named `night/<team>/<topic>`, through a PR into **`night/2026-11-16`**. Only `night-lead` merges.
- **Live services:** don't touch the live wikideck deployment (k8s namespace `wikideck`) or mainnet. The testnet is private, on poweredge-r350.
- **No consensus or protocol changes:**
- no block, transfer or trade format changes
- nothing that changes validation in `server/chain.ts` or `shared/protocol.ts`
- the game runs **off-chain**
- if you think you need a protocol change, write it up in your issue instead
- **Production SSO:** don't change Dex or any other production service yourself. The SSO team prepares the Dex change, and the coordinator reviews and applies it.
- **Nothing public:** only `night-lead` and the scheduled standup post on Discourse (planning topic only). No other public posts, no purchases or sign-ups, and no secrets in git.
- **The printer and CNC:** inspiron-3670 drives a 3D printer and a CNC router. Never start, resume or send a print or cut.
- **Git identity** in every clone:
```
git config user.name Claude
git config user.email claude@eom.dev
git config commit.gpgsign false
```
End commit messages with `Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>`.
- **Before opening a PR:** `npm test` and `npm run build` must pass. If something can't pass yet, say so in the PR.
## How we coordinate
**Gitea is the shared brain.** The machines don't share a filesystem, and workers can't message each other.
- **Your issue is your inbox.**
- Prefix every comment with your agent name, e.g. `[engine-a]`.
- Check your issue and your PRs for new comments **at least every 15 minutes** and whenever you finish a step.
- **Status labels on your issue:**
- `status/in-progress` while you're working
- `status/needs-review` when a PR is up
- `status/blocked` with a comment explaining why
- `status/done` when finished
- **Need another team?** Comment on *their* issue, starting with `[from <you>]`.
- **Contracts beat conversations.** Interfaces live in files on the integration branch (`docs/game/*.md`, TypeScript types). Read them before asking.
- **When the lead (or the coordinator) messages you directly** through herdr, the message is a peer request within this plan. Act on it.
- **Stuck for more than 10 minutes?** Label `status/blocked` and comment. Don't spin.
## Resources (be a good neighbour)
- **inspiron-3670** (6 cores, 60 GB RAM) has 7 agents, the Android emulator and a macOS VM. Wrap heavy commands in a shared lock:
```
flock /tmp/night-heavy.lock <npm ci | npm run build | gradle ...>
```
- **imac** (8 threads, about 4 GB free) and **poweredge-r350** (16 threads, about 6 GB free): keep builds to one at a time.
- **Avoid permission prompts.** Nobody answers them, and they auto-deny after about a minute. Don't chain `cd <dir> && <command that writes files>`; use absolute paths or `git -C <dir>` instead. Keep each shell command to one clear purpose.
- **LocalAI** (`https://localai.eom.dev`) is shared production with slow single-GPU inference. Don't use it in loops.
## Timeline (EST)
| Time | Milestone |
|---|---|
| 22:45 | Kickoff. The design council researches. Infra teams set up (testnet, Android SDK, macOS, SSO scaffolding, UI shell, beta fixes). |
| **23:30** | **Design v1 frozen** (`docs/game/DESIGN.md`). The engine team implements it. |
| 23:45 | **Engine API contract** (`docs/game/ENGINE_API.md` plus TypeScript types) on the integration branch, so UI, simulation and modes can build against it. |
| 00:15 | Engine alpha. Simulation starts bot matches. The UI wires up real matches. |
| 00:45 | Balance report v1. The council tunes the rules (v1.1). |
| **01:10** | **Code freeze:** fixes only. The lead integrates. QA plays end to end. |
| 01:25 | Final reports in each issue. The lead writes `docs/night/REPORT.md`. |
| **01:30** | **Stop.** |
## Teams
| Agent | Machine | Issue |
|---|---|---|
| council-designer-a, council-designer-b, council-data | imac | Game design council, card data analysis |
| council-critic, night-lead, night-review, macos-fetch | tower-plus | Design council (critic), integration and QA, iOS (download) |
| engine-a, engine-b | inspiron-3670 | Game engine |
| ui-a, ui-b | inspiron-3670 | Game UI and AI opponent, beta fixes |
| modes | inspiron-3670 | Peer-hosted game modes |
| mobile-android, mobile-ios | inspiron-3670 | Android app, iOS app and macOS VM |
| sim-testnet, sim-balance, sso | poweredge-r350 | Testnet and simulation, eom.dev accounts |
## What the cards hold (game design input)
Each card's data comes from `shared/enrichment.ts`: `PageEnrichment` and `MintStatsSnapshot`.
- **Identity:** `title`, `description`, `wikidataId` (structured facts: instance-of, dates, coordinates...) and `categories[]`
- **Size and history:** `pageBytes` and `revisionId`
- **Reach:** `languageCount`, the number of language editions
- **Pageviews:** monthly views, `total`, `dailyAverage` and `trendPercent`
- **Derived 0–100-ish scores:** `stats.attention`, `stats.depth` and `stats.reach`
- **Two values for everything:** an immutable **at-mint snapshot** and **live current values** that change over time
- **Duplicates:** the same article can be minted more than once, each copy with its own snapshot
Two facts to design around:
- **New players own no cards.** The design must give them a way to play immediately, for example starter decks of preview or loaner cards that don't touch NFTs.
- **Cards are NFTs.** A match must never move or burn cards on-chain tonight.
+109
View File
@@ -0,0 +1,109 @@
# iOS app and macOS VM (Night 1, mobile-ios, issue #9)
## Status at a glance
- **macOS VM:** QEMU/KVM on inspiron-3670, OSX-KVM style. Sequoia (15) recovery booted, disk erased (APFS volume `machd`),
licence accepted, and **"Reinstall macOS Sequoia" was downloading the full OS from Apple** over inspiron's Wi-Fi
(installer estimate ~2 h 50 min from 23:33). Per the coordinator, the VM is **shut down cleanly before 01:30**
with the disk kept. See "Progress at shutdown" below for exactly where it got to.
- **No Apple ID** was created or used. Nothing signed in to Apple.
- **iOS Capacitor project:** `ios/` (Capacitor **7.6.x**, Swift Package Manager, no CocoaPods), app id
`dev.eom.wikideck`, URL scheme `dev.eom.wikideck` registered for the SSO redirect.
- **Build script:** `scripts/ios/build-sim.sh` (run inside the VM once Xcode is installed). Not yet run: needs Xcode.
## Where things are (inspiron-3670, outside git)
| Path | What |
|---|---|
| `~/worker/jobs/macos/boot-macos.sh` | Starts the VM: 8 GB RAM, 4 vCPU, q35, OpenCore, VNC 127.0.0.1:5911, SSH fwd 127.0.0.1:10022 |
| `~/worker/jobs/macos/mac_hdd_ng.qcow2` | 80 GB macOS system disk (keep this) |
| `~/worker/jobs/macos/BaseSystem.img` | Sequoia recovery (3,221,110,784 bytes), converted from macos-fetch's verified `BaseSystem.dmg` |
| `~/worker/jobs/macos/OpenCore-osxkvm.qcow2` | OpenCore bootloader from kholia/OSX-KVM |
| `~/worker/jobs/macos/OVMF_CODE_4M.fd`, `OVMF_VARS.fd` | UEFI firmware and this VM's NVRAM |
| `~/worker/jobs/macos/vnc.sh` | `vncdotool` wrapper (`./vnc.sh capture x.png`, `./vnc.sh type ...`) |
| `~/worker/jobs/macos/guest/provision.sh` | Runs **inside** macOS: Command Line Tools + Node 24 (no Apple ID) |
| `~/worker/jobs/macos/CREDENTIALS.txt` | Local macOS user + password (mode 600, never in git), once the account exists |
## Start the VM again (resume)
On inspiron-3670:
```
cd ~/worker/jobs/macos
setsid ./boot-macos.sh > qemu.log 2>&1 < /dev/null & # installer image still attached
# after macOS is fully installed you can detach it: NO_INSTALLER=1 setsid ./boot-macos.sh ...
```
- In the OpenCore picker pick **`macOS Installer`** / **`machd`** (whichever is offered after the interrupted install;
if only "macOS Base System" shows, the download restarts: pick it → **Reinstall macOS Sequoia** → disk `machd`).
- Clean shutdown from outside: `echo system_powerdown | socat - UNIX-CONNECT:$HOME/worker/jobs/macos/qemu-monitor.sock`
(or `kill $(cat ~/worker/jobs/macos/qemu.pid)` if the guest is hung).
### See it (VNC through an SSH tunnel from tower-plus)
```
ssh -N -L 5911:127.0.0.1:5911 eric@inspiron-3670 # on tower-plus
# then open a VNC viewer on tower-plus at localhost:5911 (no VNC password; it is bound to localhost only)
```
SSH into macOS (after Remote Login is on): `ssh -p 10022 <user>@127.0.0.1` on inspiron, or tunnel `-L 10022:127.0.0.1:10022`.
Note: the VNC keyboard via `vncdotool` drops Shift (everything types lowercase, `|` becomes `\`). A normal VNC viewer is fine.
## Eric's checklist from here (~10 min of attention, plus download time)
1. **Finish the install:** start the VM (above), open VNC, let the installer finish (several reboots, pick `machd` /
`macOS Installer` in OpenCore each time if it doesn't auto-boot).
2. **Setup Assistant:** country, *skip* Apple ID ("Set Up Later"), create a local admin user. Put the user and password in
`~/worker/jobs/macos/CREDENTIALS.txt` (`chmod 600`).
3. **Enable SSH:** System Settings → General → Sharing → Remote Login: on.
4. **Prerequisites (no Apple ID):** copy and run the guest script:
```
scp -P 10022 ~/worker/jobs/macos/guest/provision.sh <user>@127.0.0.1:
ssh -p 10022 <user>@127.0.0.1 ./provision.sh
```
5. **Sign in with your Apple ID and install Xcode** (App Store, or developer.apple.com/download). Launch it once,
accept the licence, install the **iOS platform + a Simulator runtime** when prompted.
6. **Build and run:**
```
ssh -p 10022 <user>@127.0.0.1
git clone -b night/2026-11-16 https://gitea.eom.dev/claude/wikideck.git && cd wikideck
scripts/ios/build-sim.sh # optional arg: simulator name, default "iPhone 16"
```
It runs `npm ci`, `npm run build`, `npx cap sync ios`, `xcodebuild` for the simulator (no signing), installs and
launches `dev.eom.wikideck`, and saves a screenshot to `~/wikideck-ios-sim.png`.
## App notes
- Scaffold (`capacitor.config.ts`, Capacitor deps in `package.json`) is owned by mobile-android (#8); `ios/` builds on it.
Capacitor **7** to match Android (Node 20 on inspiron).
- **Swift Package Manager**, not CocoaPods: `ios/App/CapApp-SPM/Package.swift` is managed by `npx cap sync`; Xcode
resolves `capacitor-swift-pm` itself. No Ruby/CocoaPods needed in the VM.
- **Testnet build:** `build-sim.sh` sets the same defaults as `android/build-debug.sh`: `VITE_WIKIDECK_NETWORK=testnet`,
`VITE_WIKIDECK_GENESIS_TIME=1792411200`, `VITE_NODE_URL=http://192.168.1.32:3101` (override via env).
- **Plain HTTP to the LAN node:** `Info.plist` sets `NSAppTransportSecurity` → `NSAllowsLocalNetworking` (local/IP hosts
only, not arbitrary HTTP), because iOS ATS blocks cleartext by default. From the VM's simulator, 192.168.1.32 is reached
through QEMU's user-mode NAT on inspiron.
- **Plugins:** `@capacitor/app` and `@capacitor/browser` (from #35) are wired into `CapApp-SPM/Package.swift`.
- **Origin check:** mobile-android routes native requests through `CapacitorHttp`, which sends no `Origin` header, so
`server/testnet/http.ts` needs no change. Without it, the iOS WebView origin would be `capacitor://localhost`.
- **SSO (#7):** `Info.plist` registers the `dev.eom.wikideck` URL scheme; redirect `dev.eom.wikideck:/oauth/callback`
returns to the app via `@capacitor/app`'s `appUrlOpen`.
## Progress at shutdown (01:06, 2026-11-17)
- **23:21** recovery `.dmg` received and verified; **23:29** VM booted Sequoia recovery; **23:33** disk erased
(APFS `machd`), licence accepted, "Reinstall macOS Sequoia" started on `machd`.
- **23:33 → 01:03** the installer downloaded the OS over inspiron's Wi-Fi at ~75–120 MB/min. The VM disk grew from
21 MB to **9.7 GB**. Shortly before 01:03 the progress bar reached the end of its first phase and the estimate
switched to "About 7 hours and 58 minutes remaining" (the next phase; the estimate on screen had sat at
"2 hours 52 minutes" for the whole download, so don't trust it).
- **01:03** ACPI power-off was ignored by the installer, so at **01:05** I pressed **Cancel → Quit** in the installer
(it says: *"When you're ready to install, select Reinstall macOS in the menu and Continue again"*), then
Apple menu → **Shut Down** from Recovery. QEMU exited cleanly at **01:06**. `qemu-img check`: no errors.
- **macOS is not installed yet.** No user account exists, so there is no `CREDENTIALS.txt` and SSH isn't on yet.
The partial download stays on `machd` (9.7 GB qcow2); the installer may reuse it or download again (that was accepted).
**To resume** (inspiron-3670):
```
cd ~/worker/jobs/macos && setsid ./boot-macos.sh > qemu.log 2>&1 < /dev/null &
```
Then, over VNC: OpenCore → **macOS Base System** (or a `macOS Installer` entry if one appears) → **Reinstall macOS Sequoia**
→ Continue → Agree → disk **`machd`**, and continue with "Eric's checklist" above (step 1).
## What's verified
- `BaseSystem.dmg` sha256 matches macos-fetch's `SHA256SUMS`; converted `BaseSystem.img` is 3,221,110,784 bytes.
- VM boots on inspiron (KVM, no host `ignore_msrs` change needed): OpenCore → Sequoia recovery → Terminal, Disk Utility.
- Disk erase (APFS) and the installer download started.
- `npx cap add ios --packagemanager SPM` (Capacitor 7.6.10) generates the project on Linux; bundle id `dev.eom.wikideck`.
- **Not verified:** the macOS install finishing, Xcode, `build-sim.sh`, the app in the Simulator.
Binary file not shown.

After

Width:  |  Height:  |  Size: 143 KiB

+144
View File
@@ -0,0 +1,144 @@
# WikiDeck Night 2: design brief (locked decisions)
**Night 2:** 2026-11-17, 21:30 to **02:00 EST, hard stop**.
This brief merges three sources:
- the coordinator's plan, in `docs/night2/PLAN.md`
- four reviews (designer, engineer, data and critic), in `docs/night2/reviews/`
- **Eric's decisions on 2026-11-17**, which override anything in the reviews
Where a review disagrees with this file, **this file wins**. Where this file is silent, follow the reviews, in this order:
1. engineer, for architecture
2. designer, for rules
3. data, for derivation
4. critic, for process
## Eric's decisions (non-negotiable)
1. **A full 40-card, Magic-style game.**
- A deck is exactly 40 cards, and every card in it is an NFT the player owns.
- There are no loaners, Basics, filler or proxies.
2. **Players can't play until they own enough cards.**
- If a player owns fewer than 40 cards, the game shows their progress (for example "23 / 40 cards") and how to mine.
- There is no "practice with borrowed cards".
- The reviews' Basics, loaners, Signature card and the "Collection + Basics" model are **rejected**.
3. **PvP is first-class.** Bot matches exist, but only for eligible players.
4. **The mobile app must be able to mine.**
- Today `isNativeApp()` hides the mining panel, so mobile players can never earn cards.
- Fix it in the Android app.
- Write in `docs/night2/MOBILE_MINING.md` that Google Play and the App Store both prohibit on-device crypto mining, so for now the mining build is a sideloaded APK.
5. **Tone: lean in, Cards Against Humanity style.**
- Living people, tragedies, wars, disasters, crimes, diseases and dictators are all **ordinary cards with full mechanics**.
- There are no `solemn` flags, no special-casing and no softened verbs.
- The game takes itself completely seriously, and the comedy comes from the juxtapositions. Use the critic's S7 deadpan style guide: never wink, never comment on the matchup, and use encyclopedic texture with [n] citations.
- "Casualties" is fine.
- **One rule stays:** all generated text (rules text, match log, recap) is deterministic templates filled **only with real data from the card and the match**. Never use LLM prose, and never invent factual claims about real people (no made-up crimes, quotes or acts). Humour by juxtaposition is fine; fabricated facts about a real person are defamation.
6. **iOS is paused.** Don't touch `ios/` or the macOS VM.
7. **Wikimedia politeness by default** (the engineer's §4 and the data analyst's §5), including the mint-path bug (§9 below).
## Supply reality (design around it, don't fight it)
- **Mainnet:** 12 cards, all in one wallet. Consensus mints at most about 96 cards a day, network-wide. That's Eric's accepted trade-off: the game becomes playable as people mine.
- **Our testnet** (`deploy/night/`, on poweredge-r350):
- Wallets alice, bob, carol and dave own about 50 cards between them, with up to 40 more being mined now.
- The `testnet-ops` agent consolidates cards into **alice and bob** with on-chain transfers, so both reach 40 if supply allows.
- Deck size is **ruleset data** (`deckSize`, default 40). If the testnet can't reach 80 cards, the end-to-end test mode may set a lower `deckSize`. That's a testnet mode file only; production stays at 40.
- **Engine, simulation and balance work** uses the 236 real cards in `docs/game/card-sample.json` as the card pool. Build 40-card random decks from it. This needs no chain.
## Game rules core (designer §2, adapted to 40 cards)
- **Life:** each player starts with 20 Notability; at 0 you lose. Drawing from an empty library loses (the MTG rule). These are ruleset keys.
- **Opening:** hands of 7, with one mulligan (shuffle, draw 7, bottom 1). The first player skips their first draw.
- **Turn:** refresh, draw, main, combat, end.
- **Resources are a fixed income of Citations:** your maximum rises by 1 each turn, to a cap of 10. There are no land cards, so no land screw.
- Locations ramp at most +2 in total.
- Hand limit is 7.
- **Combat:**
- attack the opponent with untapped Figures, with summoning sickness ("Pending review")
- at most one blocker per attacker
- damage is simultaneous
- no trample, no first strike, no stack and no instant speed; Events are sorcery-speed
- **Clock:** 60 s per turn plus a 5-minute bank. No turn cap is required. If the sims show stalls, add `turnCap` as a ruleset key.
- **Card types** (designer §3 and data §1):
- **Figure:** a creature
- **Location:** a permanent with Territory or Garrison
- **Event:** sorcery-speed
- **Institution:** an artifact
- **Doctrine:** an Aura that attaches to a Figure, so *Tetrated dodecahedron* can buff *Cho Yang-ho*
- Classify with data §1's P31 → P279 anchor classes, falling back to categories, then the description, then Doctrine.
- Fix the `\bhuman\b` → People bug.
- Freeze each card's derived profile the first time the card is seen.
- **Factions ("Portals"):** designer §3.2. There are no colour restrictions in deckbuilding.
- **Stats:** data §2. Cost 1–6 comes from the percentile fame curve, with a fixed power+toughness budget per cost, so famous cards are bigger and pricier, not strictly better.
