mcp-server-qdrant 400: "Vector with name fast-bge-small-en-v1.5 is not configured in this collection"
I'm trying to attach mcp-server-qdrant to an MCP client on macOS, against both a local Qdrant instance and Qdrant Cloud. The server starts but a query like {"query": "日本首都是东京"} fails with HTTP 400: "Wrong input: Vector with name fast-bge-small-en-v1.5 is not configured in this collection". What causes this and how can I fix it?
Originally asked on GitHub: https://github.com/qdrant/mcp-server-qdrant/issues/1
4 answers
Status: explained — the maintainer closed the issue as completed without reproducing the error, but the 400 response points to a vector-name mismatch between your Qdrant collection and the embedding model the server expects.
The error means the collection you are querying does not contain a vector named
fast-bge-small-en-v1.5, while the server is sending exactly that name. The maintainer noted that the server expectedfast-bge-small-en-v1.5whereas the default embedding model isfast-all-minilm-l6-v2, and suspected the collection was created externally rather than through the MCP server, which would explain the mismatch. No confirmed fix was posted in the thread.What the thread gives you:
- Check how the collection was created and which vector names it actually contains. If it was created outside the MCP server (for example, in the Qdrant dashboard), its vector name may not match what the server sends. As general advice, the collection's vector name must match the embedding model the server is configured to use, otherwise you get exactly this 400.
- Review the command and environment variables used to start the server — the maintainer asked for these to diagnose the model/vector mismatch.
- If
uvx mcp-server-qdrant ...failed to start earlier, note that the package was published to PyPI on 2024-12-13, so that invocation should now work.
No working MCP JSON configuration was posted in the thread, so the collection/vector setup is the part you need to align yourself.
Did this work for you?What a GitHub user tried on GitHub (Outcome unclear)
Reported in the issue body that both a local Qdrant instance and Qdrant Cloud could not attach the MCP server on macOS, but no error output was included.
Source: https://github.com/qdrant/mcp-server-qdrant/issues/1 (opens external site)
Did this work for you?What @KashiwaByte tried on GitHub (Didn't work)
Sent a query with {"query":"日本首都是东京"} and received HTTP 400: "Vector with name fast-bge-small-en-v1.5 is not configured in this collection".
Source: https://github.com/qdrant/mcp-server-qdrant/issues/1#issuecomment-2520089335 (opens external site)
Did this work for you?What @KashiwaByte tried on GitHub (Outcome unclear)
Reported being unable to use it "in published" with only a screenshot and no error text; what "published" refers to is not stated in the thread.
Source: https://github.com/qdrant/mcp-server-qdrant/issues/1#issuecomment-2520095118 (opens external site)
Did this work for you?
Related questions
- How do I connect mcp-server-qdrant to a local Qdrant database by folder path?
- How to fix timeout when storing Japanese text with mcp-server-qdrant on Windows?
- qdrant-find fails with 'document': what payload structure does the MCP server expect?
- How to use mcp-server-qdrant with existing Qdrant collections when vector name doesn't match?
- Is there a way to edit or delete a record in mcp-server-qdrant?