Live Play
Natural-language intent, state-derived actions, consequences, continuity, and responsive narration.
GreenKappa Labs flagship project
A tabletop engine where the AI is creative—and the game remains accountable.
TableArc connects natural-language play to canonical campaign state, rules-aware procedure, characters, maps, combat, consequences, persistence, and debug proof.
Electron desktop
React + Vite
OpenAI + Ollama
Local-first direction
Active prototype · independently developed
Core ideaNatural language above structured state
Provider modelGeneric OpenAI and Ollama workflows
Latest documented validationBuild 1.1.28 acceptance path
Built byTrevor Leininger · GreenKappa Labs
What it is
Not an AI narrator with a character sheet attached.
TableArc is a local-first desktop AI tabletop engine. Players interact naturally, but the durable truth of the game lives in explicit state: actors, scenes, clocks, clues, resources, combat, spatial relationships, consequences, and saves.
The product thesisThe AI imagines.
The engine remembers, adjudicates, maps, resolves, and proves.
The product
A real desktop application, not a concept mockup.
TableArc brings campaign setup, generation, review, live play, character workflows, saves, provider settings, and debugging into one desktop shell.
Runtime architecture
State first. Procedure second. Narration third.
The model does not get to declare durable game truth by itself. The runtime owns what changed, why it changed, and what later systems are allowed to assume.
Adventure Foundry
Generation, normalization, bounded repair, validation, atomic admission, and opening-scene readiness.
Character authority
Active-PC identity, sheets, resources, inventory, passives, features, party context, and repair workflows.
Maps and combat
Spatial truth, range, exits, visibility, initiative, targeting, conditions, damage, and aftermath.
Persistent narrative state
Clocks, clues, evidence, relationships, consequences, scene memory, saves, and restart continuity.
Provider flexibility
A generic OpenAI and Ollama contract serving live play, generation, repair, rules, recap, state, and debug.
Current state
Honest about what is proven—and what is next.
TableArc is an active prototype. The site distinguishes documented validation from broader product intent instead of presenting every roadmap item as finished.
End-to-end acceptance path
- Adventure generation from an empty profile
- Bounded completion and repair
- Generated-PC promotion
- Runtime commit and Live Play start
- State-derived action rebuilding
- Save and restart continuity
Reliability before expansion
- Longer real-play acceptance
- Provider and model reliability
- Committed-scene presentation
- Accessibility and performance
- Save migration and distribution polish
- Broader DM Assistant workflows
Why it matters
Fluent prose is easy to mistake for a functioning game.
TableArc is built around the harder problem: preserving the freedom and responsiveness of tabletop play while keeping rules, state, space, consequences, saves, and authorship coherent enough to trust.
Local-first controlUse local providers and keep the desktop experience useful without a hosted platform.
Authored structure survivesAI adds responsiveness without silently replacing the adventure’s logic.
Debug is a product featureShow the causal chain instead of hiding failures behind plausible prose.
State creates continuityThe world remembers because durable systems—not conversational vibes—own the truth.
Built independently
Trevor Leininger · GreenKappa Labs
TableArc demonstrates product architecture, React/Vite/Electron development, OpenAI and Ollama integration, structured state modeling, provider contracts, testing strategy, debugging, technical documentation, interface design, and sustained iterative repair across a large application.
Contact
Discuss TableArc or adjacent work.
Reach out about AI tabletop systems, local-first applications, authored-adventure tooling, structured runtime state, interactive media, or applied AI engineering.