no idea how to get mcp-qdrant to work

[f1-outsourcing]

testing with goose for windows and created the extension with socat forwarding because it looks like this https://github.com/qdrant/mcp-server-qdrant/ env vars are not being used.

http://192.168.x.x:8001/mcp

INFO:     127.0.0.1:46810 - "POST /mcp HTTP/1.1" 307 Temporary Redirect
INFO:     127.0.0.1:46810 - "POST /mcp/ HTTP/1.1" 200 OK
[04/25/26 21:40:46] DEBUG    Call tool: qdrant-find with {'query': 'hello'}                                                                  server.py:419
                    ERROR    Error calling tool 'qdrant-find': Unexpected Response: 400 (Bad Request)                                  tool_manager.py:139
                             Raw response content:
                             b'{"status":{"error":"Wrong input: Not existing vector name error:
                             fast-all-minilm-l6-v2"},"time":0.009269893}'
                             ╭────────────────────────────────── Traceback (most recent call last) ──────────────────────────────────╮
                             │ /root/.cache/uv/archive-v0/uixYZFRm4BHd60SmiR1jv/lib/python3.11/site-packages/fastmcp/tools/tool_mana │
                             │ ger.py:130 in call_tool                                                                               │
                             │                                                                                                       │
                             │   127 │   │   │   raise NotFoundError(f"Unknown tool: {key}")                                         │
                             │   128 │   │                                                                                           │
                             │   129 │   │   try:                                                                                    │
                             │ ❱ 130 │   │   │   return await tool.run(arguments)                                                    │
                             │   131 │   │                                                                                           │
                             │   132 │   │   # raise ToolErrors as-is                                                                │
                             │   133 │   │   except ToolError as e:                                                                  │
                             │                                                                                                       │
                             │ /root/.cache/uv/archive-v0/uixYZFRm4BHd60SmiR1jv/lib/python3.11/site-packages/fastmcp/tools/tool.py:2 │
                             │ 20 in run                                                                                             │
                             │                                                                                                       │
                             │   217 │   │   type_adapter = get_cached_typeadapter(self.fn)                                          │
                             │   218 │   │   result = type_adapter.validate_python(arguments)                                        │
                             │   219 │   │   if inspect.isawaitable(result):                                                         │
                             │ ❱ 220 │   │   │   result = await result                                                               │
                             │   221 │   │                                                                                           │
                             │   222 │   │   return _convert_to_content(result, serializer=self.serializer)                          │
                             │   223                                                                                                 │
                             │                                                                                                       │
                             │ /root/.cache/uv/archive-v0/uixYZFRm4BHd60SmiR1jv/lib/python3.11/site-packages/mcp_server_qdrant/mcp_s │
                             │ erver.py:152 in find                                                                                  │
                             │                                                                                                       │
                             │   149 │   │   │                                                                                       │
                             │   150 │   │   │   await ctx.debug(f"Finding results for query {query}")                               │
                             │   151 │   │   │                                                                                       │
                             │ ❱ 152 │   │   │   entries = await self.qdrant_connector.search(                                       │
                             │   153 │   │   │   │   query,                                                                          │
                             │   154 │   │   │   │   collection_name=collection_name,                                                │
                             │   155 │   │   │   │   limit=self.qdrant_settings.search_limit,                                        │
                             │                                                                                                       │
                             │ /root/.cache/uv/archive-v0/uixYZFRm4BHd60SmiR1jv/lib/python3.11/site-packages/mcp_server_qdrant/qdran │
                             │ t.py:124 in search                                                                                    │
                             │                                                                                                       │
                             │   121 │   │   vector_name = self._embedding_provider.get_vector_name()                                │
                             │   122 │   │                                                                                           │
                             │   123 │   │   # Search in Qdrant                                                                      │
                             │ ❱ 124 │   │   search_results = await self._client.query_points(                                       │
                             │   125 │   │   │   collection_name=collection_name,                                                    │
                             │   126 │   │   │   query=query_vector,                                                                 │
                             │   127 │   │   │   using=vector_name,                                                                  │
                             │                                                                                                       │
                             │ /root/.cache/uv/archive-v0/uixYZFRm4BHd60SmiR1jv/lib/python3.11/site-packages/qdrant_client/async_qdr │

