mirror of
https://github.com/hanzoai/chat-rag-api.git
synced 2026-10-30 18:45:41 +00:00
zeekay
2e295af644
Two Qdrant-backend gaps surfaced once rag-api booted on hanzoai/vector: 1. health.py: is_health_ok() is awaited by GET /health, but the Qdrant branch returned a plain bool (pg/mongo branches return coroutines) -> "TypeError: object bool can't be used in 'await' expression", failing the readiness probe. Make is_health_ok async and await the async backends; return the sync Qdrant bool directly. 2. document_routes.py: get_cached_query_embedding read vector_store.embedding_function, which exists on PGVector but NOT on langchain QdrantVectorStore (it exposes .embeddings) -> "QdrantVector object has no attribute embedding_function" on every /query. Use the module-level `embeddings` object directly (backend- agnostic). Proven live: embed(proof-doc-001)->Fireworks 768-dim->Qdrant hanzo_chat, then query->Qdrant points/search 200 OK retrieves it at score 0.7972.