ci(docker): add DeepSeek-V3 kernels and modernize tooling

[fishcrap]

Description

Enhance Docker image with DeepSeek-V3 optimization packages to enable MoE training with optimal performance on Hopper GPUs. Also modernize development tooling and resolve dependency conflicts.

Key additions:

• DeepSeek-V3 kernels: FlashMLA (Multi-head Latent Attention), DeepGEMM (FP8 GEMM), DeepEP (Expert Parallelism), flash-linear-attention
• Modern Node.js setup: Replace nvm with fnm for faster installation
• AI coding tools: opencode, @openai/codex, @google/gemini-cli
• Dependency fixes: Resolve uvicorn version conflicts via litellm[proxy]

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Code refactoring (no functional changes)
• Performance improvement
• Test coverage improvement

Checklist

• I have read the Contributing Guide
• I have run formatting tools (pre-commit or manual)
• I have run relevant unit tests and they pass (requires Docker build)
• I have added tests for new functionality (validation script updated)
• I have updated documentation if needed
• My branch is up to date with main
• This PR introduces breaking changes (if yes, fill out details below)
• If this PR changes documentation, I have built and previewed it locally with jb build docs
• No critical issues raised by AI reviewers (/gemini review)

Breaking Change Details (if applicable):

This is a backward-compatible addition. Existing usage remains unchanged.

Changes

Dockerfile

Change	Description
DeepSeek-V3 packages	Added FlashMLA, DeepGEMM, DeepEP, flash-linear-attention for MoE optimization
Node.js tooling	Replaced nvm with fnm; moved to STAGE 2.5 after CUDA compilation
AI coding tools	Added opencode, @openai/codex, @google/gemini-cli

Python Dependencies (pyproject.toml)

Change	Before	After
PyTorch optimization	-	torchao>=0.15.0
Claude SDK	-	claude-agent-sdk
LiteLLM	litellm + litellm-proxy-extras	litellm[proxy]>=1.81.3
uvicorn	>=0.34.2 (strict)	(unconstrained)
override-dependencies	-	openai, soundfile

Documentation & Validation

• Added DeepSeek-V3 packages section to docs/tutorial/installation.md
• Added validation tests for FlashMLA, DeepGEMM, DeepEP, fla, NVSHMEM in validate_docker_installation.py

GPU Requirements

Package	Minimum GPU	Notes
FlashMLA	SM90+ (Hopper)	Multi-head Latent Attention kernels
DeepGEMM	SM90+ (Hopper)	FP8 GEMM library for MoE
DeepEP	SM80+ (Ampere)	Expert Parallelism communication
flash-linear-attention	Any GPU	Pure Triton kernels
NVSHMEM	Any GPU	Required for DeepEP internode

Additional Context

• The uvicorn version conflict was caused by litellm-proxy-extras requiring a different version than the main dependencies. Using litellm[proxy] bundles the proxy extras and manages the version constraints internally.
• fnm is significantly faster than nvm for Node.js installation and better suited for Docker builds.

[gemini-code-assist]

Summary of Changes

Hello @fishcrap, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses a dependency conflict involving litellm[proxy] and uvicorn by adjusting the pyproject.toml to specify the correct litellm variant and relaxing the uvicorn version constraint. Additionally, the Docker build process has been enhanced to leverage uv for efficient and reliable dependency synchronization, ensuring the project builds consistently.

Highlights

• Dependency Resolution: The pyproject.toml file has been updated to resolve a dependency conflict between litellm[proxy] and uvicorn. Specifically, litellm and litellm-proxy-extras were replaced with litellm[proxy], and the uvicorn version constraint was relaxed from >=0.34.2 to uvicorn.
• Dockerfile Dependency Management: The Dockerfile now incorporates a uv sync command to manage project dependencies, ensuring they are installed using the lockfile and specific project settings. This includes mounting necessary files like uv.lock, pyproject.toml, and areal/__init__.py during the build process, aligning with best practices for uv in Docker.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request addresses a conflict with litellm[proxy] and uvicore>=34.2 by adjusting dependency specifications in pyproject.toml and updating the Dockerfile to use uv sync for dependency installation. The changes correctly consolidate litellm dependencies and remove the problematic uvicorn version constraint, which is a good step towards resolving the reported conflict. However, there are some redundant instructions in the Dockerfile that should be addressed for better maintainability.