Files
wikirpg/DEVELOPMENT_PLAN.md
2026-11-01 15:22:21 -05:00

25 KiB
Executable File

WikiRPG development plan

Goal: turn the local prototype into a world people choose to inhabit, then launch that world on the existing Kubernetes platform with Dex SSO.

This is a gameplay-first roadmap for one human operator and one AI coding assistant. It does not assume an external review team, a separate operations team, or a large launch. Work is organized into playable improvements, not a certification program. Deployment is the final delivery step, not the game's main feature.

Status: proposed roadmap, November 2026. The current implementation is described in README.md. None of the milestones below should be treated as completed simply because the prototype has a related button.

What the prototype is missing

The prototype proves that a character can travel, wait for work to finish, trade, and read linked events. It does not yet give players enough reasons to do those things.

  • The three locations provide little discovery or geographic choice.
  • Characters have names and biographies but no meaningful capabilities, goals, or differences.
  • Timber and planks lead to one construction counter rather than an ongoing economy.
  • The stag is a personal discovery, not an interactive creature with behavior or consequences.
  • Players cannot communicate, make agreements, form groups, or organize work through the game.
  • Concealment hides an action, but there is no investigation, testimony, or partial revelation to make secrets playable.
  • Activity timers mostly mean waiting rather than planning around a changing world.
  • Events record outcomes but do not yet develop into projects, rivalries, discoveries, and stories worth following.

The next work should address those gaps directly.

The experience to build toward

A player arrives with an aspiration: become a builder, explore the unknown, supply the market, study creatures, or establish a settlement. Nobody assigns a mandatory quest.

They discover that another player needs fittings for a workshop. They can produce them, negotiate a delivery, cooperate with someone who has the right tools, or find a different opportunity. The materials come from another part of the valley. Travel presents a choice between a longer established path and a shorter route with uncertain conditions.

Along the way, they encounter a creature, discover evidence of a hidden journey, or learn something useful about the terrain. What they learn can help their own plans, be shared with friends, or become an authored account in the wiki.

The workshop eventually becomes a real place with an owner, useful capabilities, possessions, and a linked history. Later arrivals can read its story without being handed everyone's secrets.

The core loop is aspiration → preparation → action → consequence → discovery or exchange → a new player-chosen goal. Timers constrain that loop; they must not be its only substance.

Development approach

  • Finish one end-to-end playable slice before adding another subsystem.
  • Prefer rules that produce several uses over one-off scripted content.
  • Add a small amount of authored, coherent world content before attempting unlimited AI generation.
  • Give each new resource, skill, creature, and structure a reason to matter to another player.
  • Keep original game rules and lore; tabletop RPGs are inspiration, not a requirement to reproduce an existing ruleset.
  • Use the existing server-authoritative model. Player-authored prose and AI output cannot create mechanical advantages by themselves.
  • Add regression tests alongside mechanics, especially where actions consume resources, reveal knowledge, or affect another character. This is ordinary development work, not an external-review dependency.
  • Play together after each milestone. Fix confusing or uninteresting decisions before increasing scale.

The ordering below is intentional. Communication and a complete trade loop come before elaborate combat or a huge map.

Milestone 1 — Characters with intentions, players who can cooperate

Outcome: a player has a reason to enter the world and a way to make plans with someone else.

Build

  • A character-creation flow after first Dex login instead of silently adopting the identity provider's display name as the character name.
  • A small original character model: aptitudes, a limited selection of starting proficiencies, and an aspiration. Start with areas such as exploration, craft, and field knowledge; exact values and names are design choices to settle while implementing.
  • Useful starting equipment with explicit, bounded rules. Background text cannot grant wealth, abilities, discoveries, or claims over other characters.
  • Personal goals and notes. Players choose which goals to publish; private notes stay private.
  • In-game communication: local conversation and asynchronous correspondence with clear delivery rules. Separate out-of-character planning from messages characters can act on as knowledge.
  • A public project board for player-authored requests and recruitment. Posting “I need a workshop” is a request, not a system-generated reward or automatically binding contract.
  • A helpful onboarding page explaining the available paths, other players' published needs, and how to begin an undertaking without imposing a quest line.

