Add docker model code command for AI-powered code editing with aider

[Copilot]

Overview

Adds a new docker model code command that integrates AI-powered code editing into Docker Model Runner by running aider in an ephemeral Docker container. This enables developers to edit code with AI assistance while maintaining a clean, isolated environment that talks directly to Docker Model Runner.

Features

Git-like Workflow

The command behaves similarly to git commit, requiring execution from within a Git repository. When no prompt is provided, it opens your configured text editor (EDITOR/VISUAL, defaulting to vim) to compose a detailed prompt:

# Opens editor for prompt composition
docker model code ai/smollm2

# Or provide prompt directly
docker model code ai/smollm2 "Add error handling to the main function"

Ephemeral Container Execution

Uses docker run --rm to run aider in a disposable container with:

• Repository mounted at /workspace
• Automatic connection to Docker Model Runner
• Network configuration adjusted for Docker Desktop vs Engine
• Clean environment for each execution

Intelligent Environment Detection

Automatically detects and configures the appropriate Model Runner endpoint:

• Docker Desktop: Uses model-runner.docker.internal DNS
• Docker Engine: Uses host networking
• Custom setups: Respects MODEL_RUNNER_HOST environment variable

Usage Examples

# Interactive mode with editor
cd /path/to/your/project
docker model code ai/smollm2

# Direct prompt
docker model code ai/smollm2 "Refactor the database connection code to use a connection pool"

# With OpenAI backend
docker model code --backend openai gpt-4 "Add comprehensive unit tests for the user service"

# Custom aider image
docker model code --aider-image custom/aider:latest ai/smollm2 "Fix bug in authentication"

Implementation Details

The command:

1. Validates the current directory is a Git repository
2. Gets prompt from editor or command line
3. Validates and pulls the specified model if needed
4. Executes aider in a Docker container with proper environment variables and mounts
5. Connects aider to Docker Model Runner via OpenAI-compatible API

Files Changed

• cmd/cli/commands/code.go (276 lines): Main implementation with editor integration, container execution, and URL detection
• cmd/cli/commands/code_test.go (226 lines): Comprehensive unit tests covering command validation, flags, editor integration, and helper functions
• cmd/cli/commands/root.go (1 line): Command registration

Testing

• ✅ All existing tests pass (no regressions)
• ✅ 8 new unit tests added covering all major functionality
• ✅ Manual testing confirms git validation, editor integration, and flag handling work correctly
• ✅ CodeQL security scan: 0 vulnerabilities found

Command Line Interface

Usage:
  model code MODEL [PROMPT] [flags]

Flags:
      --backend string       inference backend to use
      --aider-image string   Docker image to use for aider (default "paulgauthier/aider")
  -h, --help                 help for code

Future Enhancements

This initial implementation provides a solid foundation that could be extended with:

• Configuration file support for project-specific settings
• Support for additional code editing tools
• Advanced networking options for complex setups
• Git commit integration for automatic change tracking

Closes #[issue-number]

Original prompt

Add command "docker model code MODEL [PROMPT]" that behaves like this. You run it in the root of a github repo and it should behave like the github version. It should execute things in an ephemeral container "docker run --rm". Assume the container has aider installed (should be this container by default paulgauthier/aider) so we can execute "aider . -m" and talks to "Docker Model Runner" by default. It should not bring up the normal "docker model run" prompt, instead it should behave like git, open up a vim text editor or other configured text editor to prompt:

github.com/copilot/agents

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.