Copilot Chat BYOK (Ollama local) fails with copilotLanguageModelWrapper 404 while Ollama OpenAI endpoint works

[bukowski777]

Select Topic Area

Bug

Copilot Feature Area

VS Code

Body

Summary

Copilot Chat fails with 404 page not found coming from copilotLanguageModelWrapper when using a local Ollama model. The same model works via Ollama’s OpenAI-compatible API (/v1/chat/completions) and also works from VS Code AI Toolkit Playground (“Local via Ollama”).

Environment
• OS: macOS (Apple Silicon)
• VS Code: 1.109.0-insider
• GitHub Copilot Chat extension: 0.37.2026020406 (pre-release)
• GitHub Copilot extension: (version from Extensions view)
• Ollama: 0.15.5-rc2 (installed from DMG, not brew)
• Ollama endpoint: http://127.0.0.1:11434/v1
• Models present in Ollama:
• llama3.2:latest
• qwen2.5-coder:latest
• kimi-k2.5:cloud

Steps to reproduce
1. Run Ollama locally and confirm OpenAI-compatible endpoint is reachable:
• curl http://127.0.0.1:11434/v1/models returns the models list (200 OK).
2. In VS Code Insiders, configure a local model via Ollama (shows as “Local via Ollama”).
3. Select that model in Copilot Chat (Ask mode).
4. Send any message (e.g., “test” / “hello”).

Expected behavior

Copilot Chat should call the local model through the configured Ollama OpenAI-compatible endpoint and return a response.

Actual behavior

Copilot Chat fails immediately with:
• UI: “Sorry, your request failed. Please try again.”
• Logs show 404 page not found from copilotLanguageModelWrapper / _provideLanguageModelResponse. Example:
[error] Server error: 404 404 page not found
[info] ... | notFound | ... | [copilotLanguageModelWrapper]

Evidence the backend is working (Ollama)

The same request works outside Copilot Chat:
• Models list:
• curl -sS http://127.0.0.1:11434/v1/models → 200 OK, returns models.
• Chat completion:
• curl -sS http://127.0.0.1:11434/v1/chat/completions -H "Content-Type: application/json" -d '{"model":"llama3.2:latest","messages":[{"role":"user","content":"say hello"}]}'
→ returns a normal completion JSON (200 OK).

Also, AI Toolkit Playground can successfully chat with “Local via Ollama” using the same model, but Copilot Chat fails with the 404 wrapper error.

Additional notes
• This looks like a routing/compatibility regression in Copilot Chat BYOK / language model access, since Ollama responds correctly to OpenAI-compatible endpoints.
• Happy to provide full logs / screenshots if needed.

[github-actions[bot]]

💬 Your Product Feedback Has Been Submitted 🎉

Thank you for taking the time to share your insights with us! Your feedback is invaluable as we build a better GitHub experience for all our users.

Here's what you can expect moving forward ⏩

• Your input will be carefully reviewed and cataloged by members of our product teams. 
  • Due to the high volume of submissions, we may not always be able to provide individual responses.
  • Rest assured, your feedback will help chart our course for product improvements.
• Other users may engage with your post, sharing their own perspectives or experiences. 
• GitHub staff may reach out for further clarification or insight. 
  • We may 'Answer' your discussion if there is a current solution, workaround, or roadmap/changelog post related to the feedback.

Where to look to see what's shipping 👀

• Read the Changelog for real-time updates on the latest GitHub features, enhancements, and calls for feedback.
• Explore our Product Roadmap, which details upcoming major releases and initiatives.

What you can do in the meantime 💻

• Upvote and comment on other user feedback Discussions that resonate with you.
• Add more information at any point! Useful details include: use cases, relevant labels, desired outcomes, and any accompanying screenshots.

As a member of the GitHub community, your participation is essential. While we can't promise that every suggestion will be implemented, we want to emphasize that your feedback is instrumental in guiding our decisions and priorities.

Thank you once again for your contribution to making GitHub even better! We're grateful for your ongoing support and collaboration in shaping the future of our platform. ⭐

[catalinconstant]

I have the same problem. currently the most reliable way to handle the mismatch is using the OAI Compatible Provider extension that acts as a bridge, ensuring the request headers and paths align perfectly with Ollama's OpenAI-compatible API.

[program-sam]

Great workaround! Thansk! Leaving the api key empty didn't work for me, even tho it is self hosted. but any value in the api key field did the trick. Didn't see this written anywhere, in case it helps someone.

[0APOCALYPSE0]

I have the same problem. currently the most reliable way to handle the mismatch is using the OAI Compatible Provider extension that acts as a bridge, ensuring the request headers and paths align perfectly with Ollama's OpenAI-compatible API.

Can you explain in detail? With OAI Compatible Provider Extension is it working with co-pilot?

[PythonPlumber]

That 404 from the copilotLanguageModelWrapper is a classic case of the extension tripping over its own feet. It’s super clear from your curl tests that Ollama is doing exactly what it should be, so the issue is definitely in the way Copilot Chat is trying to "talk" to it.

One of the weirdest things about these pre-release versions of Copilot Chat is how they handle URLs. Some versions of the extension automatically append /chat/completions to whatever you put in the box. If you put http://127.0.0.1:11434/v1/, the extension might be turning that into http://127.0.0.1:11434/v1//chat/completions. That double slash is an instant 404 for most servers.

Try setting your endpoint to just http://127.0.0.1:11434/v1 without the trailing slash. If that still fails, some people have actually had luck stripping the /v1 off entirely and just using http://127.0.0.1:11434 because the wrapper sometimes injects the /v1 itself.

Local models through Ollama usually only work in "Ask" mode. If you’re trying to use "Agent" mode (the one where it can search your files or run commands), it often breaks because those local models don't always support the specific "tools" and "functions" that the Copilot Agent expects. When the wrapper tries to call a tool-specific endpoint that doesn't exist on your local model, it kicks back that 404. Stick to standard chat and see if the error clears up.

Since you're on the Insiders build, the internal language model provider can sometimes get stuck in a weird state where it thinks it needs to route through GitHub's proxy instead of your local machine. You can force a refresh by signing out of your GitHub account in the VS Code accounts icon at the bottom left. After you sign out, run the command Developer: Reload Window from the Command Palette, and then sign back in. This reset often fixes those "wrapper" errors that feel like they're coming from nowhere.

Make sure the model name in your VS Code settings is a 100% match for what you see when you run ollama list. If Ollama shows llama3.2:latest, make sure you haven't just typed llama3.2. The extension is very picky about that string, and if it asks for a model ID that isn't an exact match, the API will return a 404 because it can't find that specific "file."
Give that URL tweak a shot first since that's usually the culprit in these Insiders regressions. If you open the Output panel in VS Code and select GitHub Copilot Chat from the dropdown, you can actually see the exact URL it's trying to hit. If you see it doubling up on /v1/v1, you'll know exactly which part of the URL to delete!