Feature Request: Non-Interactive CLI Model Selection for Claude Code Integration

[johnib]

Problem

Currently, when using --claude-code flag, users must manually select models through interactive prompts every time they start the server. This creates friction for:

• CI/CD pipelines and automation scripts
• Development workflows where the same models are used repeatedly
• Quick restarts during development sessions

Current Behavior

npx copilot-api start --claude-code
# Always prompts for model selection interactively

Proposed Solution

Add --model and --small-model flags to allow non-interactive model specification:

# Specify both models directly
npx copilot-api start --claude-code --model "claude-3.5-sonnet" --small-model "claude-3.5-haiku"

# Short aliases also work
npx copilot-api start -c -m "claude-3.5-sonnet" -s "claude-3.5-haiku"

Benefits

• Enables automation and scripting
• Faster development iteration
• Consistent model selection across environments
• Maintains backward compatibility with interactive mode

Implementation Details

• Both flags must be specified together for validation
• Models are validated against available Copilot models
• Falls back to interactive mode when flags are not provided
• Comprehensive error handling with helpful messages