- **Keywords:** designer §3.4 and data §2, at most 2 per card, each triggered by real data:
- Protected
- Featured
- Translated
- Legacy
- Stub
- Well-sourced
- Multilingual
- Local
- **Connected:** the data team's tiers (data §4): Linked, Compatriot, Contemporary and Kin, from cached Wikidata claims and links.
- It gives +1/+1, or +2/+2 for a direct link.
- Aim for it to fire in 15–30% of pairs, tuned in the simulator.
- **Flavour text:** the first sentence of the on-chain `article.extract`.
## Story
- **Match log:** past-tense, encyclopedic and per player (`logFor`), with no hidden-information leaks.
- **Recap:** a "Battle of X (2026)" page after each match, with:
- an infobox: date, belligerents (the players), strength, casualties (the cards destroyed), result
- a narrative assembled from the log with 3–4 template variants per event, seeded deterministically
- References citing every card played, with revision ids
- "WikiDeck" branding, not Wikipedia's
- a shareable URL
## Modes (kept, now on the duel engine)
- **Classic:** constructed from your own 40+ cards. It's ranked.
- **Draft:** both players must be eligible. The pool is drawn from both players' collections, and drafted cards are used for that match only. Nothing moves on-chain.
- **Trending:** live pageview momentum changes stats (designer §5), from a daily refreshed cache.
- **Peer-hosted house rules:** stay data-only, with a versioned schema. Night 1 mode files are rejected cleanly.
## PvP (engineer §3)
- **Seats and ownership:**
- Seats are tied to eom.dev accounts (Dex) and their linked wallets.
- The host verifies **ownership on the server** (`ownership.ts`, using `ledger.nft()`) and builds each card from the chain.
- **Night 1's gaps to close:**
- the browser-side ownership check
- the 128-block snapshot cap
- starter decks dealt in `host.ts:144`
- **Matchmaking:** private lobbies, invite links, quick match, and an Elo ladder keyed to the account.
- **Transport:** keep polling, with a state version so unchanged polls return 204. Don't use long-polling.
## Wikimedia (engineer §4 and data §5)
**One process-wide client:**
- 1 request per second, burst 3, 1 request in flight per host
- priority lanes: mint, then interactive, then background
- `maxlag=5`
- honour `Retry-After`
- a circuit breaker after 3 consecutive 429s
- a daily cap
- a sqlite cache on disk with ETags
- a User-Agent with an operator contact (a config key that operators must set; the testnet uses the project URL)
**The mint path changes transport only:** request parameters stay byte-identical, and the cache is never used for mint data.
**Fix the mint bug:** a 429 on pageviews currently mints a card with `attention: null` forever (`enrichment.ts:145-164`). Fail and retry instead.
**Other changes:**
- Card faces in the client no longer trigger uncached upstream calls on every render.
- An article prefetch pool for mining templates (data §8).
**Tests make zero network calls.**
## Cut order if late
1. the Elo ladder (keep W/L)
2. quick match (keep invites)
3. Trending's live stats
4. the Kin and Contemporary tiers of Connected
5. recap polish (keep the plain log)
**Never cut:**
- the duel core
- 40-card owned-deck validation on the server
- PvP over invite links
- the Wikimedia client and mint fix
- mobile mining
- the recap (the story is the point)
+88
View File
@@ -0,0 +1,88 @@
# WikiDeck Night 2: team handbook
Read this whole file, then `docs/night2/BRIEF.md` (the locked design decisions). Then read the review sections your issue names in `docs/night2/reviews/`.
**Window:** 2026-11-17 21:30 until **02:00 EST, hard stop**.
**Integration branch:** `night/2026-11-17` (from upstream main, which already contains Night 1).
**Lead/integrator:** `night-lead` on tower-plus (it can message every agent with fherdr; inspiron, imac and r350 agents can only reach it through Gitea). The coordinator is Eric's Claude session on tower-plus.
**Milestone:** "Night 2: Magic-style duel, PvP, polite Wikimedia" in claude/wikideck.
**Live log:** the Discourse topic linked in the milestone (only night-lead and the scheduled standup post there).
## Goal
WikiDeck becomes a full 40-card, Magic-style trading card game played with the cards you actually mined:
- PvP first
- deadpan, Cards Against Humanity-style juxtapositions
- an encyclopedic battle recap after every match
- an Android app that can mine
- servers that are polite to Wikimedia by default
## Hard rules (no exceptions)
- **Where to push:** only to `git@gitea.eom.dev:claude/wikideck.git`, on branches named `night2/<agent>/<topic>`, through a PR into **`night/2026-11-17`**.
- Never push to `WikiDeck/wikideck`, `main` or the Night 1 branch.
- Only `night-lead` merges, using squash.
- **No stacked PRs:** branch only from `night/2026-11-17`, and only after your dependency has merged. If you need unmerged code, stub the interface.
- **No consensus or protocol changes:** nothing that changes block, transfer or trade formats, validation in `server/chain.ts`, `server/testnet/consensus.ts` or `shared/protocol.ts`, or what a mint fetches and stores. The game runs **off-chain**, and matches never move or burn cards.
- **Live services:** don't touch mainnet, the live node (k8s namespace `wikideck`), Dex or any other production service. The testnet is private, on poweredge-r350 (`deploy/night/README.md`).
- **iOS is paused:** don't touch `ios/` or the macOS VM.
- **Nothing public:** only `night-lead` and the scheduled standup post on Discourse, in the Night 2 topic only. No purchases, sign-ups or secrets in git.
- **The printer and CNC:** inspiron-3670 drives a 3D printer and a CNC router. Never start, resume or send a print or cut.
- **Wikimedia:** tests make **zero** network calls. Don't script loops against Wikipedia or Wikidata. Use `docs/game/card-sample.json` and cached data. Our IP was rate-limited on Night 1.
- **Git identity** in every clone:
```
git config user.name Claude
git config user.email claude@eom.dev
git config commit.gpgsign false
```
End commit messages with `Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>`.
- **Before opening a PR:** run `npm test` and `npm run build` on **Node 24**. On inspiron and r350, use the `node:24-alpine` Docker image if the host's node is older: `node:sqlite` is required, and on Node 20, 12 tests fail falsely. Don't add npm dependencies; ask night-lead.
- **Tests go in** `tests/<agent>-*.test.ts` (top level).
- **Generated text:** deterministic templates filled only with real card and match data. No LLM prose, and no invented facts about real people (BRIEF, decision 5). Otherwise, lean in: no softening of tragedies or living people.
## How we coordinate
**Gitea is the shared brain.** The machines don't share a filesystem.
- **Your issue is your inbox.**
- Prefix every comment with `[<agent>]`.
- Check your issue and PRs for new comments **at least every 15 minutes** and whenever you finish a step.
- **Status labels on your issue:** `status/in-progress`, `status/needs-review` (PR up), `status/blocked` (with a comment), `status/done`.
- **Need another team?** Comment on their issue with `[from <you>]`.
- **Contracts beat conversations.** Interfaces are files on the integration branch, e.g. `shared/duel/types.ts` and `docs/game/DUEL_API.md`. Read them before asking.
- **Direct messages:** a herdr message from night-lead or the coordinator is a peer request within this plan. Act on it.
- **Stuck for more than 10 minutes?** Label `status/blocked`, comment, and move on to something else.
## Resources (be a good neighbour)
- **inspiron-3670** (6 cores, 60 GB) runs 7 agents and the Android emulator. Wrap heavy commands in `flock /tmp/night-heavy.lock <cmd>` (npm ci, builds, gradle, docker build).
- **imac** (about 4 GB free) and **poweredge-r350** (about 4 GB free, 16 threads): one build at a time, and **keep simulation to 4 workers or fewer** (Night 1 ran out of memory on r350).
- **Avoid permission prompts:** nobody answers them. Don't chain `cd <dir> && <write>`; use absolute paths or `git -C`. Keep one purpose per shell command.
## Timeline (EST)
| Time | Milestone |
|---|---|
| 21:30 | Kickoff. Preflight: Gitea MCP works, Node 24, clone. Comment `starting` on your issue. |
| **22:15** | **Contracts merged:** `shared/duel/types.ts` and `docs/game/DUEL_API.md` (duel-core), plus `docs/game/DUEL_DESIGN.md` v1 (duel-design: keywords, text templates and tone, with BRIEF.md as the base). The `ownership.ts` signature is agreed by 22:00. |
| 23:15 | Engine alpha. The simulation starts. The UI plays against a bot. |
| 00:00 | Wikimedia client merged. Host binds the duel. The PvP UI is wired to the host. |
| 00:45 | Balance v1 (tuning is data-only). Mobile mining runs in the emulator. |
| **01:15** | **Code freeze:** fixes only. Balance re-runs on the frozen head SHA. QA runs PvP end to end on the testnet with alice and bob. |
| 01:45 | Final report comment on every issue. night-lead writes `docs/night2/REPORT.md` and opens the upstream PR (never merges it). |
| **02:00** | **Stop.** Push, comment, exit. |
## Teams
| Agent | Machine | Workstream |
|---|---|---|
| night-lead | tower-plus | integration, merges, Discourse, REPORT, upstream PR |
| night-review | inspiron-3670 | code review (`VERDICT: APPROVED` / `CHANGES`), QA, Playwright end to end |
| duel-core | inspiron-3670 | `shared/duel/` engine: turns, combat, `viewFor`/`logFor`, replay |
| host-pvp | inspiron-3670 | `server/game/` multi-engine host, constructed decks, invites, quick match, Elo |
| ownership | inspiron-3670 | `server/game/ownership.ts` and `collection.ts`, server-side 40-card owned-deck validation |
| ui-duel | inspiron-3670 | `src/game/duel/` board and card faces, 360 px first, bot matches |
| mobile-mining | inspiron-3670 | Android mining (sideload build), MOBILE_MINING.md |
| duel-design | imac | DUEL_DESIGN.md, keywords, rules-text and log and recap templates, tone |
| duel-cards | imac | `shared/duel/{classify,derive,effects,text}.ts`: card profiles from data |
| wikimedia | poweredge-r350 | `server/wikimedia/` polite client, mint-path fix, cache, UA, prefetch pool |
| testnet-ops | poweredge-r350 | testnet for Night 2: autodeploy, wallet consolidation, identity pre-bake, test modes |
| sim-balance | poweredge-r350 | duel bots and simulation, `docs/game/DUEL_BALANCE.md` |
| ui-pvp | inspiron-3670 | collection and eligibility screen, lobby, invite, quick match and ladder UI, battle recap page |
**tower-plus note (night-lead):** tower-plus uses Eric's SSH key, so never `git push` from there. Clone over HTTPS read-only and do every Gitea write (comments, reviews, merges, file commits) through the gitea MCP tools, which are signed in as claude. tower-plus is short on memory: run no builds or tests there; ask night-review on inspiron.
+120
View File
@@ -0,0 +1,120 @@
# WikiDeck Night 2: proposed game redesign (for review)
Status: **draft, under review.** Written by the coordinator on 2026-11-17 from Eric's feedback on Night 1's game, Edit War. Night 2 runs unattended on 2026-11-17 from 21:30 to 02:00 EST.
## Eric's feedback on Edit War (the requirements)
1. **Real cards:** a player's deck is made of the cards they actually mined and own. Loaners existed for the demo; the real game uses real cards.
2. **PvP:** players must be able to play each other, not only the AI. Hosted lobbies exist (`src/game/OnlinePlay.tsx`, `server/game/host.ts`) but are buried.
3. **Story:** Edit War "doesn't really tell much of a story with the actual cards".
4. **Feel:** Magic: The Gathering is the model for how it should feel.
5. **Modes:** keep the different modes (classic, draft, trending); Eric likes them.
6. **Tone:** "The fact that I can use a Tetrated Dodecahedron in a game with Barack Obama is funny. Lean into this. The game should take itself seriously, but the end result is going to be entertaining for this reason." (Eric's original v2 post also mentions using his *Tetrated dodecahedron* card to buff his *Cho Yang-ho* card, and node operators running house rules.)
7. **Wikipedia limits:** our servers must respect Wikimedia rate limits by default instead of getting cut off. Night 1 hit HTTP 429 from our shared IP.
8. **iOS:** paused (Apple licence acceptance). Out of scope.
## Hard constraints
- **Consensus is fixed:**
- No changes to the chain, block format, mint stats snapshot or consensus files (see `docs/night/BOUNDARIES.md`).
- Everything game-related is derived off-chain from what a card already holds.
- **Where work happens:** in the `claude/wikideck` fork, through reviewed PRs. Eric merges upstream.
- **No live deployment:** don't touch mainnet or the live node at wikideck.eom.dev.
- **Budget:** 4.5 hours, about 10 agents across 4 machines (tower-plus, imac, inspiron-3670, poweredge-r350). Night 1 used 17 agents for 2.75 h and about 10% of the weekly usage budget; 75% remains.
## Facts the design must face
- **Card data:** what a card holds is shown in `shared/enrichment.ts` (PageEnrichment / MintStatsSnapshot) and `shared/testnet.ts` (BlockHeader).
- **Article:** pageId, title, extract, image, category, revisionId.
- **Mint snapshot:** pageBytes, languageCount, 6 months of pageviews with trendPercent and dailyAverage, and the attention, depth and reach scores.
- **Night 1 also used:** the Wikidata id, P31 "instance of", and live pageviews, all fetched off-chain.
- Real distributions are in `docs/game/CARD_DATA.md`: compressed scores, a skewed category split, and a reach of 0 for 18% of cards.
- **SUPPLY:**
- One card is minted per block, with a consensus target of one block per 15 minutes: at most about 96 new cards a day, network-wide.
- **Mainnet today has 12 cards in total, all owned by one wallet**, all mined on 2026-11-01.
- A "deck of your own mined cards" in the Magic sense (40–60 cards) is impossible for anyone for a long time.
- **This is the biggest viability question; see Open problem 1.**
- **Cards are unique NFTs:** each card is one specific mint, and the same article can be minted more than once (`pageId` repeats).
- **Night 1 code to reuse:**
- the deterministic rules engine and its rules-as-data style (`shared/game/`)
- the bots and the balance simulator (`shared/bots/`, `scripts/sim/`)
- hosted lobbies and modes (`server/game/`)
- Dex accounts (`server/account/`)
- the Android shell (`android/`)
## Proposed design
### A. Card identity from the article (MTG-style types)
The card type comes from Wikidata P31, with a fallback to Wikipedia categories:
| Article is a… | Card type | MTG analogue |
|---|---|---|
| person, character, animal species | **Figure** | creature |
| place, region, celestial body, structure | **Location** | land (resource) and/or permanent |
| event, battle, election, disaster | **Event** | instant/sorcery |
| creative work, organisation, company, team, product | **Institution** | artifact |
| concept, theory, mathematical object, ideology, field | **Doctrine** | enchantment |
**Factions (colours):** the 6 P31-based factions proposed in CARD_DATA.md, plus a neutral fallback.
**Stats from data:**
- **Power and toughness** come from percentile attention, depth and reach. For example, attention drives power, and depth drives toughness and staying power.
- **Cost** follows the curve: high-stat cards cost more.
**Rules text:**
- Generated deterministically from a fixed table of templates keyed on type, faction, category keywords, trend, language count, dates (birth/death, founding, event date), coordinates, and so on.
- It is written in MTG's terse register.
- **Flavour text:** the first sentence of the article's extract.
### B. Story mechanics
- **Connected:** two cards whose articles link to each other, or share a Wikidata property such as country, field or era, gain a bonus together. Barack Obama with the White House; Tetrated dodecahedron with Polyhedron.
- **Citations:** a resource or counter flavoured as sources.
- **Edit wars and vandalism:** removal and interaction effects.
- **Deadpan tone:** the game is completely straight, like an encyclopedia or an MTG judge. The comedy comes from the matchups, never from jokes in the text.
- **Match log:** written as a dry encyclopedic account.
- **After each match:** a Wikipedia-style article, *"Battle of the Tetrated Dodecahedron (2026)"*. It has:
- an infobox: date, belligerents, commanders (the players), strength (the cards), result
- a short deterministic narrative built from the match log
- a "References" list citing every card's real article
- These recaps are shareable.
### C. Game structure
- **Duel:** two players, a life total, turns with attack and block, and permanents on a battlefield.
- **Resources:** from Locations and/or a fixed per-turn income, designed to avoid MTG's "land screw" (no resource cards, no game).
- **Deck size and composition:** depends on Open problem 1.
### D. PvP first
- **Matchmaking:** a queue, invite-by-link and a ranked ladder (Elo or Glicko).
- **Identity:** the eom.dev account, with linked wallets as the card source.
- **Hosting:** the hosting node runs the match, so there's no client trust.
- **AI opponents:** still available for practice and onboarding.
### E. Modes (kept)
- **Classic:** constructed from owned cards.
- **Draft:** players draft from a shared pool, so they don't need large collections.
- **Trending:** stats boosted by live pageviews.
- **Peer-hosted house rules:** remain data-only.
### F. Wikimedia politeness by default
- **One client:** a single throttled Wikimedia client used by every server-side caller (`server/wiki.ts`, `server/enrichment.ts`, P31 lookups, link lookups).
- **Throttling:**
- a conservative rate limit by default (for example 1 request/s and 1 concurrent request per node, configurable)
- `maxlag`
- honour `Retry-After` with backoff
- batch up to 50 titles per request
- an on-disk cache with ETags and conditional requests
- a daily refresh for live stats
- **User-Agent:** a default UA that identifies the node and an operator contact, as Wikimedia's policy asks.
- **Volume:** a design budget for requests per match and per mint.
## Open problems for the review team
1. **Supply versus "real cards only".** What makes a real-cards game playable when a player owns 0–12 cards and the network mints at most about 96 a day? Options to evaluate (add others):
- (a) A small deck size, such as 7–10 cards, with recursion and reuse.
- (b) Your owned cards are your "commanders" or champions, and a shared pool of common public-domain article cards fills the rest. This conflicts with Eric's requirement, so it would have to be argued.
- (c) Draft as the main mode, where the pool is drawn from the whole chain's minted cards and ownership gives cosmetic credit or a guaranteed pick.
- (d) A collection-as-deck where every owned card counts, scaled by size.
- (e) Card copies or "editions" (no consensus change possible, so off-chain only?).
Recommend one, and say what Eric must decide.
2. Is MTG-style attack/block too complex to build, balance and make readable in one night? What is the smallest rules core that still feels like Magic?
3. Is P31 typing reliable enough? Check against the real sample in `docs/game/card-sample.json` and `docs/game/starter-cards.json`. Plan for unclassified cards, which were 14% of Night 1's sample.
4. Can "Connected" link lookups be done within the Wikimedia budget? Where does the data come from, and when is it fetched and cached?
5. Can the generated rules text be both deterministic and good? Does it need an LLM? LocalAI exists on the cluster, but determinism and the consensus boundary matter.
