API Server URL not working with local model providers

[hinerm]

I tested self-hosted models via LM Studio 0.4.15 and Ollama 0.30.4. Neither supported the /api/ping or /api/threads.. maybe you could shift to Open-AI compatible endpoints? /v1/models, /v1/responses, /v1/completions?

[yzx9]

Hi, thanks for your interest in testing CopilotJ.

I may need to clarify a couple of points to make sure I understand your use case correctly. From your description, it sounds like you are using Ollama to provide a local, self-hosted model with OpenAI-compatible endpoints. However, the endpoints you mentioned, such as /api/ping and /api/threads, are internal APIs used by CopilotJ to serve its own web frontend. They are not intended to be used as OpenAI-compatible model endpoints directly.

If your goal is to use Ollama as the model provider for CopilotJ, please refer to our documentation. In the current codebase, you should be able to use any OpenAI-compatible model by setting COPILOTJ_BASE_URL accordingly.

Hope this helps clarify things, and thanks again for trying out CopilotJ!

[00JackLu]

Hi,@hinerm, for my test, Ollama works fine for now with the OpenAI-compatible endpoint, for example:

COPILOTJ_MODEL=ollama/qwen3.5:4b
COPILOTJ_BASE_URL=http://localhost:11434/

[hinerm]

@yzx9 @00JackLu sorry for not providing the full context!

I was trying to use Ollama via https://copilotj.chat/#/chat

[yzx9]

Hi, copilotj.chat is currently just a static frontend page. You’ll also need to start our server and configure this address to point to it.

Since many of the latest features haven’t been officially released yet, you can clone the repository manually for now, "cd" into the "plugin" directory, and build the JAR yourself. Then install the JAR into Fiji. After opening the plugin, click Install and then Start; once the server is running, you’ll be able to see the URL.