- Cache now lives in <project>/.pi-project-map/cache/llm-cache.json
- Removed global ~/.cache/pi-project-map/ usage
- Cache travels with the project, no cross-project collisions
- Easy to invalidate: rm -rf .pi-project-map/cache/
Files changed:
- llm-cache.ts: accept cacheDir parameter, default to project-local
- llm-extract.ts: pass cacheDir through to cache functions
- init.ts: pass rootPath as cacheDir
- patch.ts: accept and pass cacheDir
- cli.ts: pass targetPath as cacheDir
- pi-extension.ts: pass ctx.cwd as cacheDir
- Tests updated to use temp dirs for cache isolation
- PiLLMClient now imports @mariozechner/pi-ai's complete() function
- Respects user's /model selection and /login auth
- No external fetch() — Pi handles transport internally
- Added src/types/pi-ai.d.ts for TypeScript declarations
- Extension reverted to real LLM calls (initProject, patchFile, reinitPath)
- Tests mock @mariozechner/pi-ai for verification
Inside Pi, the extension no longer attempts to call the LLM directly
(which Pi's ExtensionAPI doesn't support). Instead:
- project_map_init: generates maps using AST + heuristics (fast, free)
- project_map_patch: rewrites the file's directory with heuristics
- project_map_reinit: full heuristic regeneration
- project_map_validate: unchanged (no LLM needed)
The CLI still supports real LLM calls via --llm-provider=kimi|openai.
This separates concerns:
- Pi extension: deterministic structural analysis
- CLI: rich semantic analysis with configurable LLM
These grammars require tree-sitter@^0.25.0 but we use 0.21.1.
They already failed at runtime with try/catch graceful fallback.
Removing them fixes npm peer dependency conflict during pi install.
- Remove tree-sitter-go and tree-sitter-python from dependencies
- ast-extract.ts already handles missing grammars gracefully
- Heuristics + LLM layer cover Go/Python analysis anyway
- design-doc.md: Rewrote Section 3 (Pipeline Architecture) to describe the
real LLM integration: dual-provider LLM client (Pi native + external),
SHA-256 disk cache in ~/.cache/pi-project-map/, parallelization with
4-8 concurrent requests, retries with exponential backoff, hard-error
policy, context limit protection. Added new Section 4 with concrete
file-level and package-level prompts.
- implementation-plan.md: Replaced old milestone schedule with current
status and detailed M7 tasks for LLM integration: LLM client abstraction,
external API client, Pi LLM client, disk cache, parallel batching with
retries, rewriting llm-extract.ts, context limits, CLI/extension wiring,
and tests.
- Minor formatting cleanup in pi-extension.ts and src/cli.ts
All 16 tests pass. TypeScript compiles clean.
alex