6. What scope is realistic for 4.5 hours with about 10 agents, and what gets cut first?
+44
View File
@@ -0,0 +1,44 @@
"""How often random card pairs share cheap relations. Usage: python3 -I connected.py <card-sample.json>"""
import json, sys, re, itertools, collections, importlib.util, os
cards = json.load(open(sys.argv[1]))['cards']
spec = importlib.util.spec_from_file_location('ty', os.path.join(os.path.dirname(os.path.abspath(__file__)), 'types.py'))
ty = importlib.util.module_from_spec(spec); spec.loader.exec_module(ty)
GENERIC = re.compile(r'stubs|living people|\d+ (births|deaths)$|articles|wikidata|pages', re.I)
COUNTRIES = ['United States','American','England','English','British','France','French','Germany','German','Japan','Japanese','China','Chinese',
'India','Indian','Brazil','Poland','Polish','Italy','Italian','Spain','Spanish','Austria','Canada','Canadian','Australia','Australian','Russia','Russian',
'Mexico','Mexican','Iran','Pakistan','Bangladesh','Sweden','Swedish','Norway','Estonia','Ukraine','Croatia','Philippines','Cuba','Israel','Portugal','Scotland','Wales','Ireland','Kazakhstan','Malaysia','Argentin','Belgi','Iceland','Bermuda','Thailand','Singapore','Sri Lanka','Slovenia','Moldova','Cyprus','Macedonia','Mauritania','Congo','Gibraltar','Hong Kong','New Zealand','Korea','Indonesia','Aceh']
NORM = {'American':'United States','English':'England','British':'United Kingdom','England':'United Kingdom','Scotland':'United Kingdom','Wales':'United Kingdom',
'French':'France','German':'Germany','Japanese':'Japan','Chinese':'China','Indian':'India','Polish':'Poland','Italian':'Italy','Spanish':'Spain',
'Canadian':'Canada','Australian':'Australia','Russian':'Russia','Mexican':'Mexico','Swedish':'Sweden'}
Y = re.compile(r'\b(1[0-9]{3}|20[0-2][0-9])')
def feats(c):
txt = (c['description'] or '') + ' | ' + ' | '.join(c['categories'])
ctry = {NORM.get(k, k) for k in COUNTRIES if k in txt}
cents = {int(y) // 100 for y in Y.findall(c['title'] + ' ' + txt)}
cats = {x for x in c['categories'] if not GENERIC.search(x)}
p31 = {i['id'] for i in c['instanceOf']}
return dict(ctry=ctry, cent=cents, cats=cats, allcats=set(c['categories']), p31=p31, type=ty.classify(c)[0])
F = [feats(c) for c in cards]; pairs = list(itertools.combinations(range(len(cards)), 2)); N = len(pairs)
def rate(f):
k = sum(1 for i, j in pairs if f(F[i], F[j])); return f'{k}/{N} ({100*k/N:.2f}%)'
print('pairs', N)
print('share any category (incl. generic)', rate(lambda a, b: a['allcats'] & b['allcats']))
print('share specific category', rate(lambda a, b: a['cats'] & b['cats']))
print('share P31 class', rate(lambda a, b: a['p31'] & b['p31']))
print('share P31 excl human/taxon', rate(lambda a, b: a['p31'] & b['p31'] - {'Q5', 'Q16521'}))
print('share country (text proxy for P17/P27)', rate(lambda a, b: a['ctry'] & b['ctry']))
print('share century (text proxy for dates)', rate(lambda a, b: a['cent'] & b['cent']))
print('share type', rate(lambda a, b: a['type'] == b['type']))
print('country OR century OR specific cat', rate(lambda a, b: (a['ctry'] & b['ctry']) or (a['cent'] & b['cent']) or (a['cats'] & b['cats'])))
print('cards with >=1 country', sum(1 for f in F if f['ctry']), 'with >=1 century', sum(1 for f in F if f['cent']))
# deck-level: chance that a random 10-card hand-pair (10 vs 10, own deck) has >=1 connected pair within a deck
import random; random.seed(1)
def deck_hits(pred, size=10, trials=2000):
h = 0
for _ in range(trials):
s = random.sample(range(len(F)), size)
if any(pred(F[i], F[j]) for i, j in itertools.combinations(s, 2)): h += 1
return f'{100*h/trials:.0f}%'
pc = lambda a, b: (a['ctry'] & b['ctry'])
print('10-card deck has >=1 same-country pair', deck_hits(pc), '| specific category', deck_hits(lambda a, b: a['cats'] & b['cats']),
'| country-or-century', deck_hits(lambda a, b: (a['ctry'] & b['ctry']) or (a['cent'] & b['cent'])))
+23
View File
@@ -0,0 +1,23 @@
"""Field availability for rules-text generation in card-sample.json. Usage: python3 -I fields.py <card-sample.json>"""
import json, sys, re, statistics, collections
cards = json.load(open(sys.argv[1]))['cards']; n = len(cards)
Y = re.compile(r'\b(1[0-9]{3}|20[0-2][0-9])\b')
def share(f): k = sum(1 for c in cards if f(c)); return f'{k}/{n} ({100*k/n:.0f}%)'
print('description non-empty', share(lambda c: bool(c['description'])))
print('wikidataId', share(lambda c: bool(c['wikidataId'])))
print('P31 present', share(lambda c: bool(c['instanceOf'])))
print('>=1 non-hidden category', share(lambda c: len(c['categories']) > 0))
print('categories truncated (cllimit 50)', share(lambda c: c['categoriesTruncated']))
print('median categories', statistics.median(len(c['categories']) for c in cards))
print('year in title', share(lambda c: Y.search(c['title'])))
print('year in description', share(lambda c: Y.search(c['description'] or '')))
print('year in categories', share(lambda c: any(Y.search(x) for x in c['categories'])))
print('year anywhere (title/desc/cats)', share(lambda c: Y.search(c['title'] + ' ' + (c['description'] or '') + ' ' + ' '.join(c['categories']))))
print('births/deaths category', share(lambda c: any(re.search(r'\d+ (births|deaths)', x) for x in c['categories'])))
print('establishments/completed category', share(lambda c: any(re.search(r'establishments|completed in|disestablish|opened in', x) for x in c['categories'])))
print('living people', share(lambda c: 'Living people' in c['categories']))
print('stub category', share(lambda c: any('stubs' in x for x in c['categories'])))
print('country/nationality word in desc or cats', share(lambda c: re.search(r'American|British|English|French|German|Japan|China|Chinese|India|Brazil|Poland|Polish|Italy|Italian|Spain|Spanish|Austria|Canada|Australia|Russia|United States', (c['description'] or '') + ' '.join(c['categories']))))
print('legacy category on-chain', collections.Counter(c['legacyCategory'] for c in cards))
print('pageviews available', share(lambda c: c['pageviews']['status'] == 'available'))
print('languagesTruncated', share(lambda c: c['languagesTruncated']))
+59
View File
@@ -0,0 +1,59 @@
"""Proposed percentile -> cost/power/toughness derivation, evaluated on card-sample.json.
Usage: python3 -I stats.py <card-sample.json>"""
import json, sys, collections, statistics, itertools
d = json.load(open(sys.argv[1])); cards = d['cards']; P = d['percentiles']
def pct(s, v): # DESIGN 1.1 counting: cut points strictly below + half of equal
if v is None: return None
cut = P[s]; return int(sum(c < v for c in cut) + sum(c == v for c in cut) / 2)
TH = [26, 39, 51, 63, 78] # v1.1 costThresholds (fame -> cost 1..6)
BUDGET = [None, 3, 4, 6, 8, 10, 12] # P+T per cost, MTG vanilla curve (2/1, 2/2, 3/3, 4/4, 5/5, 6/6)
def derive(c):
a, dp, r = (pct(k, c['stats'][k]) for k in ('attention', 'depth', 'reach'))
vals = [x for x in (a, dp, r) if x is not None]
fame = int(sum(vals) / len(vals) + 0.5)
cost = 1 + sum(t <= fame for t in TH)
b = BUDGET[cost]
kw = 'Multilingual' if r >= 75 else ('Local' if c['stats']['reach'] == 0 else None)
if kw == 'Multilingual': b -= 1 # evasion-like keyword is paid for out of the body
if kw == 'Local': b += 0 # Local = gains +1/+1 next to a same-country Location (conditional, no tax)
share = (a if a is not None else 50) / max(1, (a if a is not None else 50) + dp) # attention vs depth
pw = min(b - 1, max(1, round(b * share))); t = b - pw
return dict(title=c['title'], a=a, d=dp, r=r, fame=fame, cost=cost, P=pw, T=t, kw=kw,
trend=c['pageviews']['trendPercent'], daily=c['pageviews']['dailyAverage'])
rows = [derive(c) for c in cards]; n = len(rows)
def naive(c): # straw man: power = attention tier, toughness = depth tier, cost from their sum
ti = c['tiers']; pw = (ti['attention'] or 1 + 1) // 2 + 0; t = (ti['depth'] + 1) // 2
return dict(cost=max(1, min(6, round((pw + t) / 2))), P=pw, T=t)
nv = [naive(c) for c in cards]
def dominated_pairs(rs, use_kw=False):
k = 0
for x, y in itertools.permutations(rs, 2):
if x['cost'] <= y['cost'] and x['P'] >= y['P'] and x['T'] >= y['T'] and \
(x['cost'] < y['cost'] or x['P'] > y['P'] or x['T'] > y['T']) and (not use_kw or x.get('kw') == y.get('kw')):
k += 1
return k
print('cost distribution', sorted(collections.Counter(r['cost'] for r in rows).items()))
print('power distribution', sorted(collections.Counter(r['P'] for r in rows).items()))
print('toughness distribution', sorted(collections.Counter(r['T'] for r in rows).items()))
print('keywords', collections.Counter(r['kw'] for r in rows))
print('P/T by cost:')
for cst in range(1, 7):
g = [r for r in rows if r['cost'] == cst]
if g: print(' ', cst, len(g), collections.Counter(f"{r['P']}/{r['T']}" for r in g).most_common(6))
pairs = n * (n - 1)
print('strict-dominance pairs proposed', dominated_pairs(rows), 'of', pairs, '; same-keyword only', dominated_pairs(rows, True))
print('strict-dominance pairs naive ', dominated_pairs(nv), 'of', pairs)
# efficiency by fame quintile
q = sorted(rows, key=lambda r: r['fame'])
print('fame quintile: mean cost, mean (P+T)/cost proposed | naive (P+T)/cost')
idx = sorted(range(n), key=lambda i: rows[i]['fame'])
for k in range(5):
sl = idx[k * n // 5:(k + 1) * n // 5]
print(f" Q{k+1} fame {rows[sl[0]]['fame']}-{rows[sl[-1]]['fame']}: cost {statistics.mean(rows[i]['cost'] for i in sl):.2f} "
f"eff {statistics.mean((rows[i]['P']+rows[i]['T'])/rows[i]['cost'] for i in sl):.2f} | naive eff "
f"{statistics.mean((nv[i]['P']+nv[i]['T'])/nv[i]['cost'] for i in sl):.2f}")
top = sorted(rows, key=lambda r: -r['fame'])[:6] + sorted(rows, key=lambda r: r['fame'])[:3]
print('examples (top fame, bottom fame):')
for r in top: print(f" {r['title'][:34]:34s} a{r['a']} d{r['d']} r{r['r']} fame{r['fame']} cost{r['cost']} {r['P']}/{r['T']} {r['kw'] or ''}")
tr = [r for r in rows if r['trend'] is not None and r['trend'] >= 25 and (r['daily'] or 0) >= 20]
print('trending (+25% and daily>=20):', len(tr), f'{100*len(tr)/n:.1f}%')
+79
View File
@@ -0,0 +1,79 @@
"""Classify card-sample.json into Figure/Location/Event/Institution/Doctrine.
Order: P31 label rules (first rule that matches ANY P31 label wins, rule order = priority),
then category keywords, then description keywords, then deterministic fallback.
Usage: python3 -I types.py <card-sample.json> [--list]"""
import json, re, sys, collections
RULES_P31 = [ # (type, regex on a P31 label). Order matters: specific before generic.
('Doctrine', r'aspect of|aspect in|disease|clinical sign|list article|chemical entity'),
('Event', r'election|protest|battle|war\b|storm|nor.easter|cyclone|tornado|earthquake|disaster|event|match|final\b|'
r'season|championship|tournament|\brace\b|grand prix|scandal|annexation|treaty|incident|festival|referendum|'
r'conference|ceremony|time period|delegation'),
('Institution', r'artificial satellite|space station|space telescope|spacecraft|space probe'),
('Location', r'planet|natural satellite|\bstar\b|galaxy|nebula|planetary system|astronomical object|rainforest|reef|'
r'civili[sz]ation|empire|historical country|archaeological site|'
r'human settlement|settlement|village|town|city|commune|municipal|parish|county|district|region|province|'
r'township|hamlet|townland|census area|local government|community of|island|river|lake|mountain|plateau|'
r'waterfall|valley|bay|peak|geographical feature|historic district|park|state park|square|road|highway|'
r'building|castle|mosque|church|temple|cathedral|house|home|museum|station|bridge|monument|historic site|'
r'military base|battery|refinery|shopping|stadium|arch|tower|fort|palace|columbari|council'),
('Figure', r'^human$|fictional character|stock character|person|deity|orisha|loa\b|taxon|species|genus|breed|'
r'^family$|organism'),
('Institution', r'company|business|organi[sz]ation|association|society|party|agency|studio|band|musical group|team|club|'
r'university|college|school|institute|diocese|ecclesiastical|manufacturer|newspaper|magazine|film|album|'
r'song|single|musical work|composition|television|series|book|novel|video game|painting|award|position|'
r'aircraft|ship|steamship|vehicle|automobile|tram|railway line|route|typeface|software|product|brand'),
('Doctrine', r'concept|theory|theorem|mathematical|programming language|language|gene|protein|chemical|compound|'
r'disease|clinical sign|ideology|religion|discipline|field|cuisine|food|dish|drink|aspect|list article|'
r'particle|process|genre|form\b|game\b|branch of|class of|type of|tabular'),
]
RULES_CAT = [ # fallback on non-hidden category titles (lowercase)
('Figure', r'\bbirth|\bdeaths\b|living people|people from|\bspecies\b|described in \d|\bgenera\b|deities|characters'),
('Event', r'\belections\b|protests|\bbattles\b|disasters|\braces\b|\bseasons\b|\bevents\b|scandals|annexations|competitions'),
('Location', r'villages|communes|parishes|local government|populated places|geography|rivers|islands|buildings and structures|bridges|'
r'stations|parks|districts|townships|cities and towns|historic places'),
('Institution', r'companies|organi[sz]ations|\bfilms\b|albums|songs|novels|newspapers|television series|musical groups|'
r'aircraft|ships|typefaces|software|awards|teams|clubs'),
('Doctrine', r'theory|structures$|algebra|terminology|cuisine|drinks|desserts|ecology|languages|diseases|conditions|'
r'concepts|ranks|etymology|maps|atlases'),
]
RULES_DESC = [
('Figure', r'\b(painter|politician|actor|writer|player|born|species|genus)\b'),
('Event', r'\b(race|election|battle|match|protest|storm|war|competition)\b'),
('Location', r'\b(village|town|city|river|island|building|church|columbarium|station|park)\b'),
('Institution', r'\b(company|band|film|album|organi[sz]ation|team|aircraft|ship|newspaper)\b'),
('Doctrine', r'\b(concept|theory|structure|dessert|drink|dish|technique|class of)\b'),
]
TYPES = ['Figure', 'Location', 'Event', 'Institution', 'Doctrine']
def classify(card):
labels = [i['label'].lower() for i in card.get('instanceOf') or []]
for t, rx in RULES_P31:
for l in labels:
if re.search(rx, l): return t, 'p31', l
cats = [c.lower() for c in card.get('categories') or [] if 'stubs' not in c.lower()]
for t, rx in RULES_CAT:
for c in cats:
if re.search(rx, c): return t, 'category', c
desc = (card.get('description') or '').lower()
for t, rx in RULES_DESC:
if re.search(rx, desc): return t, 'description', desc
# deterministic last resort: Doctrine (the residual "idea" class); never hash-random.
return 'Doctrine', 'fallback', ''
if __name__ == '__main__':
d = json.load(open(sys.argv[1])); cards = d['cards']
res = [(c, *classify(c)) for c in cards]
n = len(res)
by_t = collections.Counter(r[1] for r in res); by_src = collections.Counter(r[2] for r in res)
print('n', n)
for t in TYPES: print(f'{t:12s} {by_t[t]:4d} {100*by_t[t]/n:5.1f}%')
for s in ['p31', 'category', 'description', 'fallback']: print(f'src {s:12s} {by_src[s]:4d} {100*by_src[s]/n:5.1f}%')
print('labels w/o p31 match but has p31:', sum(1 for c,t,s,w in res if s!='p31' and c['instanceOf']))
print('cross-tab old faction x new type')
ct = collections.Counter((c['faction'], t) for c,t,s,w in res)
for f in sorted({c['faction'] for c in cards}):
print(f' {f:14s}', {t: ct[(f,t)] for t in TYPES if ct[(f,t)]})
if '--list' in sys.argv:
for c,t,s,w in res:
if [i['label'] for i in c['instanceOf']] == ['human']: continue
print(f"{t[:5]} {s[:4]} | {c['title'][:42]:42s} | {w[:40]}")
+171
View File
@@ -0,0 +1,171 @@
# Night 2 plan: critic / player-advocate review
Reviewer: critic. Sources: night2/PLAN.md, wd/docs/game/{DESIGN,CARD_DATA}.md, card-sample.json, wd/docs/night/{HANDBOOK,BOUNDARIES}.md, Night 1 report (`~/Desktop/night/night-lead/REPORT.md`; it is not in wd/docs/night). Ranked by severity, most severe first.
## Facts the plan doesn't state, but which decide everything
- **Mined cards are random articles.** `server/wiki.ts` uses `generator=random`. Nobody chooses Barack Obama. The real pool looks like card-sample.json: 34% stubs, villages, beetles, French communes, local elections. "Tetrated dodecahedron vs Obama" will be rare. The everyday matchup is *Dichromorpha viridis vs 1954 Gold Coast general election*. That's still funny in a deadpan way, but the design and its examples must be built for it, not for celebrity cards.
- **Supply is about zero, not 96 a day.** 12 cards in 16 days, all in one wallet (Eric's), all from day 1. Mining has effectively stopped.
- **Mobile can't mine.** `App.tsx` hides the mining panel when `isNativeApp()`, so an Android player can never get a real card.
- **14% of sampled cards are living people** (33/236 in "Living people"). Events include ongoing protests with deaths (*2024–2026 Cuban protests*, *2025 Tehreek-e-Labbaik Pakistan protests*), a *sex scandal*, an *annexation*, and *Organization of the Luftwaffe (1933–1945)*. The safety rules (§6) are needed tonight, not someday.
---
## S1 (blocker). "Real cards" + PvP is impossible today. This is Eric's decision, not the agents'.
Under a literal reading of point 1, exactly one person on Earth can play, and PvP has no opponent. Each option, measured against the request:
- **(a) Small decks of 7–10:** literal and honest. But only Eric qualifies, and with exactly 12 cards there's no deckbuilding, which is half of what makes MTG feel like MTG.
- **(b) Owned "commanders" plus a public filler pool:** this *is* the loaner system Eric just said was only for the demo, renamed. If agents choose it without asking, it's a quiet betrayal of the request.
- **(c) Draft from the chain's cards:** the cards are real mined cards, but the pool is the same 12 cards every match. It's degenerate until supply grows. On testnet, about 50 cards, it's fine.
- **(e) Off-chain editions:** prints cards with no proof of work and no consensus. It undermines what "mined" means. Reject it.
- **(d) Collection-as-deck:** this is (a) with scaling. Fine as the mechanism, but it doesn't solve the 0–3 card player.
**Fix:** before 21:30, put these questions to Eric. Per standing preference, agents don't start building until Eric gives an explicit go.
1. "Does Classic and ranked PvP require 100% owned mined cards? (Recommend: yes.)"
