Files in this browser session
API Integration
POST /process (URL)
POST /process-file (Upload & Process)
POST /search
Delete all vectors for one file
Removes every chunk in Milvus for the given file ID and user ID (one file per request). If the file was ingested with scoping, set Bucket ID and/or Connection ID (same values as at ingest). To drop an entire S3 link for a user, use Admin → purge with optional bucket and/or connection. To wipe all vectors for the user, leave those fields empty there.
Delete vectors for file + user
API: DELETE or POST /delete-document
.pdf name. Use Fill delete form from search on a result row, or paste the id here.
Additional API Endpoints
GET /health
GET /collections
Admin & Scalability Debug
GET /admin/stuck-jobs
GET /admin/queues
Load all vectors (summary)
API: GET /admin/vector-storage-by-user
Scans Milvus using max rows (0 = full scan for the selected filter; paginated) and shows chunk and file counts per user ID. Optional X-Admin-Key if VECTOR_STATS_ADMIN_KEY is set.
List unique files in Milvus
API: GET /unique-files (and /admin/unique-files)
Returns unique files deduplicated by file_id from scanned chunks. Use filters to narrow by tenant/scope.
Optional org_id is echoed back for BYOC correlation (Milvus rows do not store org_id; search still keys vectors by user_id).
Delete all data for a user ID
API: POST /admin/purge-user-vectors
Deletes Milvus vector rows and invalidates the Search3/Tantivy lexical index for a user ID, optionally scoped by collection (pdf/word/…/file_types/image/all),
and optionally narrowed with bucket_id and/or connectionId (AND when both are set; same values as ingest).
Leave bucket and connection empty for a tenant-wide delete within the selected collection. Remaining indexed files are rebuilt from metadata after scoped deletes. For one file, use Delete all vectors for this file above.
Same admin key as Load all vectors.