-
Notifications
You must be signed in to change notification settings - Fork 1.2k
Closed
Labels
bugSomething isn't workingSomething isn't working
Description
System Info
llama-stack version: 0.2.21
FAISS provider: SQLite kvstore backend
Embedding models: sentence-transformers/all-MiniLM-L6-v2, Ollama all-minilm
Information
- The official example scripts
- My own modified scripts
🐛 Describe the bug
The /v1/vector-io/insert API endpoint consistently fails with a KeyError on 'document_id'
Commands used:
# Basic vector database creation (works)
curl -X POST http://localhost:8321/v1/vector-dbs \
-H "Content-Type: application/json" \
-d '{
"vector_db_id": "test-faiss-db",
"embedding_model": "sentence-transformers/all-MiniLM-L6-v2",
"provider_id": "faiss"
}'
# Vector-io insert (fails)
curl -X POST http://localhost:8321/v1/vector-io/insert \
-H "Content-Type: application/json" \
-d '{
"vector_db_id": "test-faiss-db",
"chunks": [{
"content": "This is a test document",
"chunk_metadata": {
"document_id": "doc_1"
}
}]
}'
Error logs
INFO VectorIORouter.insert_chunks [StatusCode.OK] (0.26ms)
INFO error: 'document_id'
ERROR Error executing endpoint route='/v1/vector-io/insert' method='post': 'document_id'
Client response:
HTTP/1.1 500 Internal Server Error
{
"detail": "Internal server error: An unexpected error occurred."
}
Expected behavior
The /v1/vector-io/insert endpoint should successfully insert chunks into the vector database when provided with valid schema-compliant requests.
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working