2. "Can a player with fewer than N owned cards play casual or practice matches with clearly labelled **proxies**? They're never ranked, never shown as owned, and appear greyed out with the stamp 'Proxy'. (Recommend: yes, for practice, AI and invite-link casual only.)"
3. "Is Draft from the whole chain's minted cards 'real cards'? Ownership would earn the first pick and a credit line in the recap. (Recommend: yes.)"
4. "Can your 12 cards be used as public bot and ghost decks? (They're your wallet.)"
5. "Should we fix the supply side, so mobile users can get real cards, given that mobile can't mine? Options: a desktop-mining CTA, gifting by transfer (an on-chain action by Eric, not by agents), or an unlock condition such as 'mine 1 card to enter ranked'."
**Default if Eric doesn't answer:** take the most literal reading. Ranked and Classic use owned cards only. Practice and AI use proxies, which look different. Draft uses chain-minted cards only. No editions. Record the decision in DESIGN.md §0 in Eric's own words.
**Design implication either way:** the format must be fun with a small deck. Recommend **one format with no deck minimum**: a 10-card deck at most, and if you own fewer, you play what you own plus visible proxies, casual only. The engine must handle running out of cards gracefully. Draw from a reshuffled graveyard (DESIGN-style "recursion") instead of losing to an empty library.
## S2 (blocker). PvP has no population. A ranked ladder tonight is wasted work, and unsafe.
Elo with one eligible player is theatre. Night 1's REPORT (known issues 2–3) also says hosted PvP trusts the host: the host picks the seed, there's no commit–reveal, and the UI log is built from full state and leaks hand contents. A ladder on top of that is cheatable.
**Fix, in order:**
1. **Invite-by-link casual PvP** first, made prominent on the Play screen. Eric can test it with a second account on testnet.
2. **Async "ghost" matches.** You play a bot piloting a real wallet's real deck (with that owner's opt-in), and the result is posted to the owner. This is the PvP that a population of five can actually have.
3. The **queue backfills with a bot after 30 s**, *labelled as a bot*. Never pass a bot off as a human.
4. Defer the ranked ladder until commit–reveal and `viewFor`-only logs land. Cut it tonight.
## S3 (high). An MTG clone built in one night will be worse than Edit War
Edit War had a real hook: each lane is judged by one Wikipedia stat. The plan replaces it with generic power/toughness/cost. MTG is fun because of hand-designed cards, the stack and instant-speed interaction, and deckbuilding across thousands of cards. Tonight has none of that: procedurally generated text, 12 cards, and no stack.
**Fix: name the smallest core that still feels like Magic**, and only that:
- Two players, 20 life, and a resource that grows by 1 a turn automatically. Locations add +1 when played (no land screw).
- Board of 5 per side. Figures attack. The defender makes one block per attacker, chosen by tapping a blocker and then an attacker. No first strike, trample or multi-block in v1.
- Events are **sorcery-speed only**. No stack, no priority passes, no opponent-turn interaction. Every response window doubles PvP round-trips and the time needed for UI work.
- One keyword per card at most, from a list of about 8 (Night 1 used 6, with property tests).
- **Keep a Wikipedia hook that MTG doesn't have**, or this is "MTG with worse cards". Candidate: home stat as the creature's "evasion" (an attention attacker can only be blocked by a card with attention ≥ its own), or the trend arrow as haste. One hook, kept and tested.
- **Keep Edit War playable as a mode.** The code exists; deleting it removes a working fallback.
## S4 (high). 360 px readability
An MTG board (two battlefields, hand, life, resources, rules text, combat arrows) doesn't fit 360 px if the cards carry generated paragraphs.
**Fix:**
- The card face on the battlefield shows: title (2-line clamp), type icon, cost, P/T, one keyword chip. The rules text and flavour live **only** in the tap-to-zoom sheet.
- Rules text has a hard limit of 90 characters, enforced by a test over the whole sample.
- Combat uses a "declare attackers → defender blocks → preview outcome → confirm" sheet, not drawn arrows.
- The QA gate is a 360×640 screenshot of a full board (5 + 5) on Android, attached to the UI PR.
## S5 (high). Night 1 process lessons the plan doesn't apply
What went wrong (REPORT plus handbook):
- The design froze at 23:01, but the balance numbers were measured before the #38 change, so the shipped seat-advantage figure is stale.
- 3 OOMs (tower-plus, and r350 twice).
- imac's Gitea MCP was down all night, so the council had to relay through night-lead.
- Squash merges of stacked PRs caused add/add conflicts.
- Wikimedia 429s came from our own sampler.
- On Node 20, 12 tests failed falsely.
**Fixes:**
- **Freeze rules on a schedule, and gate them on evidence.** Design frozen by 22:30 (one hour, not 45 minutes: this is a bigger redesign). Engine API by 22:45. The balance re-run happens on the final head, after the last engine PR, and its numbers go in the report with the SHA they were measured on.
- **Fewer, wider agents.** 7–8, not 10:
- designer and critic as one council pair
- an engine agent
- a card-derivation and text-generator agent (it owns the safety rules)
- a UI agent
- a PvP/host agent
- a wiki-client agent
- a lead and a review agent
Night 1's 17 agents produced conflict and relay overhead.
- **Pre-flight at 21:30:**
- run `get_me` on Gitea MCP on every machine
- `node:24-alpine` is the only test image
- all heavy work runs under `flock` on inspiron
- no stacked PRs: rebase-and-merge, or wait for the parent to merge
- **No network in tests.** Every test and sim reads card-sample.json and starter-cards.json, plus a new red-team fixture (§6). The **only** process allowed to call Wikimedia tonight is the new throttled client's own smoke test, at 1 request/s.
- **Use testnet for real-cards testing.** It has about 50 mined cards across wallets the team controls. Seed 3–4 test wallets with 1, 3, 8 and 20 cards by transfer *on testnet*. Never simulate mainnet supply by copying Eric's wallet.
- **Mainnet is read-only, and only if needed at all.** Read the 12 cards' public data so Eric's own deck can go into a fixture. Nothing is written.
## S6 (high). Ethics and safety: content rules for the deadpan generator
The joke only works if the game is respectful about the subject and absurd only in the juxtaposition. These rules are **requirements** with tests, owned by the card-text agent:
1. **No free-form generated prose about real people or events.** Templates only. No LLM, including LocalAI. That also keeps the output deterministic. The only article-derived prose is the extract's first sentence, quoted verbatim and attributed.
2. **Figures are never harmed, killed or humiliated in text.** Log verbs for a Figure leaving play come from an editorial whitelist: "was reverted", "left the battlefield", "was archived", "was withdrawn from the record". Banned for any card: killed, died, destroyed (for Figures), executed, assassinated, murdered, humiliated, crushed, slaughtered, raped, lynched. "Destroy target Institution" is fine.
3. **Mechanics never derive from sensitive attributes.** Category-keyword templates for Figures come from an **allowlist** of neutral occupation and field words (politicians, mathematicians, footballers, painters, alumni). Never derive a mechanic from: crime, conviction, scandal, illness, disability, cause of death, ethnicity, religion, sexuality, gender, nationality as a trait, victimhood.
- Country is allowed **only** as a neutral "Connected" link (shared P27/P17). It must never carry a modifier such as "French: −1".
- Birth and death years may appear in the type line ("Figure — 1949–2019"). Nothing triggers on death.
4. **Solemn subjects get a `solemn` flag.** It's set by P31 class (war, battle with casualties, massacre, genocide, terrorist attack, mass shooting, disaster, epidemic/pandemic, aviation or maritime accident, protest with deaths, war crime, the Holocaust) or by category words (massacres, genocides, terrorist incidents, disasters in, deaths in, murders, sexual abuse, victims, suicides).
- Solemn cards are playable. They get one fixed neutral effect ("Commemoration: gain 1 Citation; draw a card"), no damage or removal text, a plain log line ("entered the record"), and they are **never** used to name a battle or shown in a recap headline.
- Events in general never "deal damage to" a Figure. Removal comes from Doctrines and Institutions ("Revert", "Edit war").
5. **Glorification guard.** No "leader", "inspire" or "rally" keywords keyed on heads of state, military leaders or dictators. Every Figure's keyword comes from stats, type and faction only. A *Adolf Hitler* card must be a plain statted Figure with no buff-allies text. Test it.
6. **Living people (BLP):** the flavour text is the extract sentence only, and the recap uses only the card title. Any `Living people` card mentioned in a recap is linked to its article and nothing is asserted about the person. Card actions in text are attributed to "the *Barack Obama* card" or the commander, not to the person.
7. **Mint-time vandalism.** The extract is frozen at mint and may be a vandalised revision. Render flavour from the cached **current** extract when possible. Give node operators a data-only `hiddenFlavour: [pageId]` list and a report link on the card zoom sheet.
8. **Recap pages must not impersonate Wikipedia.**
- No Wikipedia wordmark or puzzle-globe logo, and no "From Wikipedia, the free encyclopedia".
- The header reads "WikiDeck match record", with a hatnote: *"This article is about a WikiDeck match. For the polyhedron, see Tetrated dodecahedron."* That's funny, honest, and keeps us clear of the trademark.
- CC BY-SA attribution for extracts and images, in References.
9. **Player privacy.** Shareable recaps are opt-in per match. Commanders show as the account display name or "Anonymous", never a wallet address unless the player chooses.
10. **Red-team fixture plus a sign-off gate.** `tests/fixtures/sensitive-cards.json` holds hand-written PageEnrichment for: Barack Obama, Donald Trump, Adolf Hitler, Jesus, Muhammad, Yemọja, September 11 attacks, The Holocaust, Sandy Hook Elementary School shooting, 2004 Indian Ocean earthquake and tsunami, COVID-19 pandemic, a living child actor, Pullman porter affair, *2024–2026 Cuban protests*, Cho Yang-ho, and Tetrated dodecahedron.
- A test snapshots their full rules text, log lines and a recap, then asserts the banned-verb list and the solemn rules.
- The snapshot file goes into the morning report **for Eric to read before anything is shared publicly**.
## S7 (medium). Tone: how deadpan fails, with examples
- **Failure modes:**
- *Wacky:* puns, exclamation marks, emoji, winks.
- *Cringe:* the generator "noticing" the joke ("In a surprising twist…").
- *Dull:* logs that read like a stack trace.
- *Mean:* adjectives about the losing side.
- **Rule:** the text never comments on how odd the matchup is. The comedy is entirely in the nouns.
- **Card rules text.**
- DO: "*Tetrated dodecahedron* — Doctrine — Ideas. Figures you control with Depth 7 or more get +1/+1."
- DO: "*Barack Obama* — Figure — People, 1961– · 3/4 · Connected (+1/+1 while you control a Connected card)."
- DON'T: "Yes He Can! Obama gets +2 power because he's the president!"
- DON'T: "This shape is so complex it hurts your opponent's brain."
- **Match log.**
- DO: "Turn 6. *Cho Yang-ho* (4/5) attacked. *Dichromorpha viridis* (1/2) blocked and was reverted. Commander Eric cited *Tetrated dodecahedron*: *Cho Yang-ho* +1/+1 until end of turn."
- DON'T: "Cho Yang-ho SQUASHED the grasshopper lol." Also don't use "killed", or use any adjective (brutal, humiliating, heroic).
- **Recap.**
- DO: "The **Battle of the Tetrated Dodecahedron** was a WikiDeck engagement fought on 17 November 2026 between the forces of Eric and Bot (Normal). It lasted nine turns and ended in a decisive victory for Eric.[1] Contemporary accounts note the deployment of *Barack Obama* in the fourth turn.[2]"
- The infobox field is "Cards withdrawn", **not** "Casualties". Battles are named after the decisive non-Figure, non-solemn card; fall back to "Match of 17 November 2026".
- DON'T: "Obama was humiliated by a shape," or "the bloodiest battle in WikiDeck history."
- **Dull guard:** vary sentences by deterministic seed across 3–4 synonymous templates per event, and use citation numbers [n] that link to the card's real article. That small Wikipedia texture is most of the charm.
## S8 (medium). "Connected" will almost never fire on random articles
Two random articles almost never link to each other. Fetching `prop=links` per card is costly, at hundreds of links a page, and it doesn't fit the 1 request/s budget.
**Fix:**
- Connect cards on cheap shared Wikidata claims fetched once per card at first sight and cached for good: country (P17/P27), century (from P569/P571/P585), and same P31 superclass or faction. Leave out the article link graph tonight.
- Measure the fire rate on card-sample.json (random pairs) and on Eric's 12. Target 15–30% of pairs. If it's lower, the mechanic is invisible.
## S9 (medium). Wikimedia politeness (point 7): good plan, three gaps
1. The plan never says **who decides the production UA**. REPORT next step 8 leaves it to Eric. Ship the default as `WikiDeck/<version> (node: <PUBLIC_ORIGIN>; operator contact required)`, refuse to start the enrichment client without a contact configured, and leave production config to Eric.
2. Rate limits must be **per egress IP**, but all of Eric's nodes share one IP. Use one shared limiter per host IP: a single node acts as the Wikimedia proxy for the cluster, or the budget is split statically.
3. Set a request budget per match: **zero Wikimedia calls during a match.** All P31, claims and trend data are fetched at deck validation or lobby time from the cache. Trending's live stats come from the daily refresh.
## S10 (medium). Modes under the new engine (point 5)
Draft and Trending must be re-specified, not just "kept":
- Draft with a 12-card pool must deal from **all chain-minted cards**, or Eric must decide (S1-3).
- Trending uses the daily cache (S9).
- The house-rules JSON schema changes with the engine. Version it (`rulesetVersion: 2`) so Night 1 mode files on testnet nodes are rejected cleanly instead of misread.
## S11 (low). Story (point 3) depends on the recap
The plan's "story" is Connected plus the recap. Given S8, the recap and log carry the whole of point 3, so prioritise them over the ladder. The recap is cheap (a deterministic template over the match log) and is the most "Eric" deliverable.
## Scope: what gets built, and what gets cut first
**Build, in this order:**
1. The throttled Wikimedia client and cache. Nothing else calls Wikipedia.
2. Card derivation (type, faction, P/T, cost, keyword) with safety rules and the red-team test.
3. The smallest MTG core engine plus bots.
4. A 360 px match UI.
5. Invite-link PvP and ghost or bot matches.
6. The match log and battle recap.
7. Draft and Trending ported.
**Cut first:** the ranked ladder and Elo, the Wikipedia link graph, instant-speed Events, the Citations resource as a second currency (fold it into one keyword), and the "editions" idea.
## Decisions for Eric (to ask before kickoff, all in one message)
1. Real-cards strictness (S1 questions 1–3).
2. Use of his 12 cards as public bot and ghost decks.
3. A mobile path to owning cards.
4. Keep Edit War as a mode? (Recommend: yes.)
5. Sign-off on the safety rules in S6, especially solemn Events and living persons.
6. Production User-Agent and contact.
7. Are shareable recaps public by default? (Recommend: opt-in.)
+272
View File
@@ -0,0 +1,272 @@
# Night 2 plan review: data analyst
Inputs: `docs/game/card-sample.json` (236 random illustrated cards), `starter-cards.json`, `shared/enrichment.ts`, `shared/testnet.ts`, `server/{wiki,enrichment,wikimedia-user-agent}.ts`, `server/testnet/{http,miner}.ts`.
Scripts (all `python3 -I`, offline, no API calls): `night2/analysis/types.py`, `stats.py`, `fields.py`, `connected.py`.
**Caveat:** the type rules were tuned on this same sample, so the in-sample accuracy is optimistic. Anything marked *(memory)* comes from my knowledge of the Wikimedia APIs, not from a live check, because we are rate-limited.
## 1. Typing (P31, then categories, then description, then a residual)
The classifier in `analysis/types.py` checks the rules in **priority order**. The first rule that matches *any* P31 label wins. If nothing matches, it tries non-stub category titles, then the description. If all three fail, the card becomes **Doctrine**, which is the residual "idea" class and is deterministic, never hash-random.
| Type | n | share | Mostly |
|---|---|---|---|
| Figure | 86 | 36.4% | human 64, taxon 16, breed, deity, stock character, family |
| Location | 78 | 33.1% | villages, communes, rivers, buildings, parks, roads |
| Institution | 32 | 13.6% | companies, bands, films, parties, aircraft, ships, awards |
| Event | 21 | 8.9% | elections, protests, storms, matches, team seasons |
| Doctrine | 19 | 8.1% | genes, diseases, maths, cuisine, lists, etymology |
| Decided by | n | share |
|---|---|---|
| P31 | 222 | 94.1% |
| category fallback | 12 | 5.1% |
| description fallback | 2 | 0.8% |
| residual (Doctrine) | 0 | 0% |
- **Unclassified rate:** 0% in-sample, down from 14% ("Other") under Night 1's faction rules. 13 cards have no P31, and one card's P31 matched no rule.
- **Accuracy:** I judge about 8 of 236 (about 3–4%) to be debatable or wrong.
- **Expected out of sample:** about 5–10% would reach the category or residual steps on regex alone. That is why the P279 step below matters.
**Hard or misclassified examples:**
| # | Card | P31 | Result | Problem |
|---|---|---|---|---|
| 1 | Buky, Lunugala, Žižki, Laukhella | human settlement | Location | **Existing bug:** Night 1's `\bhuman\b` faction rule puts these 4 villages in *People*. Use `^human$`. |
| 2 | Lewis and Elizabeth Bolton House | single-family detached home | Location | Night 1 calls it *Works* because the label contains "single". Substring rules bite. |
| 3 | Corruption in Kazakhstan | aspect in a geographic region | Doctrine | "region" pulls it to Location unless an "aspect of/in" rule runs first. |
| 4 | 2023 Seville municipal election | municipal election | Event | "municipal" pulls it to Location unless Event outranks Location. Rule order is load-bearing. |
| 5 | Agda (programming language) | functional PL; free software | Institution | Several P31 labels disagree (Doctrine vs product). Tie-breaking is by type priority, not by label order. |
| 6 | Diocese of Mindelo | Diocese of the Catholic Church | Location | "church" matches. It is really an organisation (Institution). |
| 7 | Region Östergötland | county council of Sweden | Location | A governing body, so Institution. "county" matches first. |
| 8 | Christianity in the 5th century | events in a specific year or time period | Event | A history overview. Doctrine fits better. |
| 9 | NRHP listings in Yates County | Wikimedia list article | Doctrine | A list is not a thing. Consider a neutral "Index" type or rerolling at mint (consensus says no). |
| 10 | Dojikko; Tracheal collapse | stock character; type of animal disease | Figure; Doctrine | A trope as a Figure is fun but debatable. The disease was a Figure until "animal" was removed from the Figure rule. |
| 11 | Figueiredo (Guimarães) | former freguesia of Portugal | Location (via category) | Local admin-unit names defeat English regexes. Only the categories rescued it. |
| 12 | *(memory)* Rosetta Stone (stele), Apollo 11 (crewed spaceflight), ISS (space station) | | residual, residual, Institution | Famous starter cards fall through, or hit "station" and become a Location, unless explicit rules come first. |
**Proposed rules** (as implemented in `analysis/types.py`):
1. **Doctrine overrides first:** `aspect of|in`, `disease`, `clinical sign`, `list article`, `chemical entity`.
2. **Event:** `election`, `protest`, `battle`, `war`, `storm`, `match`, `final`, `season`, `race`, `grand prix`, `scandal`, `annexation`, `time period`, `delegation`.
3. **Institution overrides:** artificial satellites, space stations and space probes.
4. **Location:** settlement and administrative-unit words, geography, celestial bodies, `civilization`, `empire`, `historical country`, built structures.
