Skip to content

roadmap.md

Latest commit

 

History

History
74 lines (55 loc) · 3.11 KB

roadmap.md

File metadata and controls

74 lines (55 loc) · 3.11 KB

Roadmap

Planned features and their current status.

Feature Status
React Visualization Component Done
Studio Application Done
AI-Ready Integrations (Phase 1 — MCP) Done
AI-Ready Integrations (Phase 2–4) Planned

React Visualization Component

@bt-studio/react - A React component for real-time tree execution visualization. See React Debugger docs.

  • Live tree rendering with node status coloring
  • Time-travel debugging: navigate previous ticks and inspect full tree state at any point
  • Node detail sidebar with result distribution, display state, and tick history
  • Ref mutation tracing panel
  • Display of stateful decorator/composite internal state (via getDisplayState())
  • Powered by the existing Inspector system

Studio Application

@bt-studio/studio-ui + @bt-studio/cli - Standalone debugging app (bt-studio CLI) that launches the server and browser UI.

  • Connect to running behavior trees via WebSocket or TCP transports
  • Real-time execution trace visualization with time-travel debugging
  • CPU profiling with flame graphs
  • Tick retention (configurable, default 100k ticks per tree)
  • Node search and filtering by name, tag, or flag
  • Depends on: React Visualization Component

AI-Ready Integrations ( Draft )

Make bt-studio first-class for AI agents implementing and debugging behavior trees.

Phase 1 - MCP foundation

  • Build a studio-server MCP for runtime interaction (tree list, tree snapshots, tick playback, profiling, node/ref inspection).
  • Build a web MCP for Studio UI so agents can drive UI workflows (open tree, select node, inspect timeline, export traces).
  • Expose MCP tools + resources + prompts (not only tools), so different clients can choose the best interaction mode.
  • Support both MCP transports:
    • stdio for local/dev workflows.
    • Streamable HTTP for hosted/remote workflows.

Phase 2 - Compatibility + security

  • Add a ChatGPT apps/deep-research compatibility mode with read-only search/fetch tools.
  • Implement OAuth 2.1 for remote MCP usage, including PKCE and auth metadata discovery.
  • Add transport hardening for remote/local servers:
    • Origin validation for HTTP transport.
    • Localhost-first defaults for local runs.
    • Explicit auth for non-local access.
  • Keep human confirmation for sensitive/mutating tool calls.

Phase 3 - Documentation for agents

  • Ship a documentation MCP (read-only) with searchable docs + examples.
  • Publish /llms.txt and keep docs consumable as clean markdown where possible.
  • Provide importable agent guidance assets:
    • AGENTS.md snippets for preferred MCP usage patterns.
    • Optional SKILL.md package for “implement/debug behavior trees with this library”.

Phase 4 - Discoverability + quality

  • Publish official MCP servers in the Official MCP Registry.
  • Add an AI compatibility test matrix (Codex, OpenAI API/Apps, Claude clients).
  • Add AI evals for common tasks:
    • Build a tree from plain-language requirements.
    • Diagnose failing/stuck nodes.
    • Explain lifecycle/ref changes with citations to docs/resources.