Document search API
Semantic retrieval over indexed document collections. Queries are embedded, matched against the vector index and returned as ranked passages with source references.
Streaming endpoint
/ws operational
Long-running retrieval sessions use a persistent WebSocket channel. Access requires a project token issued by your workspace administrator.
# open a retrieval session wss://this-host/ws { "op": "query", "collection": "docs-main", "text": "...", "top_k": 8 }
[gateway] index shards attached: 6/6
[gateway] embedding worker pool ready
Collections
| Name | Documents | Chunks | Reindexed |
|---|---|---|---|
| docs-main | 18 402 | 611 930 | 2 h ago |
| api-reference | 2 117 | 48 265 | 9 h ago |
| support-archive | 54 880 | 1 204 771 | 1 d ago |
Service status
p50 latency38 ms
Uptime 30d99.95%
Index size41.6 GB
Sessions127
This node serves authenticated API traffic only. There is no public sign-up; unauthenticated requests to API paths are rejected.
Request access