Keep it small

Start with a few proficiencies and fixed rules. Avoid extensive class trees, mandatory daily tasks, artificial profession monopolies, or a full messaging platform. Every character must have useful ways to act alone as well as opportunities to specialize with others.

Playable check

Two players can create different characters, explain what they want, exchange messages, agree on a task, and find the agreement again later. A new player can choose a first goal without needing an explanation outside the game.

Milestone 2 — A complete production, trade, and delivery loop

Outcome: useful player work creates demand, payment, and something lasting in the world.

Build

  • A compact regional resource chain. For example: timber → planks, ore → fittings, then planks + fittings → equipment or structures. Start with roughly six to eight meaningful recipes, not hundreds of interchangeable items.
  • Tools and workshops that change which tasks are possible or improve their efficiency. Give them costs, locations, and ownership rather than treating them as descriptive page decorations.
  • Item pages with known properties, location, possession, and discoverable provenance. Group fungible stock sensibly so the wiki does not need a separate page for every stick.
  • Carrying capacity, local storage, and transport. Buying remotely can acquire goods at their storage location; it cannot move them to the buyer's pocket.
  • Expanded local market orders with expiration, partial fills, cancellation, and reserved goods or funds. Preserve the prototype's atomic settlement behavior.
  • Delivery contracts with deposited rewards, specified goods, destinations, deadlines, and objective completion conditions. Begin with material delivery rather than subjective services or complicated collateral.
  • Construction that creates functional world entities. A repaired bridge should change travel; a workshop should enable production; a storehouse should provide storage.
  • Recurring demand through useful consumables and reasonable repair costs. Start with a few understandable costs rather than an exhaustive wear simulation.
  • A documented currency supply rule and a viable path for new players to earn from existing players. The demo's 100-crown allocations are not the production monetary policy.

Economic rules to settle here

  • What is initially issued, to whom, and under what account-scoped limit?
  • How can a player with no money earn it through useful work?
  • If currency is removed by system fees, what supplies it over the long term? A fee paid to a player treasury is redistribution, not destruction.
  • How do resources regenerate or move without allowing unlimited effortless extraction?
  • Which prices and market histories are intentionally published, and who can see them?

Keep currency closed to the game. Blockchain, real-money cash-out, lending, insurance, and sophisticated financial instruments are not necessary to make this loop interesting.

Playable check

One player produces something another genuinely needs. They agree on a price or delivery, complete it asynchronously, and the buyer uses the goods in a structure or activity that has a real effect. Failed or canceled contracts have understandable outcomes. A player can explain why a material has value without referring to an arbitrary system reward.

Milestone 3 — A valley worth exploring and time worth planning around

Outcome: movement and timing create choices, discoveries, and opportunities rather than interchangeable waits.

Build

  • Expand to a coherent small region: approximately eight to twelve places across a few terrain types, with routes, shelters, resources, and landmarks. Treat these numbers as a content budget, not a promise that more pages mean more fun.
  • Distinguish public geography from locations that require discovery. The current prototype publishes its entire three-place atlas; that cannot remain the rule for undiscovered areas.
  • Route progress and stopping points so canceled travel no longer teleports a character back to the origin. Show only the route information a character knows.
  • Give routes different time, carrying, and environmental trade-offs. Improve a path or bridge through player work rather than adding arbitrary travel fees everywhere.
  • Make the shared day/night cycle affect visibility, exploration, resource access, or creature behavior through explicit rules.
  • Add a small weather model only where it changes a decision: whether to travel, shelter, investigate, or prepare equipment. Avoid weather text that has no consequence.
  • Exploration results such as tracks, ruins, useful routes, material deposits, and field observations. Record observations with source and time; a past sighting is not live tracking.
  • An activity planner showing knowable duration, inputs, cancellation costs, and relevant daylight forecasts before commitment.
  • Bounded offline intentions and nonlethal interruption responses for travel. Do not require players to be awake at a particular real-world hour to avoid disaster.