5. **Figure:** `^human$`, `taxon`, `species`, `genus`, `breed`, `deity`, `character`, `^family$`.
6. **Institution:** organisations, companies, teams, creative works, vehicles, positions, awards, typefaces.
7. **Doctrine:** concepts, theories, genes, languages, cuisine, genre, form, process, `type of`, `class of`.
**Production recommendation** (the regex rules above are the fallback):
- Resolve each P31 class through **P279 (subclass of), up to 4 hops**, against a curated table of about 200 anchor classes. For example: Q5 → Figure, Q16521 → Figure, Q486972 (human settlement) → Location, Q1656682 (event) → Event, Q43229 (organisation) → Institution, Q17537576 (creative work) → Institution, Q151885 (concept) → Doctrine.
- **Class cache:** the sample has only **141 distinct P31 classes in 236 cards**. Class lookups saturate quickly, cost under 0.1 requests per card amortised, and are cached forever.
- **Fallback chain:** P279 anchors → label regex → category regex → description regex → **Doctrine**.
- **Freeze the result per card:** store it in the card profile (§3) so that a later Wikidata edit cannot retype a card someone owns.
**Deck consequence:** Event and Doctrine are scarce.
| Type | P(none in 10 random cards) | P(none in 7) |
|---|---|---|
| Event | 39% | 52% |
| Doctrine | 43% | 55% |
| Institution | 23% | 36% |
| Location | 1.8% | 6% |
| Figure | 1.1% | 4% |
- **Instants and enchantments:** a deck built from random mints often has none. The core loop must not need them.
- **Resources:** this argues for a fixed per-turn income (as plan C says). With Locations as the only resource, about 6% of 7-card decks would have no resource at all.
## 2. Stats: percentiles → cost, power, toughness (`analysis/stats.py`)
**Facts that drive the design:**
- **No independent signal in the raw fields.** pageBytes, languageCount and monthly views are monotonic in depth, reach and attention, so their percentiles are identical. They add nothing beyond the three scores.
- **attention and depth correlate** (ρ 0.61), and **reach is independent** of both.
- **trendPercent is live, small and noisy.** Only 2.1% of cards pass the Night 1 gate (+25% and ≥20 views a day).
- **A new independent-ish signal** is available from the on-chain `months[]`: the volatility (coefficient of variation) of the 6 monthly views, median 0.22. Its ρ is −0.52 with attention, −0.30 with depth and −0.19 with reach. It could drive a "Volatile" keyword.
**Proposal:**
1. `pct_s` for attention, depth and reach, using the §1.1 counting against the sample's 99 cut points (unchanged from Night 1).
2. `fame` = the mean of the three percentiles. Cost = 1–6, using the v1.1 `costThresholds` [26, 39, 51, 63, 78] (unchanged).
3. **The body budget depends on cost only:** P+T = [3, 4, 6, 8, 10, 12] for costs 1–6, which is MTG's vanilla curve (2/1, 2/2, 3/3, 4/4, 5/5, 6/6).
4. **The shape comes from the stats:** power = round(budget × pct_a / (pct_a + pct_d)), clamped to [1, budget−1], and toughness is the remainder. Attention reads as "loud", depth as "sturdy".
5. **Reach gives a keyword, paid for from the body.**
- **pct_r ≥ 75 → Multilingual** (evasion-like), costing −1 from the budget.
- **reach = 0 → Local** (a conditional bonus with a same-country Location, no tax).
- Local covers the 18% of cards with reach 0.
6. **Non-Figures use the same budget as effect size:**
- **Event:** deals ⌈budget/2⌉.
- **Location:** P=0 with T=budget, or a per-turn bonus.
- **Doctrine:** ±1 effects, scaled by cost.
- **Institution:** a utility ability.
7. **Trend:** ±1 power, in Trending mode only, using live data frozen at match start.
**Resulting distribution (236 cards, every card rated as if it had a body):**
| cost | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|
| cards | 30 (13%) | 48 (20%) | 51 (22%) | 48 (20%) | 35 (15%) | 24 (10%) |
| most common P/T | 2/1, 1/2 | 1/3, 3/1, 2/2 | 3/3, 4/2, 2/4, 5/1 | 3/5, 4/4, 5/3 | 5/5, 4/5, 5/4 | 6/5, 6/6, 5/6 |
| Measure | Value |
|---|---|
| power, 1..7 | 38 / 43 / 53 / 33 / 42 / 26 / 1 |
| toughness, 1..7 | 43 / 43 / 50 / 30 / 53 / 16 / 1 |
| keywords | none 139, Multilingual 55, Local 42 |
**"Famous isn't strictly better" check:**
- **Strict dominance** means another card with cost ≤, power ≥ and toughness ≥, and strictly better in at least one.
- **Proposal:** 0 of 55,460 ordered pairs among cards with the same keyword. All 945 raw pairs are cases where a keyword costs one point of body, which is intended.
- **Straw man:** with P and T taken directly from the attention and depth tiers and cost derived from their sum, **3,058 pairs**.
- **Efficiency, (P+T)/cost by fame quintile:** Q1 2.64 (the cheap 2/1s), then Q2–Q5 at 1.90, 1.96, 1.95 and 1.89.
- **Famous cards are bigger and pricier, never more efficient.** Pallas's cat (fame 98) is a cost-6 6/5 Multilingual. Dr. John L. Butler House (fame 8) is a cost-1 1/2 Local.
- **Remaining risk:** in a small-deck game with fixed income, a 6-drop can still dominate the late game. Balance that in the simulator (Night 1's `scripts/sim/`), not in derivation.
## 3. Rules-text inputs (`analysis/fields.py`)
| Field | Where | Reliability (sample) | Use for abilities | Extra cost |
|---|---|---|---|---|
| title, extract (≤900 chars), image, legacy category, pageId, revisionId | **on-chain** (`Article`) | 100%. Extract ≥40 chars by filter. | Flavour text is the first sentence of the extract. Years and "(born …)" can be parsed from the extract. | 0 |
| pageBytes, languageCount, 6 months of views, dailyAverage, trendPercent, attention/depth/reach | **on-chain** (`MintStatsSnapshot`) | 100% (attention null under 1%) | Cost and body, Multilingual/Local, Volatile, Rising/Fading | 0 |
| description (short) | live: WP `prop=description` | 90% | Keywords such as "Species of", "Village in", "film by" | in the batched meta query |
| non-hidden categories | live: WP `prop=categories` | 100%, median 5 | Year in 55% of cards; births/deaths 25%; establishments 10%; stub 34%; country words 52% | same query |
| wikidataId | live: `pageprops` | 100% | Key for all Wikidata data | same query |
| P31 (instance of) | live: Wikidata | 94% | **Type** | 1 per 50 cards |
| Dates: P569/P570 birth and death, P571 inception, P585/P580 point in time and start | live: Wikidata | *(memory)* Common for humans, organisations and events; rare for taxa and concepts | Era keywords ("Ancient", "19th-c."), "Contemporary" links | same `wbgetentities` |
| Coordinates: P625, or WP `prop=coordinates` | live | *(memory)* Nearly all Locations, few others | "Adjacent" (distance), hemisphere, lat/long flavour | in the batched meta query |
| Country: P17 / P27 | live | *(memory)* Most places, people and organisations | "Compatriot" connection | same `wbgetentities` |
| Outlinks: `prop=links` | live | 100% | Connected (§4) | 1–6 per card, continuation |
**Request cost per card, first sight, cached forever:**
- **Batched, amortised:** about 1.5 requests per card.
- 1/50 for the WP meta query (description, categories, pageprops, coordinates)
- 1/50 for `wbgetentities&props=claims`
- under 0.1 for class/P279 lookups
- about 1.2 for outlinks
- **Unbatched (a single card at a time):** about 4 requests.
- **Payload warning *(memory)*:** `props=claims` for famous items can be **hundreds of KB to over 1 MB** (Q76, Barack Obama). It cannot filter by property.
- **Option 1:** a WDQS SPARQL `VALUES` query for the ~12 needed properties, covering up to about 100 items per query.
- **Option 2:** accept the size and only ever fetch each item once.
**Request cost per match**, two decks of 10 cards:
- **Warm caches:** **0–1**. That is the links matrix (§4) if per-card outlinks aren't cached.
- **Trending mode:** add up to **20 per-article pageview requests** for today's figures. These can't be batched, so refresh them in a daily background job, never at match start.
- **Cold worst case:** about 23 requests, which is 23 s at 1 request per second. So prefetch when a card is minted or received, or when a deck is saved, never at queue time.
**Frozen vs live:**
- **On-chain (frozen by consensus):** only `article` and `statsSnapshot`.
- **Everything else is mutable.** That covers type, dates, country, categories and links.
- **Card profile:** freeze it once into an off-chain record, `{tokenId, pageId, mintRevisionId, fetchedRevisionId, wikidataRev, type, faction, dates, country, coords, outlinks hash}`. Cache it forever, and log it in the match record so replays and recaps are deterministic.
- **Why not the mint revision:** fetching categories or links *as of* the mint revision needs `action=parse&oldid=`, which is costly because of the parser-cache miss *(memory)*. Freeze at first sight instead, and record which revision was used.
- **Agreement between hosts:** different hosts could freeze different profiles. To fix that, a node publishes its signed profile file per tokenId, and hosts adopt the first one seen.
## 4. Connected: cheap relationship detection (`analysis/connected.py`)
**Measured on random pairs** (27,730 pairs):
| Relation | Pairs | In a 10-card deck (≥1 pair) |
|---|---|---|
| any shared category (including generic ones such as "Living people" or "1948 births") | 1.96% | |
| shared **specific** category | **0.04%** | 2% |
| same P31 class | 7.9% (mostly human/taxon) | |
| same P31, excluding human and taxon | 0.22% | |
| same country (text proxy for P17/P27) | 7.4% | 91% |
| same century (text proxy for dates) | 16.6% | |
| country OR century OR specific category | 21.5% | 99% |
**Direct article links between two random cards will almost never fire.**
- Shared specific categories are already 0.04%, and links are rarer still between random stubs.
- Obama ↔ White House only works because both are famous starter-style titles.
- **Proposed tiers:**
- **Linked (rare, strong):** a direct link in either direction.
- **Compatriot:** same P17 or P27.
- **Contemporary:** same century, from P569, P571 or P585.
- **Kin:** same P31, excluding human and taxon.
**Methods compared:**
| Method | Cost | Notes |
|---|---|---|
| Pairwise links matrix: `action=query&titles=A1…An&prop=links&pltitles=A1…An&pllimit=max` | **1 request per match** for up to 50 titles; at most 400 edges for 20 cards, under the 500 limit | Gives both directions for every pair. Uses current revisions. Misses links that go through redirects, unless you add `redirects` and also check the redirect titles. |
| Per-card outlinks (`prop=links`, ns0), cached | 1–6 requests per card, once | Intersection is local and free, so it scales to the whole chain. Network-wide this is at most 96 × about 2 ≈ 200 requests a day. **Recommended.** |
| `prop=linkshere` (backlinks) | Famous pages: thousands to millions, many continuations | **Avoid.** Outlinks from both sides already cover A↔B. |
| Shared Wikidata values (P17, P27, P131, P106, P101, P361, dates) | Included in the card profile fetch | Most useful in practice. Gives readable reasons ("both Polish"). |
| Categories | Included in the meta fetch | Too sparse except for generic categories, which are meaningless. Use only for era and country when Wikidata lacks them. |
**Caching:**
- **On first use:** compute each card's outlinks and profile once, keyed by `(pageId, fetchedRevisionId)`, and pin them to the `tokenId` forever.
- **During a match:** the host computes the edge list at match start from cached profiles, which needs no requests, and writes it into the match log.
- **Re-fetching:** never re-fetch for an owned card. An optional "re-sync" can be a deliberate, logged action.
## 5. Wikimedia politeness *(memory: not checked live tonight)*
**Rules:**
- **User-Agent policy:**
- The policy asks for a descriptive UA with contact details: `Tool/version (https://site; contact@…) lib/ver`.
- Generic or library-default UAs may be throttled or blocked. Browsers must use `Api-User-Agent`.
- **Finding:** our defaults have **no contact at all**. That covers `Wikideck/0.1 (experimental Wikipedia article cards)`, the enrichment UA, and the sampler UA `WikiDeckCardDataSampler/0.1`. Contact is opt-in only, through `WIKIDECK_USER_AGENT`. This is likely part of why the shared IP got 429.
- **Action API etiquette:**
- Make requests **serially**, not in parallel. There is no fixed read limit, but use "reasonable" volume.
- Send `maxlag=5` on non-interactive requests, and on a `maxlag` error wait for `Retry-After`.
- Batch: `titles`/`pageids` take 50 per request (500 with the bot flag), and `pltitles` 50 likewise.
- Use continuation, `Accept-Encoding: gzip`, and `formatversion=2`.
- **REST API (`/api/rest_v1`):** the historical guidance was ≤200 requests per second per client. Since 2025 Wikimedia has tightened limits on unauthenticated automated traffic and enforces them per IP/UA. Treat any published number as a ceiling, not a target.
- **API Portal or token traffic:** I recall anonymous traffic at about 500 requests an hour and a personal token at about 5,000 an hour. **Unverified.**
- **Pageviews (AQS):**
- The documented limit was about 100 requests per second per client.
- It is **per article and has no batching**.
- Daily data lands once a day, a few hours after 00:00 UTC, so refreshing more than once a day is waste.
- **Wikidata:**
- `wbgetentities` takes up to 50 ids per request; use `props=` to slim the response.
- WDQS allows about 60 s of query time per minute and 5 concurrent queries per client, and blocks after repeated errors.
- **Always:** honour `429`/`503` with `Retry-After`, use exponential backoff, and cache.
**Concrete node defaults:**
| Setting | Default |
|---|---|
| Client | one shared `WikimediaClient` per **process, and budget per egress IP** (several nodes or agents behind one IP must share it) |
| Rate | token bucket: **1 request/s sustained, burst 3**, **1 in-flight per host** (en.wikipedia.org, wikimedia.org, wikidata.org) |
| Daily cap | 20,000 requests a day per node (configurable). Past it, serve from cache or 503. |
| maxlag | `maxlag=5` on every Action API call from background jobs |
| 429/503 | honour `Retry-After`, else 5 s × 2ⁿ up to 15 min. **Circuit breaker:** after 3 consecutive 429s, stop all Wikimedia calls for 10 min. |
| Batching | 50 pageids/titles, 50 Wikidata ids, links matrix per match |
| UA | `WikiDeck-node/<ver> (+https://<node host>; <operator contact>)`. **Start-up warns loudly**, or refuses in prod, unless a contact is set. |
| Cache | on disk. Profiles and outlinks are immutable (keyed by tokenId and revision). Pageviews keyed by (title, day). ETag / `If-None-Match` on REST. |
| Background | one daily job after about 06:00 UTC refreshes live views for cards in active decks only |
**Mining load: a real issue the plan doesn't list.**
- **What a template costs:** `POST /api/mining/template` calls `wikipediaArticles.random()`, plus `getFresh()` enrichment, on **every template**. That is 3 or more Wikimedia requests (`server/testnet/http.ts:122`).
- **How often:** each miner asks for a new template whenever the tip changes (`server/testnet/miner.ts:89`).
- **Load:** M miners make **about 3M requests per block, all in a burst at each new tip**, or 288·M a day. With 100 miners that is 28,800 a day, which breaks the cap above.
- **Fix:** prefetch a small pool of article+snapshot pairs in the background at 1 request/s, and serve each template from the pool.
- `grnlimit=10` yields about 4.4 usable articles per request.
- This changes no consensus rules.
- Wikimedia load then stays at about 2–3 requests per block, whatever the number of miners.
## 6. Supply
- **Production:** at most 96 blocks a day means at most 96 cards a day: 672 a week, about 2,900 a month, 35k a year.
- **Mainnet reality:** 12 cards, all mined on 2026-11-01. That is **0.75 a day averaged over 16 days**; mining has effectively stopped.
- **Model:** with M equal-hashrate miners, each expects 96/M cards a day, so the time to N cards = **N·M/96 days**. This assumes continuous full block production.
| Active miners M | N=7 | N=10 | N=20 | N=40 | N=60 |
|---|---|---|---|---|---|
| 1 | 1.8 h | 2.5 h | 5 h | 10 h | 15 h |
| 10 | 0.7 d | 1.0 d | 2.1 d | 4.2 d | 6.3 d |
| 50 | 3.6 d | 5.2 d | 10 d | 21 d | 31 d |
| 100 | 7.3 d | 10 d | 21 d | 42 d | 63 d |
| 500 | 36 d | 52 d | 104 d | 208 d | 313 d |
**Caveats:**
- **Mining is a Poisson lottery.** At M=100, 10 cards take a mean of 10.4 days, but about 15 days at the 90th percentile.
- **Hashrate is skewed.** The *median* miner gets well under 96/M; a phone-class CPU gets about 0. Non-miners only get cards through trades.
**Implications for deck size:**
- **Rule of thumb:** a deck size D is reachable in T days only if D ≤ 96·T/M.
- For a 1-week ramp with 100 miners, D ≤ 6.7.
- **40–60-card constructed decks** need M ≤ about 12, or months of play.
- **Recommendation: a 7–10-card constructed deck**, plan option (a).
- Every card is unique anyway, so this is effectively Highlander, with recursion and reuse.
- **Collection-as-deck**, option (d), covers players with fewer than 7 cards.
- **Draft, option (c), is supply-friendly**, but only once the chain holds enough cards.
- An 8-player draft of 3×15 needs 360 cards (about 4 days of full production); a 2-player draft needs about 60.
- **Today's 12 cards cannot feed even one draft.**
- **The cold start is the blocker, not the steady state.** Eric must decide one of two things for launch week:
- accept a loaner or shared pool (plan option (b), which conflicts with requirement 1), or
- accept that PvP only works after someone mines continuously for a day or two.
+278
View File
@@ -0,0 +1,278 @@
# Night 2 review: game designer
**Verdict: approve the direction with changes.** Do the MTG-style duel, the article-derived types, the Connected bonus and the battle recap. The plan has one gap: it has no answer to supply, and Open problem 1 decides whether any of the rest gets played. My answer is a **15-card deck made of your owned article cards plus Basics**. Basics are non-article "process" cards, the equivalent of MTG basic lands. **Draft** from the chain's whole register is the front door for players who own few cards. Rules core: **no stack, no land cards, fixed income, attack/block with single blocks, 16 Notability.** Everything below is data in the existing rules-as-data style.
---
## 1. Supply (Open problem 1)
**The facts:** mainnet holds 12 cards, all in one wallet. At most about 96 a day can be mined, and the real rate is far lower. For months nearly every player will own 0–3 cards. No version of "real cards only" fills a 40-card deck. The question is how the rules bend.
### Recommendation: "Collection + Basics" (options a and d, with the MTG basic-land argument)
- **Deck:** exactly **15 cards**.
- **Article slots:** any number of your owned mints, up to 15. Each mint can appear once, since it is unique. Different mints of the same `pageId` can appear up to 2 times.
- **Basics** fill the rest, at most 3 copies of each.
- Owned and loaner copies no longer mix (v1 §2 is retired).
- **Basics are not articles.** They are a fixed set of 8 Wikipedia process cards that nobody owns and anyone can use in any number. Listed in §2.4.
- **The argument for Eric:** in Magic, about 40% of every deck is basic lands that nobody collects, and the deck is still "your deck". Here, **every article in your deck is one you mined**. The letter of the requirement holds for articles. It bends only in that the deck isn't made *only* of articles.
