Knowledge Bases API

Knowledge bases store documents that ground AI replies. Backed by Qdrant; uploaded files are chunked, embedded, and indexed.

Endpoints

Upload Flow

curl -X POST https://api.anvilhk.com/api/v1/knowledge-bases/kb_01H.../upload \
  -H "Authorization: Bearer $ANVIL_KEY" \
  -H "X-Tenant-ID: $TENANT" \
  -F "file=@product-faq.pdf"

Returns { documentId, chunks }. Embedding is async; poll /stats until tokens increases.

Test Query

curl -X POST https://api.anvilhk.com/api/v1/knowledge-bases/kb_01H.../test \
  -H "Authorization: Bearer $ANVIL_KEY" \
  -H "X-Tenant-ID: $TENANT" \
  -H "Content-Type: application/json" \
  -d '{ "question": "What is your MOQ?" }'

Every response uses the standard envelope:

{
  "success": true,
  "data": { /* resource or array */ },
  "meta": { "total": 100, "cursor": "...", "hasMore": true },
  "error": { "code": "ERROR_CODE", "message": "Human readable" }
}

error is omitted on success; meta is omitted when not paginating.

Rate Limits

Subject to the standard 200 req/min/tenant authenticated limit. See [Rate Limits](/docs/api-reference/rate-limits) for full details and the per-endpoint overrides.