Time rules

Retain the proposed 20-real-hour day while testing. Choose production activity durations based on how much meaningful planning and cooperation they allow, not by multiplying every prototype timer by the same constant. Mix short activities with longer projects and separate contributions.

Players can read and plan while occupied. Do not solve waiting by allowing incompatible physical actions or limitless unattended queues. Long activities need either useful preparation and anticipation or intermediate decisions—not constant compulsory check-ins.

Playable check

Two routes have intelligible advantages. Preparation or timing changes the outcome. A discovery helps a player achieve a goal or gives them information worth sharing. Players in different timezones can contribute to the same project without needing to log in simultaneously.

Milestone 4 — Creatures, encounters, and character growth

Outcome: the environment provides interesting encounters even when other players are offline.

Build

  • A few distinct creature types with habitats, schedules, motivations, and rule-defined capabilities. Start with a peaceful creature, an avoidable territorial threat, and an encounter that rewards cooperation.
  • Encounter choices: observe, prepare, avoid, retreat, or intervene where the rules permit. Do not make every creature a combat resource dispenser.
  • A small original resolution system with skill, equipment, uncertainty, and clear consequences. Save committed outcomes so refreshing a page cannot reroll an encounter.
  • Nonlethal setbacks initially: lost time, damaged equipment, temporary injury, or a blocked route. These can support crafting and preparation without immediately creating permanent loss of a character.
  • Progression through relevant practice and discoveries, with explicit limits against repeating trivial actions for unlimited gain. Aim for broader options and specialization, not runaway power gaps.
  • Creature knowledge pages that grow through observation rather than exposing every statistic and weakness on first contact.
  • Encounter histories that show choices and consequences to permitted viewers.

AI can vary descriptions and help propose encounter content; mechanics remain authoritative. Until offline responses are implemented, do not introduce unattended encounters that require an immediate player decision.

Playable check

An encounter supports more than one reasonable approach. Preparation, skill, or cooperation matters. Its consequences give the player a new decision rather than simply ending the timer or awarding coins.

Milestone 5 — Secrets that can become stories

Outcome: hidden travel, investigation, and information exchange become gameplay—not just missing entries in a feed.

Build

  • A knowledge model for facts, observations, rumors, and authored accounts. Track who learned what, from whom, and when.
  • Concealment and detection rules that consider relevant location, conditions, and character capabilities. Do not automatically reveal a hidden character merely because someone browses a page.
  • Clues and partial event accounts. Discovering missing goods may reveal a loss without naming the thief; tracks may suggest a route without revealing every traveler.
  • Investigation actions that produce useful evidence or honest uncertainty. Evidence must follow established events instead of AI inventing a culprit.
  • Explicit ways to share testimony, publish discoveries, write books, and keep records private. Receiving testimony does not automatically grant all underlying secret records.
  • Knowledge-aware search, backlinks, timelines, and notifications. Keep unavailable entities and revealing metadata out of the response rather than cosmetically censoring them.
  • Small, explicitly agreed player-conflict scenarios before enabling theft generally. Define what can be taken, opportunities to notice or respond, and protections for offline characters.

Keep it small

First prove hidden travel → clue → investigation → partial disclosure → player response. Enable theft only after that loop works and its rules are agreed. Protected accounts and carried physical valuables are different targets; do not make every balance stealable by default.

Playable check

One player keeps a journey secret; another learns something through a valid action. The discovery changes a decision without exposing unrelated private history. A player can share or dispute an account, and the wiki distinguishes that account from objective truth.

Milestone 6 — Groups, ownership, and a world players can shape

Outcome: cooperation develops into persistent social institutions and player-authored places.