- **Signature card:** before the match you name one owned article as your Signature. It **starts in your opening hand**, in addition to the 5 cards you draw.
- So a player with 1 card plays that card in every game.
- The recap's infobox lists it under "Commanders".
- **Small deck, high visibility:** a typical game lasts about 8 turns each, so you see 5 + 8 = 13 of your 15 cards. Your owned cards nearly always show up.
- **Running out of cards doesn't lose the game.** When your library is empty, shuffle your Deleted pile (the graveyard) into it. This is option (a)'s "recursion and reuse".
- **Power parity:** an article and a Basic of the same cost are worth about the same, with articles roughly 10% ahead from keywords and Connected (sim target: an all-Basics deck wins 42–48% against a 12-article deck).
- Owning cards buys **variety, story and a modest edge**. It doesn't decide games. Ranked Glicko absorbs the rest.
### What a new player does
| Owns | Practice / AI | Casual PvP | Ranked | Draft |
|---|---|---|---|---|
| 0 cards | Tutorial: the Basics deck plus 3 Archive cards, the existing loaners in a grey "Not owned" frame | Basics-only deck allowed | no (one article needed) | **yes: main path** |
| 1–5 | any | Collection + Basics; the Signature guarantees their card | yes | yes, plus the owner perk |
| 6–15 | any | mostly articles | yes | yes |
Casual matchmaking groups players by article count (0–2 / 3–7 / 8+). Ranked has one ladder.
### Fallbacks, in order
- **Fallback A (if all-Basics games test as dull):** the "Archive ramp".
- Up to `max(0, 8 − owned)` Archive loaner articles may fill slots, and they shrink as you own more.
- Archive cards are grey-framed and listed in the recap as "(Archive, not owned)".
- This is option (b) with a sunset clause. **Eric must accept it explicitly.**
- **Fallback B (strictest "real mints" reading):** "Register as shared library".
- Both players draw from one shared library: every mint on the chain, whoever owns it, plus Basics.
- Your owned cards sit in a Command zone. You can cast them at any time, and they return there with +1 cost tax.
- Every article in play is then a real mint, but your *deck* is no longer yours. It also needs ≥ 30 mints before games stop repeating.
### What Eric must decide
1. Are Basics acceptable as a deck's "lands"? I recommend yes.
2. Archive loaners below 8 owned cards: allowed, or Basics only? I recommend Basics only at launch, with Fallback A behind a ruleset flag.
3. Can owning cards give a small edge in ranked? I recommend yes, capped by the parity target above.
---
## 2. Rules core (Open problem 2)
**Is MTG combat too complex for one night?** Not if the rules are cut to the bone: no stack, single blocks and one main phase. It is the only part that gives "feels like Magic", so keep it. The lane engine, `Topic` and `Portal` are retired. `rng.ts`, `ruleset.ts`, `modes.ts`, `validateDeck`, `replay`, `viewFor` and serialisation carry over.
### 2.1 Setup and turn
- **Notability:** each player starts at **16** Notability (the life total). Reaching 0 loses.
- The log reads: "*X* was nominated for deletion. The result was **delete**."
- **Opening hand:** draw 5, plus your Signature. One free mulligan: shuffle, draw 5, put 1 on the bottom.
- **First player:** skips the first draw. The second player gets **Second Opinion**, a 0-cost token: "Gain 1 Citation this turn".
- **Turn:** (1) **Refresh:** Citation maximum +1 (to at most 7, plus Territory), Citations refill, permanents untap. (2) **Draw** 1. (3) **Main:** play cards in any order. (4) **Combat.** (5) **End:** damage clears, "until end of turn" effects end, hand limit 7. No second main phase.
- **Resources:** fixed income, the Hearthstone model. **No land cards, so no land screw.**
- Locations add ramp (+1 maximum each, at most +2 from Territory in total). Max Citations is 9.
- **Summoning sickness:** a Figure can't attack the turn it enters ("Pending review"). Trending removes this.
### 2.2 Combat
- **Attack:** the active player taps any untapped Figures to attack. They attack the opponent only, not permanents.
- **Block:** the defender assigns at most one blocker to each attacker. Each blocker blocks one attacker. Locations with Garrison may block.
- **Damage:** simultaneous. Blocked pairs trade damage, and an unblocked attacker deals its power to Notability.
- No trample and no first strike in v1. That is about 150 lines of engine, and it is easy for bots.
### 2.3 Interaction without a stack
- **There is no priority, and nothing happens at instant speed.** That is the right call for online PvP on phones, because priority passes cost latency and clock time. Events are sorcery-speed.
- **v2, cut tonight:** *Pending changes*, Events set face-down that trigger on a condition (like Hearthstone's Secrets). They add interaction on the opponent's turn without priority passes.
### 2.4 Basics (the 8 process cards, max 3 copies each)
| Basic | Type | Cost | Text |
|---|---|---|---|
| Anonymous Editor | Figure | 1 | 1/2 |
| Established Editor | Figure | 3 | 3/3 |
| Administrator | Figure | 5 | 4/5. When this enters, destroy target Figure with power 2 or less. |
| Citation | Event | 1 | Put a +1/+1 counter on target Figure. |
| Revert | Event | 2 | Return target Figure or Doctrine with cost 3 or less to its owner's hand. |
| Talk Page | Event | 2 | Draw 2 cards, then discard 1. |
| Speedy Deletion | Event | 4 | Destroy target Figure. |
| Sandbox | Location | 1 | Territory. Garrison 0/1. |
### 2.5 Length
- **Clock:** 60 s per turn plus a 5-minute bank per player.
- **Turn cap:** at the end of turn 12 (each player's), the player with more Notability wins. A tie goes to the higher total power on the battlefield, then to the second player.
- **Sim target:** median 7–9 turns each, 6–12 minutes, decided before the turn cap in ≥ 85% of games.
- **Tunable:** Notability (14–20) and the maximum Citations are ruleset keys.
---
## 3. Derivation: article to card (Open problems 3 and 5)
All of it is deterministic: no LLM, fixed tables, and a versioned `deriveVersion`.
**Inputs:**
- the snapshot (as v1)
- `description`, `categories` and `pageBytes`, already in `PageEnrichment`
- the Wikidata entity's **claims**, fetched in the same `wbgetentities` call (≤ 50 ids per request) that Night 1 already used for P31, so no extra requests
The host caches the claims per `(wikidataId, lastrevid)` and freezes them into the `CardInput` at deck validation.
### 3.1 Type: the first rule that matches wins
1. **Exact Q-id table:**
- Q5 human, Q95074 fictional character, Q16521 taxon, Q23038290 fossil taxon, Q8436 family, deities → **Figure**
- Q486972 human settlement → **Location**
2. **P31 label regex,** reusing `factionRules`, grouped by type:
- People and Nature taxa → Figure
- Places and Structures → Location
- Events and Sport matches/seasons/finals → Event
- Works, Organisations, Sport teams and companies → Institution
- Ideas & Tech → Doctrine
3. **P279 "subclass of" labels,** with the same regexes. Concepts such as *Monoid* have no P31 and only P279. This is why 13 sample cards have no P31.
4. **Description regex,** present on 90% of the sample:
- `^Species of|genus of` → Figure
- `(actor|politician|player|singer|writer)\b` → Figure
- `^(Commune|Village|Town|Municipality|River) in` → Location
- `election|storm|battle` → Event
5. **Categories:** `\d+ births|deaths|Living people` → Figure, `Communes of|Villages in|Populated places` → Location, `establishments` → Institution, `elections|disasters|battles` → Event.
6. **Otherwise Doctrine (Neutral).** An abstract thing is the safest default.
**Bug to fix:** the People regex `\bhuman\b` matches "human settlement". As a result *Buky*, *Lunugala*, *Žižki* and *Laukhella* are People in the sample today (4 of 68). Run the exact Q-id step first. Then score the sample's 236 cards in CI and hand-check the 34 "Other" cards. Target ≥ 95% plausible.
### 3.2 Portal (the faction or colour), orthogonal to type
Five **Portals**, each with an MTG colour analogue, plus **Neutral**:
- **Arts** (red)
- **Earth** (green)
- **History** (black)
- **Science** (blue)
- **Society** (white)
**How a card gets one:**
- **Score:** count keyword hits in description + categories + P31/P106 labels, using fixed regex lists in data. For example:
- History: `\d+(st|nd|rd|th)-century|deaths|sultan|empire|castle|medieval`
- Science: `mathemat|algebra|programming|gene|chemi|disease|scientist`
- **Ties** go to the type default: Figure → Society, Location → Earth, Event → History, Institution → Arts, Doctrine → Science.
- **Zero hits:** the card is Neutral.
**Deck rule: none.** You can't ask for colour discipline from a 3-card collection. Portals matter only through effects, for example "Figures from your Portal get +1/+0".
### 3.3 Cost and stats
- **Percentiles, `fame` and cost 1–6:** exactly as v1 §1.1–1.2, using the existing `costThresholds`.
- **Figure:**
- budget `B = 2·cost + 1` (3/5/7/9/11/13)
- `s = clamp(pct_att/(pct_att+pct_depth), 0.3, 0.7)`
- `power = max(1, round(B·s))`, `toughness = B − power`
- each keyword costs 1 point from the larger stat (minimum 1)
- at most 2 keywords
- fame sets cost, never efficiency: v1's anti-inflation principle is kept
- **Location:**
- cost `ceil(cost/2)` (1–3)
- Territory
- Garrison 0/T, where `T = 1 + floor(pct_depth/25)`
- can't attack
- **Event, Institution and Doctrine:** magnitude `k = ceil(cost/2)` (1–3) inside the templates in §3.5.
### 3.4 Keywords, in priority order, max 2
Rates are measured on the 236-card sample.
| Keyword | Trigger | Rules text | Sample rate |
|---|---|---|---|
| **Protected** | category `Living people` | Can't be targeted by Events an opponent controls. | 14% |
| **Featured** (flying) | pct_attention ≥ 85 | Can be blocked only by Featured or Translated Figures. | ~15% |
| **Translated** (reach) | languageCount ≥ 20 | Can block Featured Figures. | 12% |
| **Legacy** | P570 present or a `\d+ deaths` category | When this is destroyed, draw a card. | 12% |
| **Stub** | category contains "stubs" or pageBytes < 3000 | At the start of your turn, put a +1/+1 counter on this (at most 2). | 34% |
| **Well-sourced** | pct_depth ≥ 85 | Prevent the first 1 damage dealt to this each turn. | ~15% |
| **Trending** (haste) | Trending mode only: live ▲ | Can attack the turn it enters. | mode |
### 3.5 Templates by type
Text is set in MTG's terse register. Subtype regexes run over the P31/P279 labels and the description.
- **Event:**
- conflict (`battle|war|siege|annexation|protest|riot`): "Deal k+1 damage to target Figure or Location."
- disaster (`storm|nor.easter|cyclone|earthquake|flood|tornado`): "Deal k−1 (min 1) damage to each Figure and each Location."
- election (`election|referendum|straw poll`): "Create k 1/1 Society Figure tokens named Elected Member."
- sport (`match|final|season|tournament`): "Target Figure you control gets +k/+0 until end of turn, then fights target Figure you don't control."
- period (`events in a specific year|ceremony|festival`): "Draw k cards."
- default: "Target Figure gets +k/+k until end of turn."
- **Institution** (a permanent, like an artifact):
- organisation/company/party/military: "At the start of your turn, if you control fewer than 3 tokens, create a 1/1 Figure token named Staff Member." (for cost ≥ 4; below that, "When this enters, …")
- creative work: "Figures you control that are Connected to this get +1/+1. When this enters, draw a card."
- team/club: "Figures you control with power 2 or less get +1/+0."
- vehicles and ships (v2: Crew): "Figures you control have Featured while attacking."
- **Doctrine** (an Aura, the mechanic in Eric's Tetrated dodecahedron → Cho Yang-ho anecdote):
- **Adopt:** "Attach to target Figure. Adopted Figure gets +k/+k." If the Figure leaves, the Doctrine goes to Deleted.
- Riders by subtype:
- mathematics: "and has Well-sourced"
- language/software: "Whenever it deals combat damage to a player, draw a card"
- ideology/religion: "Other Figures you control from the same Portal get +1/+0"
- **Pathology** (`disease|clinical sign|gene|chemical`) is inverted: "Attach to target Figure. It gets −k/−k."
- **Flavour text:** the first sentence of the extract, or `description` if the extract is missing (the sample has none).
### 3.6 Example cards from `card-sample.json`
Tiers and stats are taken from the file. The final numbers come from the real percentile function.
| Card | Derivation | Generated card |
|---|---|---|
| **Tom Ewell** (person) | Q5 → Figure. "actor" → **Arts**. Tiers 10/9/10 → fame ≈ 92 → cost 6. B = 13, s ≈ .53 → 7/6. Featured (att), Translated (30 langs), Legacy (1994 deaths) apply; capped at 2 by priority: Featured, Legacy. | **Tom Ewell**, 6, Figure — Arts. **6/5**. *Featured. Legacy.* "American actor (1909–1994)." *(description; extract not in sample)* |
| **Salahuddin of Aceh** (person) | Q5 → Figure. "16th-century", "Sultans" → **History**. 7/6/7 → fame ≈ 62 → cost 4. B = 9 → 5/4. Legacy (1548 deaths) → 4/4. | **Salahuddin of Aceh**, 4, Figure — History. **4/4**. *Legacy.* "Sultan of Aceh (1530–1537/9)." |
| **Monoid** (maths/concept) | No P31 → P279 "algebraic structure" → Doctrine. "Algebraic structures", "Category theory" → **Science**. 10/10/10 → cost 6, k = 3. Maths rider. | **Monoid**, 6, Doctrine — Science. *Adopt.* Adopted Figure gets +3/+3 and has Well-sourced. "Algebraic structure with an associative operation and an identity element." |
| **Agda (programming language)** | P31 programming language → Doctrine. **Science**. 10/9/9 → cost 6. Language rider. | **Agda**, 6, Doctrine — Science. Adopted Figure gets +3/+3. Whenever it deals combat damage to a player, draw a card. |
| **Dichromorpha viridis** | Q16521 taxon → Figure. "Species of grasshopper" → **Earth**. 7/1/6 → fame ≈ 45 → cost 3. B = 7, s clamped to .7 → 5/2. Stub (1,347 B) → 4/2. | **Dichromorpha viridis**, 3, Figure — Earth. **4/2**. *Stub.* "Species of grasshopper." |
| **April 2021 nor'easter** | P31 nor'easter → Event, disaster. "2021 in Maine" … → **Earth**. 6/10/2 → cost 4, k = 2. | **April 2021 nor'easter**, 4, Event — Earth. Deal 1 damage to each Figure and each Location. *(k−1 = 1: the sim decides whether disasters need k instead)* |
| **1954 Gold Coast general election** | legislative election → Event, election. **Society**. 7/4/5 → cost 4, k = 2. | 4, Event — Society. Create two 1/1 Society Figure tokens named Elected Member. |
| **Pathrail Mosque** | mosque → Location. "15th-century mosques" → **History** (2 hits) vs Society (1). Derived cost 4 → Location cost 2. pct_depth ≈ 55 → Garrison 0/3. | **Pathrail Mosque**, 2, Location — History. Territory. Garrison 0/3. |
| **Tatsunoko Production** | animation studio → Institution, organisation. **Arts**. 10/10/9 → cost 6. | **Tatsunoko Production**, 6, Institution — Arts. At the start of your turn, if you control fewer than 3 tokens, create a 1/1 Figure token named Staff Member. *(the sim may want 2/2 at cost 6)* |
| **Charmont-en-Beauce** | commune of France → Location. **Earth**. 2/2/10 → cost 1. Stub ignored (Locations take no keywords). | **Charmont-en-Beauce**, 1, Location — Earth. Territory. Garrison 0/2. |
**Readability:** a Figure shows cost, P/T, Portal, the type line and up to 2 keyword words. The full text is behind a tap. This is still one phone-sized face.
---
## 4. Story mechanics (Open problem 4, design side)
### 4.1 Connected
- **Data:** the frozen claims (§3). Each card gets `qid` and `connectKeys` at derivation. The engine intersects keys without any I/O.
- Keys come from P17/P27/P131/P495 (place), P106/P101/P921/P136 (field), P361 (part of), P171 (parent taxon) and `C<century>` (from P569/P570/P571/P580/P585).
- Century counts only between cards of **different** types, because it is too common otherwise.
- Wikipedia link tables are not used in v1, which gives zero extra requests.
- **Direct connection:** one card's claims cite the other card's `qid`, for example Barack Obama with the White House (residence).
- **Bonuses:**
- A Figure gets **+1/+1** while you control another card Connected to it; **+2/+2** if the connection is Direct.
- A Doctrine adopted by a Figure Connected to it gives its bonus twice.
- A Connected creative-work Institution buffs, as in §3.5.
- **The card shows why:** "Connected — Pathrail Mosque (country: Bangladesh)".
- **Rivalry (log only, v1):** an opposing pair that is Connected is noted in the log and the recap ("The two had previously been associated through …").
- **Tuning:** sim on random 15-card decks from the sample. Target 0.5–2 Connected pairs per deck. Drop the noisiest keys (P17 = United States, or France) if the rate is above that.
### 4.2 Match log
Past tense, third person, no adjectives about real people, numbered references. Card names are *italic* links, and the first mention is "the card *X*". Examples:
> Turn 3. *Tom Ewell* (American actor, 1909–1994) was deployed by eric.[1]
> Turn 4. *Monoid* was adopted by *Salahuddin of Aceh*, which became a 7/7.[2][3]
> Turn 5. The *April 2021 nor'easter* dealt 1 damage to each Figure. *Dichromorpha viridis* was destroyed.[4][5]
There is one template per engine event type (about 25 strings). It is generated from the deterministic event list, so `replay` regenerates it exactly.
### 4.3 Battle recap ("Battle of X (2026)")
- **Title:**
- "Battle of {Location}": the winner's highest-cost Location at the end, else any Location in play.
- With no Location: "{Signature of the winner} campaign".
- The prefix is "Skirmish at" under 6 turns and "Siege of" over 10.
- **Infobox:** date, Location, result ("Decisive victory" if the margin is ≥ 10, "Pyrrhic victory" if the winner has ≤ 3 Notability, "Victory by decision" at the turn cap), belligerents (players), commanders (Signatures), strength (cards deployed), casualties (cards destroyed).
- **Sections:** Background (Portals, Signatures); Course of the battle (the 3–5 largest swings); Aftermath (final Notability, rating change); See also (Connected pairs); References (every card's real article with `revisionId`, plus the mint id).
- **Branding:** Wikipedia-*style* typography, but **no Wikipedia or Wikimedia logo or name in the masthead.** Footer: "This page records a WikiDeck match. Card names refer to Wikipedia articles; the events described are game events."
---
## 5. Modes in the new structure
- **Classic:** Collection + Basics, 15 cards, Signature. Ranked (Glicko-2) and casual queue, plus invite-by-link.
- **Draft (front door):**
- **Pool:** the host's **Register**, every mint on its chain whoever owns it. That means real mined cards for players with none.
- **Picks:** an open pool of 20 cards, made of min(Register, 16) random mints plus Basics. The two players snake-pick 10 each, at 20 s per pick, and the deck is their 10 picks + 5 Basics of their choice.
- **Owner perk:** if you own a mint in the pool, you may take it as your first pick, once, and the recap credits it "from the collection of X".
