docs: document Claude Code local model workflows - #25641
Conversation
✅ Deploy Preview for docsdocker ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟡 NEEDS ATTENTION
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟢 APPROVE
The PR adds a well-written "Local models" section to workflows.md and updates the cross-reference in claude-code.md. The new content:
- Follows Docker documentation style (sentence-case heading,
consolecode blocks with$prompt, no hedge words, no marketing language) - Correctly discloses the experimental feature status and the
claude-agent-only restriction - Accurately explains the recreate-on-model-change behavior and what persists (workspace, kit-owned volumes)
- Keeps scope appropriate — the new section is concise and consistent with the rest of the page
No confirmed or likely issues were found.
The sandbox docs only described local models through Docker Model Runner. Document the experimental --model workflows for llmman and Ollama, including model swaps, and link them from the Claude Code page.
1057780 to
0db5506
Compare
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟢 APPROVE
The changes document the experimental Claude Code sbx run --model workflows (bundled llmman, Ollama, and model swapping) clearly and accurately. The updated description and keywords match the new content, the replacement Model Runner cross-reference points to the correct sandbox-specific guide, and no AI-isms, broken Markdown, or inaccurate commands were found.
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟢 APPROVE
The PR adds a new Use a local model section documenting the experimental sbx run --model workflows (bundled llmman, Ollama, and model swapping). The content is clear, accurate, and follows Docker documentation style. No AI-isms, no broken markdown, no vendored files touched, and the scope matches the existing document.
No issues requiring changes were found.
| Docker Model Runner, see | ||
| The `--model` flag routes Claude Code's Anthropic API requests to a model | ||
| served on your host. This feature is experimental and isn't supported on | ||
| Windows. |
There was a problem hiding this comment.
Could we change this to only "This feature is experimental."? I don't want to set the expectation that Windows is out because it's not, it's an important platform. Could be a follow on PR.
There was a problem hiding this comment.
Hmm, I heard that we are not shipping llmman + llama in Windows MSI. I took that as Windows is not supported. Can you confirm? cc @ilopezluna
There was a problem hiding this comment.
it's not, i can confirm 👍
There was a problem hiding this comment.
@ilopezluna it's not supported for just this release?
It should at least say windows support is pending, not that we are not supporting it.
There was a problem hiding this comment.
I can confirm its not supported in this release we might say something like "coming soon" that's what we did in DMR. So being clear that is not supported I think its fine and also mentioning that will be here eventually so we set the right expectations
There was a problem hiding this comment.
Local models IMO isn't supported at all, experimental means, have fun, but don't expect any support yet, but approved regardless, we might be too early even documenting this. But it's done now, we should merge and avoid wasted effort.
|
I'm actually making a breaking change, to use ollama, it will become |
Summary
Document the experimental Claude Code
sbx run --modelworkflows introduced by docker/sandboxes#4216, including bundled llmman, Ollama, and model swapping. Retain Docker Model Runner as an alternative on the Claude Code page.@netlify /ai/sandboxes/agents/claude-code/
Preview
Generated by Codex