Build

  • Organizations with explicit roles, permissions, membership history, shared storage, and a treasury.
  • Joint projects with independent contributions and clear ownership outcomes. Do not require all contributors to be present together.
  • Controlled customization of owned structures: names, descriptions, permitted functions, and authored records. Mechanical changes require construction or another supported action.
  • Agreements about access and use, followed by a small set of settlement capabilities. Do not begin with unrestricted territorial conquest.
  • Worldbuilding proposals checked against existing lore and mechanics. Keep drafts distinct from canon; players cannot invent deposits or royal authority by editing a page.
  • A coherent AI worldbuilding workflow: generate constrained drafts, validate them, and accept them into canon through the same domain rules as authored additions. Start with bounded expansion beyond the established region.
  • An ambition/project view that connects player needs and discoveries without turning the game into mandatory AI-generated quests.

AI integration order

  1. Narrate already resolved events using only the viewer's authorized context, with deterministic fallback.
  2. Produce editable drafts for descriptions and worldbuilding proposals.
  3. Interpret player intentions into supported action proposals that show costs and require commitment.

Do not give a model direct balance-editing, database, or unrestricted world-mutation tools. Apply per-request and daily cost limits. AI failure should not prevent trading, travel completion, or reading established history. Log only what is necessary and do not send another character's secrets to a player-facing prompt.

Playable check

A group completes something no one member could conveniently supply alone. The result creates new play opportunities and can be found through its wiki history. Player writing enriches the world without contradicting established facts or granting free advantages.

Milestone 7 — Lives, legacies, and a readable chronicle

Outcome: characters matter as people with histories, and the wiki makes those histories worth exploring.

Build

  • Richer event records: intention, relevant commitments, outcome, participants, known causes, and consequences, separated from viewer-specific accounts.
  • Linked project and story arcs assembled from real events, not invented retroactive history. Readers can go from a place to an event to a participant to another part of their story.
  • Useful histories for objects, structures, discoveries, and organizations—not just chronological lists of generic “work begun” entries.
  • A distinction between changing public biography and a character's accumulated historical record.
  • A complete successor flow when a character dies, preserving one living character per account and keeping the previous life available only as knowledge permits.
  • Explicit estate rules before lethal gameplay: carried goods, protected funds, escrow orders, contract obligations, and organization roles. Remove the prototype's assumption that a successor simply controls the same account balance.
  • Death warnings, permissible causes, offline protections, and consequences agreed before enabling permanent death. The game must remain playable when a player is left without a living character.

Permanent death is part of the intended lifecycle, but lethal play is not a requirement for the first release. Keep it disabled until succession and estate settlement are playable; do not rush it just to add stakes. Character growth, choices, friendships, and construction already provide meaningful stakes.

Playable check

A player can browse a structure's page and understand how it came to exist, who contributed, and what happened next. Character histories contain recognizable decisions. A successor cannot obtain private knowledge or repeated starter rewards from a previous life, and unresolved obligations do not disappear through death.

First release scope

The first release should be a small persistent sandbox, not the completion of every milestone.

Needed for an inviting first release

  • Character creation, goals, and a few useful proficiencies.
  • Communication and player-authored project requests.
  • A coherent small region with discovery and meaningful routes.
  • A short production chain, useful tools and structures, local trade, storage, and delivery.
  • Day/night with at least one real gameplay consequence and clear activity timing.
  • A few interactive creatures with nonlethal outcomes.
  • Linked histories, private knowledge, and the first investigation/share-information loop.
  • Basic group cooperation and player-authored descriptions.
  • Dex login and a persistent deployment on the existing platform.

Safe to defer

  • General PvP theft, lethal encounters, and full inheritance.
  • Territorial warfare, elaborate combat, large skill trees, and complex finance.
  • A vast map, detailed seasonal simulation, and unlimited procedural generation.
  • Fully autonomous AI adjudication. Rules-based narration can remain available even after a model is connected.
  • Blockchain and real-money currency exchange.