- **Small Register:** mainnet's 12 mints still work, padded with Basics. If the Register has fewer than 16 mints, the ruleset may allow Archive padding (Fallback A).
- **Trending:** any deck policy. Live trend frozen at match start, through the throttled client and its daily cache:
- **▲** (trend ≥ +25% and daily ≥ 20): Trending (haste) and +1/+0
- **▼** (≤ −50%, daily ≥ 20): −1/−0
- the 20/day floor keeps noise out (CARD_DATA finding 6)
- **House rules:** unchanged, data-only. New keys: `startingNotability`, `deckSize`, `maxCitations`, `turnCap`, `basicsAllowed`, `archiveRamp`, `connectedBonus`.
## 6. Scope for tonight (designer's cut order)
**Must:** the duel engine (§2); derivation (§3.1–3.5) with the classifier test on the sample; the 8 Basics; Connected with frozen claims; the match log; the recap page; Classic + Draft through PvP lobbies; a greedy combat bot.
**Cut first:** Rivalry text, Trending's ▼, Pathology inversion, Institution subtypes beyond the default. **Defer:** Pending changes, vehicles/Crew, the Archive ramp (flag only).
## 7. Risks
1. **The engine rewrite is the night's critical path.** The lane engine, bots and golden tests don't transfer. Mitigation: freeze the §2 rules at 21:45, then one engine owner and one bot/sim owner from the first hour.
2. **Famous cards all cost 6** (Tom Ewell, Monoid, Agda, Tatsunoko). Collections of notable articles are top-heavy. Mitigation: the Signature in hand plus fixed income make 6-drops castable by turn 6. Watch the sim's curve.
3. **Basics games are generic.** For 0–2 card owners, Classic has little story. That is exactly why Draft is the front door, and why Fallback A exists.
4. **The classifier is wrong ~5–10% of the time** (the human-settlement bug, P279-only concepts). Show the type line with a "derived from: P31 = …" tooltip so mistakes read as data, not bugs. Correct it by `deriveVersion` bumps, never by hand.
5. **Claims are live data, not the mint snapshot.** Card text can drift between nodes. Freeze the claims per card at first sight with `claimsRevision`, and log it. The host's copy is authoritative for its matches.
6. **Real people in shareable recaps:** this is a living-persons and tone risk. Use mechanical verbs only ("was destroyed", "attacked"), use "the card X", and add no commentary.
7. **Collector advantage in ranked:** enforce the parity target in the sim before ranked opens.
8. **Connected density** may be near zero in random collections (good for surprise) or everywhere through country keys (noise). Tune the keys from the sim. Don't guess.
+246
View File
@@ -0,0 +1,246 @@
# Night 2 plan review: lead engineer (feasibility and scope)
Reviewed against `wd` @ 4f7a8a7 (upstream main with Night 1). Paths are repo-relative. `docs/night/REPORT.md` does **not exist** in this checkout, so Night 1 lessons come from HANDBOOK/BOUNDARIES and the coordinator's notes.
**Verdict:** the plan is feasible in 4.5 h **if** the duel engine is a *new sibling package* with no stack and sorcery-speed only, PvP stays on the self host, and the Wikimedia client is handled as transport-only for the mint path. Three facts change the plan:
1. **Owned cards cannot reach PvP today.** The host never accepts a player deck (see §2).
2. **Card data is not trustworthy.** Article and stats are claims made by the miner: consensus only checks internal consistency.
3. **The browser drives most Wikimedia traffic.** Every card face triggers a server enrichment fetch.
---
## 1. Engine: reuse, rewrite, smallest change
**Reusable as is:**
- `shared/game/rng.ts` (mulberry32, `Rng.fromSeed`, exported via `index.ts:3`)
- the clone-on-apply pattern
- the `(ruleset, decks, seed, actions)` replay contract: `replay` at `engine.ts:433`, `applyAction` at `engine.ts:291`
- zod `parseRuleset` (`ruleset.ts:83`)
- `percentile`/`fame`/`costFor` in `derive.ts:52-79`
- the generated sample percentiles (`card-data.ts`, `scripts/game-card-data.mjs`)
The host is **engine-agnostic** through `EnginePort` (`server/game/engine-port.ts:9-43`), so a second engine plugs in with a new binding file instead of host surgery. Bots are also engine-agnostic (`shared/bots/types.ts` `EngineAdapter`, with `toAct`, `assume` and `evaluate`), and a sequential duel fits that adapter *better* than Edit War does: `game-adapter.ts:243` fakes `toAct` from `committed`.
**Must be rewritten:**
- **Everything lane- or Topic-shaped:**
- `MatchState.lanes` / `LaneState` (`types.ts:136-170`)
- the simultaneous queue/commit/reveal turn (`engine.ts:291-363`)
- Topic scoring (`finish`, `engine.ts:366`)
- Portals and keywords (`keywords.ts`, 241 lines, hooks assume lanes and `onReveal`)
- **`GameCard`** (`types.ts:24-43`): a duel card needs `type`, `toughness`, rules-text ids and flavour, and `validateDeck` (`engine.ts:27`) hard-codes cost 0–6, power 0–30, Faction and lane fields.
- **UI boards** (`src/game/MatchBoard.tsx`, `adapter.ts`).
- **`parseModeRuleset`** host checks (`engine-binding.ts:31-43`), which are Edit War-specific: `turns × turnSeconds ≤ 240`, `topicCount`.
**Recommendation: build `shared/duel/` as a sibling package and leave `shared/game/` untouched.**
- Edit War stays playable: its 6 test files keep passing and its modes keep working.
- Two engine agents don't conflict with anyone.
- Duel exposes the same API names (`createMatch`, `legalActions`, `applyAction`, `viewFor`, `replay`, `validateDeck`, `isOver`, `result`), so `EnginePort` and `EngineAdapter` bindings are thin files.
**Smallest core that still feels like Magic (one night):**
- **Turn:** start (untap, income, draw) → main → combat → end. No second main phase, no stack, no instants and no priority passing. Events are sorcery-speed only.
- **Triggers** resolve immediately in a fixed order: active player first, then reveal `seq`, as Edit War's `onReveal` already does.
- **Resources:** fixed income ramp, `incomeByTurn` as data (Edit War's `energyByTurn` idea, `types.ts:70`), with no carry. A Location is a permanent that gives +1 income. No land cards means no land screw.
- **Figures** have power, toughness and summoning sickness; damage clears at end of turn. **Institutions and Doctrines** are permanents with `ongoing` statics (port the pure `ongoing` hook contract, `keywords.ts:16`). Life is 20 by default (data).
- **Actions must be atomic** so that `legalActions` stays small. That matters because the host checks each action by deep-equal against the full list (`host.ts:187-189`), and the bots enumerate it:
- `play {cardId, target?}`
- `toggleAttacker {id}`, then `confirmAttack`
- `block {blocker, attacker}`, then `confirmBlocks`
- `endPhase`, `concede`, `timeout`
Never enumerate whole block assignments, which explode combinatorially.
- **Empty library** behaviour is data: `libraryEmpty: 'lose' | 'reshuffle' | 'fatigue'`. This keeps option (a), small decks with recursion, open with no code change.
- **`legalActions` is `[]` for the seat without priority.** The host sweep already treats "has legal actions" as "is the one waiting" (`host.ts:264`), so timeouts work without host changes.
**Determinism, rules-as-data, `viewFor` and information hiding all carry over.**
- Hidden information is simpler than in Edit War: there is no simultaneous hidden queue, only the hand and library order.
- **New requirement: `logFor(state, seat)`.** Edit War's `viewFor` returns no log at all (`engine.ts:395-419`). The plan's "dry encyclopedic match log" and the recap both need a per-seat log that masks the opponent's `draw` card ids.
- **Card typing (P31) and "Connected" are not on-chain**, so they must be *frozen into the deck at match start*, exactly as the trend arrow is today (`host.ts:318-334`). Match records already store the derived decks (`host.ts:361-363`), so replay stays exact even if Wikidata changes later.
- **Other host limits:** the host's 2000-action cap (`host.ts:344`) is fine for about 15 turns of atomic actions. `MatchRecord` needs an `engine` field.
- **Multi-engine host** (about 15 lines):
- `ModeHost` holds one `engine` (`host.ts:53`); change to `HostedMode {mode, rules, engine}` and `Lobby.engine`.
- Add an optional `engine: 'editwar' | 'duel'` to `ModeSchema` (`modes.ts:88`, default `editwar`).
- Caveat: the schema is `.strict()`, so un-upgraded peers will count duel modes as `rejected` in discovery. That is acceptable on the testnet.
## 2. "The cards a player owns": the trace and the gaps
**Today's path:**
1. The client computes ownership itself. `src/App.tsx:130` filters `snapshot.blocks` (from `GET /api/state`, `http.ts:68` → `legacySnapshot(limits.legacyBlocks)`) by the **browser-local wallet** address.
2. `PlaySection` gets `owned` (`App.tsx:203`) and merges `cardFromBlock` into the deck-builder pool (`PlaySection.tsx:42`). `cardFromBlock` derives in the browser, marks the card `unchecked` with its cost capped, and computes rarity from `powHash` client-side (`cards.ts:39-46`).
3. Owned cards are used **only in browser-local bot matches** (`MatchScreen`).
**Gaps:**
- **G1. No player decks in PvP.** `ModeHost.join` deals `engine.starterDeck` to both seats for every non-draft policy (`host.ts:144`). The `constructed` policy (`modes.ts:64`, which is also the default, `modes.ts:99`) is **unimplemented**: no create/join route takes a deck (`router.ts:123-131`). Ownership is therefore never verified server-side, because it is never used.
- **G2. Accounts are not wired to the game.**
- Linked wallets live in `accounts.sqlite` (`server/account/store.ts:126-130`), but nothing in `server/game/` reads them.
- `gameRouter` receives only `{dataDir, enrichment}` (`http.ts:154`), with no ledger and no account store.
- The `AccountStore` is created privately inside `createAccountRouter` (`router.ts:19`), so it must be injectable to be shared.
- The `auth` middleware is local to that router (`router.ts:25-33`); export it, or call `verifyIdToken` directly.
- **G3. Host identity is not the account.** The player id is `sha256(localStorage secret HMAC hostOrigin)` (`online.ts:161`, `host.ts:380`), so a ladder keyed on it can be farmed trivially by sybils. Also, `Authorization: Bearer` is already taken by seat tokens with a 16–64 character regex (`router.ts:98-101`). Pass the Dex ID token in the **body** at create/join instead.
- **G4. Collection truncation.** `/api/state` returns at most 128 blocks (`server/resources.ts:20`), so older owned cards silently disappear from Play. Use `ledger.deck(address)`, paged by 24 (`ledger.ts:293`), or a new authed `GET /api/game/collection`.
- **G5. Chain card data is a miner claim.**
- The ledger accepts any peer-valid block. `template()` (`ledger.ts:255`) is advisory, and nothing binds a block to a node-issued template.
- `articleSchema` (`consensus.ts:43-56`) does not cross-check `pageId` against `title`.
- `statsSnapshotSchema` (`consensus.ts:59-93`) only checks internal consistency (`deriveStats` of the declared bytes, languages and views).
- So a custom miner can mint "Barack Obama" with any stats. DESIGN §1.10 plausibility exists only as types: `verifyStats` and `check: 'verified'` are never read (no hits outside `types.ts`, `ruleset.ts`, `engine.ts:22` and `derive.ts`).
- **G6. Per-node account stores.** Peer hosts have their own `accounts.sqlite`, so a wallet linked on node A is unknown on node B. Sending a Dex ID token to a peer also lets that peer replay it on other nodes until it expires, since every node shares the `wikideck` audience.
**Recommended (MVP, self host only):**
- Create and join take `{idToken, deck: tokenId[]}`. The host:
1. verifies the token (`verifyIdToken`);
2. loads `wallets(sub)` from the account store;
3. for each token, checks `ledger.nft(token)` (`canonical-index.ts:163`): owner ∈ wallets and `confirmations ≥ rules.minConfirmations` (1 casual, 3 rated);
4. builds `CardInput` **from the block**, never from client card JSON, then derives and freezes.
- Put this in a new `server/game/ownership.ts` exposing a pure `resolveDeck(deps, sub, tokenIds)`. `host.ts` calls it from `startMatch`.
- Plausibility (G5), stretch:
- Check `pageId → title` and `revisionId → size` with **one batched request per 50 cards** (`prop=revisions&revids=…&rvprop=size|ids`). Revisions are immutable, so the result can be cached forever and verifies `pageBytes` exactly.
- Past monthly pageviews are also immutable, so they can be checked exactly at one request per card.
- On a mismatch, the card is marked `unverified` and its cost is capped.
- Peers (stretch): replace the ID token with a wallet-signed deck claim `["wikideck:deck-claim:v1", NETWORK, hostOrigin, nonce, tokenIds]`. The client builds the message itself, like `shared/account-link.ts`. This needs no account on the peer and cannot be replayed on another host.
## 3. PvP: what exists, what's needed, transport
**What exists:**
- **`server/game/host.ts`:** in-memory lobbies, snake draft, the turn timer and AFK handling via `timeoutAction` (`host.ts:263-270`), legal-action enforcement, live-lobby persistence across restarts (`index.ts:205-226`), JSONL match records with replayable audit (`store.ts`) and a W/L leaderboard (`store.ts:64-68`).
- **`router.ts`:** rate-limited (`router.ts:90-94`).
- **UI:** `OnlinePlay.tsx` is a lobby list (polled every 4 s), create/join, draft and a match board. It is reachable only from inside the Play section.
**What's needed:**
| Feature | Minimal design | Files | Size |
|---|---|---|---|
| Invite link | `createLobby({private:true})` is excluded from `openLobbies` (`host.ts:108-112`); the UI route `/play/join/<lobbyId>` works because the SPA catch-all serves `index.html` (`http.ts:157`), and `join` already takes an id | host, router, OnlinePlay | S |
| Quick match | `POST /modes/:id/quick` joins the oldest open lobby within the rating window, else creates one; no separate queue structure | host, router | S |
| Rated ladder | Elo (K=24, start 1200) per mode, keyed by account id, recomputed by replaying `matches.jsonl` in order at startup (deterministic, no new store); rated only when both seats are account-verified and `rated:true` in mode limits; Glicko-2 is not worth the night | store.ts, host.ts | M |
| Account seat identity | `idToken` in the create/join body → player id `sha256('acct:'+iss+sub)`; guests keep the secret but are unrated | host.ts, online.ts | S |
| Duel binding | `server/game/duel-binding.ts` plus multi-engine `HostedMode` | new file, host.ts | S |
**Transport is polling:** the client polls every 1 s (`OnlinePlay.tsx:11, 45-50`), with no WebSocket or SSE anywhere. It is adequate for a sequential duel: about 0.5 s average hand-off latency.
Three caveats:
1. **Cost.** Each poll runs `viewFor` and `legalActions` and returns the full view. Add a monotonic `version` to the lobby and `GET …?v=N` → `204` when unchanged. Poll every 500 ms while waiting on the opponent and every 3 s otherwise.
2. **Do not long-poll under `/api`.** `RpcBudget` caps the whole node at 32 active `/api` requests (`resources.ts:17`, applied at `http.ts:54`), so 16 matches holding long-polls would starve the node. Real push would need SSE mounted *outside* `/api` (a `server/testnet/http.ts` shared-hook change): a stretch only.
3. **Rate limits.** Two humans behind one NAT, or QA agents on one machine, see 1 req/s × 2 against `plays` 240/min and the global 300/min per IP (`http.ts:56`). That is OK, but QA should not run more than 2 matches per IP.
## 4. Wikimedia politeness
**Every call site** (grep over `server`, `src`, `scripts`, `shared` and `android`; the browser never calls Wikimedia APIs directly, because CSP `connect-src 'self'` at `http.ts:50` forbids it):
| # | Where | Upstream | Trigger | Limits today |
|---|---|---|---|---|
| W1 | `server/wiki.ts:73-78` `random()` | en Action API, `generator=random` ×10 | **mint template** `http.ts:122` | ≤4 concurrent, 3 attempts, 3.5 s timeout, no 429/Retry-After, no maxlag |
| W2 | `server/enrichment.ts:134-138` `load()` metadata | en Action API, `pageids=1` | `getFresh` at **mint** `http.ts:123`; `get` from `/api/wikipedia/:id` `http.ts:93-98`; TrendCache | ≤4 active, 8 pending, then **reject**; in-memory LRU of 128 entries, 6 h TTL; errors cached 30 s |
| W3 | `server/enrichment.ts:148` pageviews | `wikimedia.org` REST, monthly | same as W2 | same |
| W4 | `server/game/trends.ts:37-56` via `index.ts:174-186` | W2+W3 | background, one page per 20 s | about 8.6k requests/day steady if the queue stays warm |
| W5 | **browser:** `src/Card.tsx:24` → `src/useEnrichment.ts:10` → `/api/wikipedia/:id` | W2+W3 | **every card face rendered** (collection, explorer, hero) | per-tab cache 5 min; server limit is 60/min **per client IP** only, with no node-wide upstream budget |
| W6 | `scripts/night/card_sample.py:13-38` | en API, Wikidata `wbgetentities`, REST daily views | offline research | 1 req/s, honours Retry-After, disk cache (its docstring says "~4 req/s"; the code does 1) |
| W7 | browser `<img>` from `upload.wikimedia.org` (`Card.tsx:28`, CSP `http.ts:49`) | CDN | each render | hotlinking; acceptable, but the plan should cache thumbnails or use `pithumbsize` |
**Likely cause of the Night 1 429:** W5 fan-out (each grid page means N cards × 2 upstream requests, with a 128-entry LRU thrashing across users) plus W4, multiplied by several testnet nodes sharing one egress IP. Each node also uses the default UA, `Wikideck/0.1 (experimental …)` (`wikimedia-user-agent.ts:4-5`), which has **no contact**. That breaks the Wikimedia UA policy and is the kind of client that gets blocked first.
**Design: `server/wikimedia/client.ts`, one process-wide `WikimediaClient`.**
- **Rate:** a token bucket per upstream host. Defaults (env-configurable):
- `en.wikipedia.org`: **1 req/s, burst 2**
- `www.wikidata.org`: 1 req/s
- `wikimedia.org` REST: 5 req/s
- **Concurrency:** **1 in flight per host** by default (`WIKIDECK_WIKIMEDIA_CONCURRENCY`). The Action API etiquette is serial requests.
- **Co-located nodes:** document `WIKIDECK_WIKIMEDIA_RPS=0.3` for nodes behind a shared IP (testnet compose).
- **Priority lanes**, with a bounded queue (64):
- `mint`: always first; the timeout starts at dispatch; the queue wait is capped at 15 s, and the miner tolerates it (30 s timeout, `miner.ts:57`)
- `interactive`: on overflow, serve stale cache or 503
- `background`: trends, P31, links, plausibility; dropped when the queue is full
- **maxlag:** add `maxlag=5` to every Action API request (en and Wikidata). Detect `error.code==='maxlag'` (HTTP 200 with a JSON error), wait for `Retry-After` (default 5 s) and retry at most 3 times. REST has no maxlag.
- **429 and 503:** parse `Retry-After` (seconds or HTTP-date), then **pause the whole host bucket** (circuit-open) for max(Retry-After, exponential backoff 2^n s with jitter, cap 10 min). Log once per pause and expose a counter in `/healthz`-adjacent ops output.
