Expand description
Auto-pruning + dedup for the knowledge vector store (#688).
Strategies:
- Stale pruning: remove entries older than a threshold by source_type
- Content dedup: merge entries with near-identical content (prefix match)
- Source dedup: if same source_type+source_id has multiple entries, keep newest
Structs§
- Prune
Report - Prune report returned by the pruning endpoint.
- Prune
Request - Request for the prune endpoint.
Functions§
- prune
- Run the full prune pipeline.