- Add workerPath config option to specify custom worker-service.cjs location
- Check custom path first, then fall back to Claude plugin cache auto-discovery
- Update docs with simpler manual install instructions (no need to copy to cache)
- Update manifest with workerPath schema and UI hints
Add clawdbot claude-mem CLI commands:
- status: check if worker is responding
- search <query>: search memories with optional --limit
Part of claude-mem integration (Phase 6/7).
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add two memory tools following claude-mem's 3-layer pattern:
- memory_search: Layer 1, compact results (~50-100 tokens)
Returns [#id] title for filtering
- memory_observations: Layer 3, full details (~500-1000 tokens)
Returns narrative with files_modified
Both use @sinclair/typebox for schema definition.
Part of claude-mem integration (Phase 5/7).
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add two hooks:
- after_tool_call: fire-and-forget observation to claude-mem
(skips memory_* tools to prevent recursion)
- before_agent_start: context injection from memory search
(wraps in <claude-mem-context> tags)
Both hooks have error handling and graceful degradation.
Part of claude-mem integration (Phase 4/7).
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add ClaudeMemClient with methods:
- ping(): health check at /api/health
- observe(): fire-and-forget to /api/sessions/observations
- search(): query /api/search with index format
- getObservations(): batch fetch from /api/observations/batch
All methods use AbortSignal.timeout for request timeouts
and handle errors gracefully (worker may be offline).
Part of claude-mem integration (Phase 3/7).
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>