- **Batching:**
- `pageids=`/`titles=`/`revids=` up to 50 for cheap props: `info`, `pageprops` (`wikibase_item`), `description`, `revisions|size`.
- P31 via `wbgetentities&ids=Q1|…(50)&props=claims` is heavy (Obama's claims alone are megabytes). Use `wbgetclaims&property=P31` per entity (small, cached for 30 days), or one WDQS `VALUES` query per ≤50 QIDs.
- **"Connected"** costs one request per deck pair: `prop=links&titles=<≤50 deck titles>&pltitles=<same>&pllimit=max` returns exactly the intra-set links. Cache it by the sorted pageId set.
- **Never batch** `langlinks` or `categories`: their limits are shared across the batch, which forces continuation and changes `languagesTruncated` semantics.
- **Persistent cache:** `node:sqlite` (already used by accounts; Node 24) at `$WIKIDECK_DATA_DIR/wikimedia-cache.sqlite`. Key = canonical URL with sorted params and without `maxlag`. Store the body, ETag and Last-Modified; send `If-None-Match` (REST returns ETags; the Action API mostly doesn't); serve stale-while-revalidate for `interactive`. TTLs:
| Data | TTL |
|---|---|
| metadata | 24 h |
| monthly pageviews | until month rollover (already the rule, `enrichment.ts:70`); closed months are immutable |
| P31 | 30 d |
| links | 7 d |
| revision size | forever |
| random | never |
Raise the 128-entry cap (`enrichment.ts:51`) once the cache is on disk.
- **UA:** `WikiDeck/<pkg version> (+https://gitea.eom.dev/WikiDeck/wikideck; <WIKIDECK_OPERATOR_CONTACT or PUBLIC_ORIGIN>) node/<ver>`.
- Keep `WIKIDECK_USER_AGENT` as an override, still sanitized as at `wikimedia-user-agent.ts:8-12`.
- Warn at startup if there is no operator contact.
- Update `tests/testnet-user-agent.test.ts`, which pins today's fallback strings.
- **Volume budget:**
- mint: 3 requests (≤5 with random retries)
- match: **0 synchronous** (cache-only at match start; unknown data means "no bonus" or a fallback type, as `trends.ts` already does)
- per new card first seen: ≤1 batched metadata/plausibility share + 1 P31
- per deck save: 1 links request
- card-face render: 0 when cached
**Consensus boundary (checked carefully):**
- **Validation never touches Wikimedia.** `consensus.ts:43-93` is pure, and the only callers of `wikipediaArticles` and `wikipediaEnrichment` are `http.ts:96, 122-123, 154`. Throttling, caching and retry therefore **cannot affect block validation or fork choice**.
- **The mint template's *content* becomes immutable chain data.** The client must be *transport-only* on the mint path:
1. Keep W1/W2/W3 request params byte-identical: `grnlimit`, `exchars=900`, `pithumbsize`, `cllimit=20` (no hidden filter, which feeds `category()` at `wiki.ts:17`), `clshow=!hidden`, `cllimit=50`, `lllimit=500`. They determine `article.category`, `languageCount` and `languagesTruncated`, and the percentile calibration. `tests/enrichment.test.ts:42-52` already pins part of this; add a golden URL test for W1. Adding `maxlag` is safe because it changes only error behaviour.
2. `getFresh` must bypass the persistent cache. `snapshotStats` stamps `capturedAt = new Date()` (`shared/enrichment.ts:41`), so cached data would be misdated.
3. **Existing bug that throttling would amplify.** A 429/5xx on the pageviews call is swallowed into `status:'unavailable'` (`enrichment.ts:145-164`). On the mint path that **permanently mints a card with `attention: null`**. With the new client, mint must fail the template (503, the miner retries) on 429/5xx/timeout, and record "unavailable" only on a genuine 404 or no-data. Both outcomes are valid blocks, so this is template policy, not a consensus change; it needs one owner and a test.
## 5. Scope: 4.5 h, about 10 agents, one integration branch
**Rules learned from Night 1:**
- **No stacked PRs.** Squash-merge makes every child PR conflict. Branch only from `night/2026-11-17` after the dependency has merged.
- **Contracts first:** `shared/duel/types.ts` and `docs/game/DUEL_API.md` merge by 22:15.
- **Node 24 everywhere** (`node:sqlite`; the Dockerfile uses `node:24-alpine`). Check `node -v` before `npm ci`.
- **No new npm dependencies.** The lead alone owns lockfile regeneration.
- **Test files:** top-level `tests/<team>-*.test.ts` only (`scripts/run-tests.mjs:7`).
**Workstreams** (9 workers plus the lead; ✱ = touches a shared file, one line or a small edit):
| WS | Owner paths | Delivers | Depends on |
|---|---|---|---|
| A duel-core | `shared/duel/{types,engine,ruleset,index}.ts`, `tests/duel-core*.test.ts`, `docs/game/DUEL_API.md` | turn and combat engine, `viewFor`/`logFor`, `replay`; **types PR by 22:15**, alpha by 23:15 | — |
| B duel-cards | `shared/duel/{derive,classify,effects,text}.ts`, `tests/duel-cards*.test.ts` | P31 → type table (from `card-sample.json` `instanceOf`, 236 cards), category fallback, P/T and cost curve, ≤8 effect templates, deterministic MTG-register rules text, flavour from `article.extract` | A types |
| C sim-balance | `shared/bots/duel-adapter.ts`, `scripts/sim/*duel*`, `docs/game/DUEL_BALANCE.md` | greedy and MC bots on duel; win rate by type and faction; ruleset number tuning (data-only PRs) | A alpha (23:15) |
| D wikimedia | `server/wikimedia/*`, `server/wiki.ts`, `server/enrichment.ts`, `server/wikimedia-user-agent.ts`, `tests/wikimedia-*.test.ts` | §4 client, mint-path fix, sqlite cache, UA; `trends.ts` unchanged (it goes through enrichment) | — (starts 21:30) |
| E identity | `server/game/identity.ts`, `scripts/night2/prebake-identity.ts` | background P31, links and plausibility via D's `background` lane; persistent per-token identity; pre-bakes testnet and mainnet cards | D interface (stub until merged), B input type |
| F ownership | `server/game/ownership.ts`, `server/game/collection.ts`, ✱`server/account/router.ts` (inject store, export auth), ✱`http.ts` (pass `{ledger, accounts}` into `gameRouter`) | `resolveDeck`, `GET /api/game/collection`, chain-pool draft source (option c) | — |
| G host-pvp | `server/game/{host,router,store,index,duel-binding}.ts`, `shared/game/modes.ts` (`engine` field), `deploy/night2/modes/*.json` | multi-engine host, constructed decks via F, account seats, private lobbies, invite, quick match, Elo, `?v=` polling | F function signature (agree by 22:00), A types |
| H ui-duel | `src/game/duel/*` (board, card face with type line, P/T, text, flavour; adapter; local bot practice) | playable vs bot in the browser | A types → A alpha |
| I ui-pvp+story | `src/game/OnlinePlay.tsx`, `online.ts`, `PlaySection.tsx`, `Leaderboards.tsx`, `src/game/recap/*`, `shared/duel/recap.ts` | collection deck picker, invite and quick-match UI, ladder, "Battle of X (2026)" recap rendered client-side by `replay` of `GET /matches/:id` (`router.ts:137`) | G API (stub), A log format |
| Lead | integration branch, `docs/night/*`, QA, REPORT | merges, lockfile, end-to-end on testnet with 2 Dex accounts | — |
Overlap is limited to the ✱ lines. `host.ts` is G's alone: F ships pure functions that G calls.
**Timeline (EST):**
| Time | Milestone |
|---|---|
| 21:30 | A, D and F start |
| 22:15 | types contract merged; B, G, H and I build against it |
| 23:15 | engine alpha; C starts sims; H plays vs bot |
| 00:00 | D merged; E live; G binds duel; I wired to G |
| 00:45 | balance v1 (data-only tuning) |
| 01:15 | **freeze** |
| 01:15–01:45 | QA PvP plus recap |
| 01:50 | REPORT |
| 02:00 | stop |
**Cut first, in order:**
1. WDQS/shared-property "Connected": keep article-link Connected only, or none.
2. Plausibility verification (G5): trust chain stats and document the gap.
3. Quick match: invite link and open lobbies suffice.
4. Elo ladder: keep W/L keyed by account.
5. Peer-hosted duel and wallet-signed deck claims.
6. Recap infobox polish: keep the plain encyclopedic log.
**Never cut:**
- the Wikimedia client (requirement 7: low risk, independent, and protects every later night)
- server-side ownership verification
- the duel core
- self-hosted PvP with owned or draft decks
- bot practice
**Biggest risks:**
1. Duel engine plus balance in about 2 h of real build time. Mitigation: no stack, atomic actions, effects as ≤8 templates.
2. UI combat readability. H must start on mock views at 22:15, not wait for the alpha.
3. Supply: 12 mainnet cards make a constructed duel unplayable without option (c) or (a). Make deck size and `libraryEmpty` ruleset data so Eric's decision is a mode file, not code.
+45
View File
@@ -0,0 +1,45 @@
# Node resource controls — stage 1
This stage bounds public RPC transport/admission and browser state loading without changing consensus, chain identity, or launch approvals. **It does not make the node production-ready.**
## Admission and transport
API and P2P traffic have separate process-local token buckets. API saturation does not spend P2P tokens. Both lanes still share the event loop, ledger mutation queue, native engine and host resources: this is admission separation, not CPU/process isolation. Budgets are global per process rather than per wallet/IP, so changing wallet or source IP cannot bypass the global limit. Existing per-IP limits remain in place; no permissive proxy trust was enabled.
Admission precedes JSON parsing, signature verification and route work. Excess traffic receives 503 plus `Retry-After: 1`; there is no unbounded retry queue. JSON request bodies remain capped at 68 KiB and compressed request inflation is disabled. Response JSON is capped before transmission. Byte budget is charged on serialized JSON bytes, not network/TLS framing. Small error responses are not charged recursively when rejecting a payload/budget; edge limits are still needed.
| Variable | Default | Accepted range |
| --- | --- | --- |
| `WIKIDECK_RPC_REQUESTS_PER_SECOND` | 120 | 1–1000 |
| `WIKIDECK_RPC_WRITES_PER_SECOND` | 10 | 1–100 |
| `WIKIDECK_RPC_MAX_ACTIVE` | 32 | 1–256 |
| `WIKIDECK_RPC_RESPONSE_BYTES` | 2097152 (2 MiB) | 128 KiB–2 MiB |
| `WIKIDECK_RPC_BYTES_PER_SECOND` | 8388608 (8 MiB/s) | 2–128 MiB/s |
| `WIKIDECK_RPC_LEGACY_BLOCKS` | 128 | 0–128 |
Token bursts equal one second of the configured rate. P2P uses a separate fixed lane of 40 requests/s, 10 writes/s and 16 active HTTP responses, with the same configured byte limits. Active slots are released on response finish/connection close; asynchronous route work already admitted may still be completing after disconnection. Existing bounded native, ledger and Wikimedia queues/concurrency remain necessary.
HTTP has at most 256 accepted connections, 10-second headers timeout, 15-second request timeout, 5-second idle keepalive and a 30-second socket inactivity timeout. These are process transport protections, not a substitute for proxy timeouts/connection limits, global ingress limits, OS limits or load-balanced per-instance policy. Body parsing is no longer ahead of admission. Health bypasses the API lanes for liveness, but still shares server connections and CPU.
Configuration is validated before opening node resources. Lower limits can shed legitimate traffic, including sync; test settings against the intended topology before deploying. Do not raise limits merely to hide overload. Custom settings require container environment forwarding when deploying with Compose.
## Bounded state API
- `GET /api/summary`: network/chain/genesis, canonical tip/height/work, process-local revision, peers/maturity and bounded pending transactions. No full-chain copy. Revision is only a local cache invalidation counter, not consensus or a durable sequence.
- `GET /api/canonical?tip=<hash>&offset=0&limit=16`: up to **16 canonical blocks**, pinned to the current tip. Tip changes return 409 so the caller restarts. Offset cannot exceed current height. Each page uses the cached canonical branch and clones only its slice. The first page after a new tip may rebuild that branch in O(height).
- Legacy `/api/state` and `/api/chain` still return full snapshots only up to the configured legacy height and response-byte cap. Beyond it, they return 413 with paging instructions **before full snapshot cloning**. They never silently truncate ownership history. Legacy scripts/API clients must migrate to paging for longer chains.
- `/p2p/status` and `/p2p/mempool` now obtain only the fields needed instead of cloning the chain. P2P block inventory semantics are unchanged.
The browser and trade signing refresh use bounded streaming response reads (2 MiB/page), pagination/network/tip/parent continuity checks, one restart on tip conflict, shared in-flight loads, and append-history cache reuse. A final summary confirms the loaded tip. These checks are **not independent PoW/consensus verification**: the UI still trusts the node.
The existing snapshot-based UI still assembles complete ownership history, capped at **10,000 blocks or 64 MiB of encoded history** with a 30-second fetch deadline. If exceeded, it fails explicitly rather than presenting partial ownership as complete. That is a client resource guard, not a solution for an archival-sized wallet/explorer. Full state objects and parsed JS memory cost more than encoded JSON. A truly indexed wallet and fully paged explorer remain required before closing SEC-01. An unchanged refresh requests two summaries and no block pages; append loading verifies the cached prefix and requests only new blocks.
## Tests and safe load tooling
Final validation: **86 tests passed, 0 failed, 0 skipped** in the Node 24 Docker test image, including real RandomX HTTP integration. TypeScript/Vite build passed. The suite includes deterministic token refill/write/active/byte limits; a held-response 40-request loopback burst (4 admitted, 36 shed); health/P2P-lane availability during API saturation; canonical paging defensive copies/tip conflicts/legacy guards; client append caching, malformed pages/continuity, bounded history, single-flight loading and response size checks; non-recursive oversized-response errors; and paging against the real RandomX HTTP fixture.
`test:rpc-load` invokes the read-only [local smoke-load script](../scripts/rpc-load.ts). It requires `--consent`, accepts only **literal loopback IP origins** (not remote hosts or DNS names), at most 32 workers and 1,000 total requests. Options: `--node`, `--concurrency`, `--requests`. It reads summary and canonical-chain endpoints and checks health, with response/time bounds. It never submits a block or transaction. Use only a disposable isolated node: even read traffic consumes CPU/IO. Output includes status counts, bytes and p50/p95 latency; rejected load is visible, not counted as successful capacity.
Initial smoke measurements on a disposable empty-chain loopback Docker node are historical local baselines, not performance guarantees. They do not measure hashing load, long-chain costs, or hostile Internet traffic. The deterministic held-response test separately verifies overload shedding and peer/health-lane availability. Final test totals must be rerun against the selected release artifact.
Neither measurement establishes that a realistic long-chain public deployment meets its latency/throughput SLO. Production-scale chain/journal fixtures, slow clients, adversarial payloads, sustained concurrency, disk pressure and multi-host recovery measurements remain outstanding. No launch-readiness finding is marked resolved by this stage.
+42
View File
@@ -0,0 +1,42 @@
# Deployment, private monitoring, and recovery
**Production remains NO-GO.** These procedures are operator preparation, not an installed public edge, configured alerts, independent review, or deployment approval. Do not remove persistent node volumes as part of source cleanup.
## Reverse-proxy trust
`WIKIDECK_TRUSTED_PROXIES` is empty/untrusted by default. It accepts up to 16 comma-separated literal IPs or bounded CIDRs. Configure only the actual source address of a trusted edge connection. The edge must overwrite forwarded headers, and firewall rules must prevent direct backend access. A trusted-proxy setting does not configure the firewall.
The [nginx example](../deploy/nginx.conf.example) is not installed or deployment-validated. Replace its reserved hostname and certificate paths, and validate TLS, renewal, forwarded-IP handling, body/connection/time limits, and edge admission controls on the intended host. Container proxy source addresses may differ from host loopback; inspect the actual topology instead of copying a trust list blindly.
## Private diagnostics
Set `WIKIDECK_OPS_PORT` to an unused port (1–65535) to enable a separate HTTP listener bound to `127.0.0.1`; leave it unset to disable it. Never publish or reverse-proxy this port. Only `GET /status` is supported. It returns bounded, no-store aggregate JSON; any local process with loopback access can read it.
Reported diagnostics include node uptime/RSS, event-loop delay, canonical height and tip age, DAG and pending counts, native-engine status, configured/connected peers, RPC budgets, and available disk plus ledger database/WAL sizes. Native subprocess memory is not included in Node RSS. Peer count is not a security score. Health is liveness, not sync readiness or consensus assurance.
`ops:check -- --port <private-port>` reads only literal loopback with bounded time and response size. It emits warnings for stale/empty chain, peer isolation, native-engine failure, queue pressure, event-loop delay, and unavailable/low disk. Thresholds are starting points, not production SLOs. Route alerts to an external system and test delivery; this repository does not install a scheduler or alert destination.
## Stopped-node ledger backup
`node:backup` supports `create`, `verify`, and `restore`. New version-2 packages contain only the selected network's ledger database. Stop the node cleanly first; the tool takes the same exclusive network lock as startup, uses SQLite `VACUUM INTO`, checks SQLite integrity and chain metadata, hashes the snapshot, and writes the manifest last. It refuses to overwrite a destination, restore in place, or run when a node lock exists. Restore targets must not exist.
Example workflow, using the `node:backup` package command with explicit `--source`, `--destination`, and `--consent` as required by the CLI:
1. Stop the node and confirm all writers to its data directory are stopped.
2. Create the backup in a new directory outside the node data directory. Keep wallet backups separately; this tool does not include browser wallets, secrets, configuration, binaries, or logs.
3. Run `verify` against the backup before copying it off-host.
4. Restore only to a new isolated data directory. The tool reports `proofsRevalidated: false`; reopen with the trusted RandomX implementation and verify the chain before resuming service.
Backups are not encrypted or authenticated by this tool. Store them in restricted, encrypted off-host storage, protect the manifest separately, and test cold restoration. Hashes detect accidental corruption, not a coherent malicious replacement. Filesystem, power-loss, Windows ACL, and network-share durability need validation on the actual storage platform.
## Retiring legacy financial data safely
Removing market/custody source code does **not** delete or resolve data left by an older deployment. The card-only node does not open or reconcile those files. If the selected network directory contains `market.sqlite`, `custody-sandbox.sqlite`, or a `custody-postgres.json` binding, new ledger-only backup creation refuses to proceed rather than silently omit those records. Preserve the existing volume and any PostgreSQL database; arrange an operator-reviewed archive before retiring it.
Historical version-1 paired backups containing `ledger.sqlite` and `market.sqlite` remain verifiable and restorable as opaque files for preservation. Restoration does not reactivate market services or imply that financial obligations have been resolved. External PostgreSQL data is not contained in a SQLite package. Do not delete old volumes, journals, database credentials, or backups as part of this source cleanup.
An operator must choose retention, recovery point/time objectives, off-host storage, and restore ownership. Never remove a lock or replace a running volume just to make recovery succeed.
## Remaining operational evidence
Verify the intended HTTPS/proxy/firewall topology, peer and tip freshness, alert delivery, disk growth, backup encryption/retention, and restore/replay time on the actual host. Assign an incident responder and document stop/rollback criteria. See the [card-only production plan](card-only-production-plan.md), [operator launch register](operator-launch.md), and [security review](security-review.md). No startup or payment approval is granted by these tools.

Some files were not shown because too many files have changed in this diff Show More