INFO:     127.0.0.1:59930 - "POST /mcp HTTP/1.1" 307 Temporary Redirect
INFO:     127.0.0.1:59930 - "POST /mcp/ HTTP/1.1" 200 OK
WARNING:root:Failed to validate request: 30 validation errors for ClientRequest
PingRequest.method
  Input should be 'ping' [type=literal_error, input_value='resources/read', input_type=str]
    For further information visit https://errors.pydantic.dev/2.11/v/literal_error
InitializeRequest.method
  Input should be 'initialize' [type=literal_error, input_value='resources/read', input_type=str]
    For further information visit https://errors.pydantic.dev/2.11/v/literal_error
InitializeRequest.params.protocolVersion
  Field required [type=missing, input_value={'_meta': {'agent-session...ssToken': 2}, 'uri': ''}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/missing
InitializeRequest.params.capabilities
  Field required [type=missing, input_value={'_meta': {'agent-session...ssToken': 2}, 'uri': ''}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/missing
InitializeRequest.params.clientInfo
  Field required [type=missing, input_value={'_meta': {'agent-session...ssToken': 2}, 'uri': ''}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/missing
CompleteRequest.method
  Input should be 'completion/complete' [type=literal_error, input_value='resources/read', input_type=str]
    For further information visit https://errors.pydantic.dev/2.11/v/literal_error
CompleteRequest.params.ref
  Field required [type=missing, input_value={'_meta': {'agent-session...ssToken': 2}, 'uri': ''}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/missing
CompleteRequest.params.argument
  Field required [type=missing, input_value={'_meta': {'agent-session...ssToken': 2}, 'uri': ''}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/missing
SetLevelRequest.method
  Input should be 'logging/setLevel' [type=literal_error, input_value='resources/read', input_type=str]
    For further information visit https://errors.pydantic.dev/2.11/v/literal_error
SetLevelRequest.params.level
  Field required [type=missing, input_value={'_meta': {'agent-session...ssToken': 2}, 'uri': ''}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/missing
GetPromptRequest.method
  Input should be 'prompts/get' [type=literal_error, input_value='resources/read', input_type=str]
    For further information visit https

{"name": "extensionmanager__list_resources", "parameters": {"extension_name": "mcp-qdrant"}}

0:28
Edit

Copy

Read Resource uri: (ui://mcp-qdrant_find)

Tool Details

uri
(ui://mcp-qdrant_find)

Output
Extension operation failed: Failed to read resource: Resource with uri '(ui://mcp-qdrant_find)' not found. Here are the available extensions: computercontroller, mcp-qdrant_testing_, mcp-qdrant, tom, apps, skills, extensionmanager, developer, summon, todo, analyze
0:28
It seems like the extensionmanager__read_resource function had an error reading the resource again. The error message indicates that it could not find a resource with the specified URI ('(ui://mcp-qdrant_find)').

Unfortunately, I couldn't find any information on how to proceed using the provided functions and arguments.

To continue our conversation, I'll need to figure out what caused this issue and how to resolve it. Can you please provide more information about your setup and data?

One possible approach would be to use the extensionmanager__list_resources function again with a valid URI from the QDRANT extension:

{"name": "extensionmanager__list_resources", "parameters": {"extension_name": "mcp-qdrant"}}

0:28

Copy
{"name": "extensionmanager__list_resources", "parameters": {"extension_name": "mcp-qdrant"}}

0:30
Edit

Copy

List Resources extension_name: mcp-qdrant

Tool Details

extension_name
mcp-qdrant

Output
apps - clock, uri: (ui://apps/clock)

Output

Output

Output
0:30
It seems like the extensionmanager__list_resources function has returned a list of resources from the mcp-qdrant extension. The output includes an app named "clock" with its corresponding URI.

If you want to get more information about this resource, such as its description or usage instructions, you can use the extensionmanager__read_resource function with its proper arguments:

{"name": "extensionmanager__read_resource", "parameters": {"uri": "(ui://apps/clock)"}}

0:30

[cookesan]

The log suggests transport/env propagation is no longer the first problem. The request reaches the qdrant-find tool, and Qdrant itself returns the 400.

mcp-server-qdrant derives the vector name from the embedding provider. With the default FastEmbed model, that name is fast-all-minilm-l6-v2; both store and search use that named vector. The server only creates the collection if it does not already exist, so an existing Qdrant collection that was created with an unnamed vector, a different vector name, or a different embedding model will produce exactly this kind of "Not existing vector name" error.

The quickest test is to point the extension at a brand-new collection name and try store/find again. If that works, the connection settings are fine and the existing collection schema is the mismatch. The alternatives are to recreate the collection, or configure the embedding model/provider to match the vector name already present in that collection.

The /mcp to /mcp/ redirect in the log is probably incidental, since the redirected request returns 200 and the tool call is executed.

[f1-outsourcing]

Yes was indeed some mis configuration and deleting the existing collection helped. I have ollama / qdrant / mcp-qdrant on the linux server and goose on windows. Should such setup allow for documents to be parsed? I have the impression only meta data is being send (path of the file)