localai not outputting responses in Continue chat.

[drewdrew21b]

Am I missing some basic setting to allow Continue to display responses from my localai service? This is what I have configured:

name: Local Config
version: 1.0.0
schema: v1
models:
  - name: QWEN3 - Coder (25b)
    provider: openai
    model: qwen3-coder-reap-25b-a3b-i1
    apiBase: http://192.168.0.5:18080/v1/
    apiKey: EMPTY
    capabilities:
      - chat
      - tool_use
    requestOptions:
      timeout: 60000

I have confirmed that when I check the console output, if I apply the same Options json output in the API directly, it works just fine.

Additional Details:

• LocalAI: Docker Container v3.8.0
• apiBase: I have tried both internal IP and internal HTTPS endpoint. (Note this is running on a server on my local network, not the same device as my vscode instance.)
• I have tried no defined capabilities and the list above.
• I have tried a few models on my server.

[drewdrew21b]

API Side:

[drewdrew21b]

Looks like this is a sample of the json output in case there is any mismatches on the Continue side:

{
  "created": 1765312222,
  "object": "chat.completion",
  "id": "c05fe112-b0a3-46f1-864b-ba08e32c40a2",
  "model": "qwen3-coder-reap-25b-a3b-i1",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": {
        "role": "assistant",
        "content": "<Response>"
      }
    }
  ],
  "usage": {
    "prompt_tokens": 3,
    "completion_tokens": 1129,
    "total_tokens": 1132
  }
}

[bdougie]

@uinstinct can you assist in debugging this.

[uinstinct]

@drewdrew21b your config seems alright to me. can you do few of the things below and tell me what you get?

• can you check in your qwen3 instance if continue is sending any requests? how does the request object look like?
• if the above does not work, can you try removing the ending / in your apiBase? (i.e. http://192.168.0.5:18080/v1 instead of http://192.168.0.5:18080/v1/)

[drewdrew21b]

When I run this through chat in VSCode I can hear the fans on my server spinning up while it is thinking. It seems to take it when there is a tailing slash or not. When I run the same command directly against my server's API, it outputs without any issues

"can you check in your qwen3 instance if continue is sending any requests? how does the request object look like?"

Is there something I can do here specifically that you were looking for?

[uinstinct]

can you once try changing the provider to ollama from openai? also try removing the apiKey key if it is not required

[drewdrew21b]

Getting errors when having this applied.

Leaving the APIKey field removed, still see the same problems:

In discord, somebody on the team mentioned this might a known issue:

[uinstinct]

Ah, that's awesome!
Then he might have solved that problem earlier. Let me know if you are able resolve the problem with his help, otherwise am here to look deeper.

[drewdrew21b]

Thanks I haven't heard or seen anything back yet in Discord. I will wait a few weeks and come back here if I still haven't.