- project_map_reinit now regenerates only the target subtree + ancestors
by default, falling back to full reinit when subtree file count exceeds
reinitFullThresholdPercent (default 10%).
- Add reinitFullThresholdPercent config option.
- Expose fix=true on project_map_validate Pi tool for localized repair.
- Update docs and runtime guidance to prefer patch / validate --fix
before full reinit.
- Add integration tests for smart reinit and update typebox mock.
The before_agent_start handler now scans the active session context via
ctx.sessionManager.buildSessionContext() for an existing pi-project-map-hint
custom message and skips injection when one is already present in the
current branch. This prevents duplicate visible hints in advisory/pre-init
modes and duplicate hidden hints in strong/strict modes. The hint is
automatically re-injected after compaction or /tree navigation removes it
from the active path.
Also removes the unused hooks/on-prompt.ts prompt-text injector.
The before_agent_start handler now inspects the current session context
and skips injection if a pi-project-map-hint custom message is already
present in the active branch. This prevents duplicate hints on every
prompt while still re-injecting after compaction or tree navigation.
Also removes the unused hooks/on-prompt.ts prompt-text injector.
- Call ctx.modelRegistry.getApiKeyAndHeaders(model) to resolve auth
- Pass resolved apiKey and headers to complete() options
- Provides clear error message if auth fails ( directs user to /login )
- Update test mock to include ok: true on auth response
- No external fetch() — Pi handles all transport and auth
- 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
Developer
alex