Basic safety, privacy, and correct accounting are requirements of the playable mechanics, not separate feature expansions. We should fix a confirmed leak or duplication bug before inviting players, but do not need an external sign-off to continue development.

How we decide whether it is becoming fun

After each slice, play a few ordinary sessions together. Once the core loops work, invite a small group to play rather than to perform a formal external review.

Ask:

  • Did players choose their own goals, or run out of things to do after onboarding?
  • Did someone want what another player made or discovered?
  • Did travel, preparation, information, or cooperation change a decision?
  • Did occupied time create useful planning, or just an empty wait?
  • Did events create a reason to return and follow the story?
  • Could an occasional player still participate without rearranging their real life?
  • Could players understand the rules and consequences through the interface?

Watch for concrete failures: no buyers, irrelevant resources, compulsory chores, overwhelming text, unreadable histories, repetitive encounters, and social coordination that only works outside the game. Fix those before adding content volume. Returning by choice matters more than maximizing daily logins.

Do not interpret a busy market or an elaborate ruleset as proof of a good game. Small groups produce thin markets; adapt the initial recipe and resource mix to actual demand rather than inventing NPC shoppers to fake it.

A deliberately short path to Kubernetes

Use the platform's existing Dex and deployment conventions. Verify the necessary cluster settings when we implement deployment; this plan does not claim the cluster already has particular database, ingress, storage, or backup capabilities.

Before the first persistent invitation:

  • Package a reproducible application image and deploy it with production configuration and Dex, never the demo account switcher. Start a fresh world without the seeded demo accounts and allocations; the current constructor seeds them even in Dex mode.
  • Support the actual Service/Ingress connection model. The current loopback listener is suitable for local preview, not an ordinary separate Kubernetes Service. Configure container binding, HTTPS ingress, the expected public host, and the exact Dex callback together.
  • Select a transactional persistence arrangement appropriate for a small world. Do not run multiple copies of the current in-memory/SQLite snapshot writer or rely on a replica count alone to prevent overlapping writers during updates. Use an enforced single-writer lifecycle initially, or move shared state into database transactions before allowing concurrent writers.
  • Preserve deadlines and complete activities once across restarts; keep the world epoch stable. Make data migrations and the supported request-retry window explicit instead of silently evicting old command protection after 1,000 requests.
  • Provide a basic health check, bounded request handling, and an understandable maintenance mode. Address proxy-aware rate limits: the current per-socket limit would group players behind one ingress address.
  • Keep secrets out of the repo, verify live Dex login/logout and stable account binding, and make account restrictions/revocation effective within a documented interval. Decide whether restarting sessions is acceptable for the small release or whether sessions need shared persistence.
  • Save recoverable backups outside the application's failure domain and perform an actual restore before calling the world persistent. Keep the previous working image and a data-compatible rollback or forward-fix procedure.
  • Run gameplay regression tests and a manual two-account pass for trade, hidden information, activity completion, and character availability. Fix observed failures before exposing the affected mechanic.
  • Set basic storage/resource limits, report serious errors and activity-processing failures, and provide one way for players to report problems. Add an operator-only maintenance control to pause affected mutations during a problem; do not silently rewrite player history to conceal a repair.

No separate external-review milestone is assumed. We will launch modestly, be honest about unfinished features and possible resets, and keep the enabled feature set within what one operator can support. Real-money assets remain outside that scope.

Next implementation slice

Start with Milestone 1 and the smallest part of Milestone 2 needed to complete a player-requested job.

  1. Replace automatic character creation on first login with a character setup screen and basic proficiencies.
  2. Add private notes, a published aspiration, and local/asynchronous communication.
  3. Add a player-created material request with a funded, objective delivery condition.
  4. Add one additional material and a functional workshop that consumes delivered inputs.
  5. Connect request, messages, delivery, construction, and participants into the wiki history with appropriate visibility.

The deliverable is not “five new screens.” It is two players making an agreement, supplying something useful, changing the world, and being able to read that story afterward.