Persistent Memory Explorer
Semantic search over every incident, consequence and lesson Sentinel has stored. This is the same retrieval path the agent uses during a live incident.
Try a demo query:
Results
Run a search to retrieve memories.
Sentinel embeds the query and ranks stored memories by cosine distance.
Retrieval telemetry
Retrieved from the CockroachDB distributed vector index.
- Query latency
- —
- Memories searched
- —
- Total in store
- 18
- Vector width
- —
- Index
- —
- Embeddings
- —
How ranking works
The query is embedded, then CockroachDB orders memory_embeddings by embedding <=> query — cosine distance. Similarity shown here is 1 − distance. Structured filters narrow the candidate set before ranking.