| name | reindex-all-vectors |
| description | Reindex all vector databases for configured projects. Keeps semantic code search fresh. Use when user says "reindex vectors", "refresh code index", or "rebuild vector index". Use when this capability is needed. |
| metadata | {"author":"dmzoneill"} |
Reindex All Vectors - Update Semantic Code Indexes
Iterates config.json repositories, runs code_index per project, restarts watchers.
Inputs
| Input | Type | Default | Purpose |
|---|
force | bool | false | Full re-index (not just changed) |
projects | string | "" | Comma-separated; empty = all |
restart_watchers | bool | true | Restart file watchers after |
Workflow
1. Get Project List
- Load config.json → repositories
- Filter by
projects if provided
- Include only where path exists
- Output: to_index, skipped
2. Reindex Each Project
- For each project:
- Get stats before: chunks_count
code_index(project, force=force) - or _index_project from aa_code_search
- Parse: files_indexed, chunks_created, files_skipped
- Get stats after
- Aggregate: total_files, total_chunks, success_count, error_count
3. Restart Watchers (if restart_watchers)
- For each successfully indexed project:
code_watch(project, action="start") - or start_watcher from aa_code_search
- Track: started, failed
4. Build Summary
- Timestamp
- Projects indexed: N / total
- Total files processed, total chunks
- Per-project table: Project | Status | Files | Chunks | Index Type
- Skipped projects, watcher status, errors
5. Log and Track
memory_session_log("Vector reindex completed", "N projects, X files, Y chunks")
- Update learned/patterns: vector_reindexes (last 50)
6. Output Format
## 🔄 Vector Reindex Complete
**Timestamp:** YYYY-MM-DD HH:MM:SS
### 📊 Summary
- Projects indexed: N / total
- Total files processed: N
- Total chunks created: N
### 📁 Project Details
| Project | Status | Files | Chunks | Index Type |
### 👁️ Watchers Started
- ✅ project1
- ✅ project2
### ⏭️ Skipped Projects
- project: reason
---
Use code_stats() for detailed statistics.
Use code_health() for vector search health.
Key Details
- Chains to:
knowledge_refresh
- Validated by:
gather_context, find_similar_code, explain_code
- Uses aa_code_search: _index_project, _get_index_stats, start_watcher
Converted and distributed by TomeVault — claim your Tome and manage your conversions.