Bootstrap Knowledge - Build Project Knowledge

Scans project, generates persona-specific knowledge, creates vector index for code search.

Inputs

Workflow

1. Detect Project

• If not provided: detect from cwd via config.json repositories
• Validate project exists in config and path exists

2. Check Known Issues

• check_known_issues("code_index", "") - vector indexing issues

3. Knowledge Scan per Persona

• For each in personas_list: knowledge_scan(project, persona, force=false)
• Personas: developer, devops, tester, release

4. Create Vector Index

• code_index(project) - semantic embeddings for code search
• Parse result: files indexed, chunks created

5. Start File Watcher

• code_watch(project, action="start") - auto-reindex on changes
• Only if vector index created successfully

6. Deep Scan (if enabled)

• code_search(query="class definition pattern implementation", project, limit=10)
• code_search(query="exception handling try except raise error", project, limit=5)
• Read README.md (first 5000 chars)
• Scan tests/ for test_*.py, *.test.js, conftest.py

7. Build Summary

• List generated files: memory/knowledge/personas/<persona>/<project>.yaml
• Vector index stats: files, chunks, watcher status
• Deep scan: README length, test counts, patterns, error handlers

8. Learn from Failures

• If "permission denied": learn_tool_fix("code_index", "permission denied", "Cannot write to vector cache", "Check ~/.cache/aa-workflow/vectors/")
• If "out of memory": learn fix for memory
• If "already running": note watcher is active

9. Log and Track

• memory_session_log("Bootstrapped knowledge for <project>", "Personas: X, Vector: Y")
• Update state/knowledge with project, last_bootstrap, vector_indexed

10. Output Format

## 📚 Knowledge Bootstrap Complete: <project>

**Project Path:** `path`
**Personas:** developer, devops, tester, release

### Generated Knowledge Files
- memory/knowledge/personas/developer/<project>.yaml
- ...

### 🔍 Vector Index Created
- Files indexed: N
- Chunks created: N
- File watcher: ✅ Running

### Next Steps
1. knowledge_query(project='...', persona='developer')
2. code_search(query='...', project='...')
3. knowledge_update() for gotchas

Key Details

• Chains to: knowledge_refresh, learn_architecture
• Provides context for: gather_context, review_pr, start_work
• Project path from config.json repositories