This is the front door for the repository. Start here when you need to understand where CoreAI lives, which package owns a feature, or which document answers a specific integration question.
The short version: CoreAI is split into a portable AI/runtime layer, a Unity-facing package, an example game, and planning material. Each package keeps its own deep index; this page points you to the right one before you spend time reading the wrong folder.
| Area | Entry Point | Use When |
|---|---|---|
| CoreAI Unity package | Assets/CoreAiUnity/Docs/DOCS_INDEX.md | You need setup, chat UI, streaming, settings, architecture, tests, or Unity integration docs. |
| Portable CoreAI package | Assets/CoreAI/Docs/README.md | You need host-agnostic contracts: agents, tools, routing, MEAI, Lua sandbox, and tool-calling rules. |
| CoreAI Mods (Lua, optional) | INSTALL.md §3 · LUA_SANDBOX_SECURITY.md | You need the com.neoxider.coreaimods package: Lua sandbox, execute_lua/manage_mods tools, mod runtime. |
| CoreAI Hub (UI Toolkit, optional) | INSTALL.md §4 | You need the com.neoxider.coreaihub tabbed Hub window and its page registry. |
| CoreAI Benchmark (dev-only) | Assets/CoreAIBenchmark/README.md | You need the com.neoxider.coreaibenchmark LLM game-creation benchmark harness. |
| Example game | Assets/_exampleGame/README.md | You need the RogueliteArena sample, scene setup, progression, or AI wave planning notes. |
| NeoxiderTools | Assets/NeoxiderTools/Docs/README.md | You need shared utility/toolkit documentation used by this Unity project. |
The full five-package dependency graph and install profiles (Base / +Mods / +Hub / Full) live in INSTALL.md.
Use these paths when you have a concrete task:
| Goal | Read In This Order |
|---|---|
| Install or wire CoreAI in a Unity scene | CoreAiUnity docs index -> quick start -> settings/runtime architecture |
| Debug LLM requests, streaming, or WebGL SSE | CoreAiUnity streaming docs -> CoreAI routing docs -> transport-specific source comments |
| Add a new LLM tool | Tool-calling best practices -> AgentBuilder -> relevant tests |
| Expose Lua or AI-authored scripts | Lua sandbox security -> runtime binding code -> sandbox tests |
| Decide what to build next | Module audit -> TODO -> orchestration plan |
| Document | Purpose |
|---|---|
| TODO.md | Live backlog, completed archive, and remaining feature-level debt. Open audit findings (CI, package isolation, durability, full-tier query) live here. |
| BACKLOG.md | Non-blocking future work and feature ideas (director-AI scene, replay, mod gallery, and more). |
| TODO/MultiAgent_Orchestration_v2.0.md | Multi-agent orchestration plan. |
| BENCHMARK_LEADERBOARD.md | Public community leaderboard for the Game-Creation Benchmark: ranked results, suite versioning, and the score-submission workflow. |
| CoreAIMods/MOD_SHARING.md | Shareable mod bundle format, Hub export/import flow, import safety, and the proposed community mod gallery process. |
Files under Docs/LocalBusinessPlans/ are intentionally local/untracked (gitignored) and
are not part of the public package documentation surface. Do not link them from
user-facing docs; if a planning conclusion becomes product guidance, copy the stable
technical decision into a public package document instead.
- Keep detailed package docs in English unless the file is explicitly marked
_RU. - Keep changelogs historical; move current user guidance into stable docs.
- Link every new stable guide from the nearest package index and, when it is repository-wide, from this file.
- Keep XML documentation in source concise and contract-focused; avoid mechanical summaries such as "Executes X API operation".