Tips for Gemini CLI Users

This guide provides helpful tips for users who are migrating from or using both Gemini CLI and LLxprt Code.

Syncing Configurations with Symbolic Links

If you're using both Gemini CLI and LLxprt Code, you can keep your configurations in sync using symbolic links. This prevents the need for manual copying or rsync.

Setup Instructions

# Link settings file
ln -s ~/.gemini/settings.json ~/.llxprt/settings.json

# Link context file (GEMINI.md/LLXPRT.md)
ln -s ~/.gemini/GEMINI.md ~/.llxprt/LLXPRT.md

# Link environment variables
ln -s ~/.gemini/.env ~/.llxprt/.env

Platform-Specific Examples

Termux (Android)

ln -s /data/data/com.termux/files/home/.gemini/settings.json /data/data/com.termux/files/home/.llxprt/settings.json
ln -s /data/data/com.termux/files/home/.gemini/GEMINI.md /data/data/com.termux/files/home/.llxprt/LLXPRT.md
ln -s /data/data/com.termux/files/home/.gemini/.env /data/data/com.termux/files/home/.llxprt/.env

macOS/Linux

ln -s ~/.gemini/settings.json ~/.llxprt/settings.json
ln -s ~/.gemini/GEMINI.md ~/.llxprt/LLXPRT.md
ln -s ~/.gemini/.env ~/.llxprt/.env

Important Compatibility Notes

While the symlink approach works for many use cases, please be aware that LLxprt Code and Gemini CLI configurations are not 100% compatible due to the following differences:

Key Differences

  1. Authentication

    • LLxprt Code supports multiple auth providers (OAuth, API keys for different providers)
    • Auth configuration structure differs from Gemini CLI
    • Use /auth command to set up authentication in LLxprt Code

  2. Model Configuration

    • LLxprt Code uses a profile-based system instead of a single default model
    • Multiple models can be configured and switched between easily
    • Use /profile save and /profile set-default to manage model profiles

  3. Provider Support

    • LLxprt Code supports multiple AI providers (OpenAI, Anthropic, Google)
    • Provider-specific settings may not transfer from Gemini CLI
    • Each provider has its own configuration options

  4. Settings Structure

    • Some settings have different names or structures
    • New settings in LLxprt Code that don't exist in Gemini CLI
    • Provider-specific tool format overrides

Testing Your Setup

After creating the symbolic links, test the configuration:

  1. Start LLxprt Code and check if your settings loaded correctly
  2. Test authentication with /auth status
  3. Check your model profiles with /profile list
  4. If you encounter issues, you may need to:
    • Set up authentication again using /auth
    • Create new model profiles using /profile save

When to Use Separate Configurations

Consider maintaining separate configurations if you:

  • Use different models or providers in each tool
  • Need different authentication methods
  • Have customized settings specific to each tool
  • Experience conflicts with the shared configuration

Understanding Authentication in LLxprt Code

Mixed Authentication Strategy

LLxprt Code offers a unique capability that allows you to use Google authentication for Google-specific features (web search, web fetch) while using API keys for other providers. This gives you the best of both worlds:

  • Google Authentication (/auth): Provides access to Google Search grounding and other Google-specific features
  • Provider API Keys: Use different AI providers (OpenAI, Anthropic, etc.) while maintaining Google features

Setting API Keys

LLxprt Code provides two commands for managing API keys:

/key Command

Sets an API key directly in the current session:

/key sk-your-api-key-here
  • The key is stored in memory for the current session only
  • Useful for quick testing or temporary access
  • Does not persist between sessions

/keyfile Command

Loads an API key from a file:

/keyfile ~/.keys/openai.txt
  • More secure than typing keys directly
  • Keeps sensitive keys out of command history
  • Allows different key files for different providers

Example: Using Google Auth with OpenAI

  1. First, authenticate with Google for web features:

    /auth

  2. Switch to OpenAI as your AI provider:

    /provider openai

  3. Set your OpenAI API key:

    /keyfile ~/.keys/openai.txt

Now you can use OpenAI models while still having access to Google's web search and web fetch capabilities!

Migration Tips

If you're migrating from Gemini CLI to LLxprt Code:

  1. Start Fresh (Recommended)

    • Run LLxprt Code without symlinks first
    • Set up authentication: /auth
    • Configure your preferred model: /model <model-name>
    • Save as default profile: /profile save default and /profile set-default default

  2. Gradual Migration

    • Start with symlinking only .env for API keys
    • Manually configure LLxprt Code-specific settings
    • Test thoroughly before symlinking other files

  3. Check MCP Servers

    • MCP server configurations should work if paths are correct
    • You may need to update paths if they're absolute and platform-specific

Troubleshooting

Common Issues

  1. Authentication Failures

    • Remove the symlinked settings.json
    • Run /auth to set up authentication properly for LLxprt Code
    • Save your configuration

  2. Model Not Found

    • LLxprt Code may not recognize Gemini CLI's default model setting
    • Use /model to see available models
    • Set your preferred model and save a profile

  3. MCP Server Errors

    • Check that MCP server paths exist and are executable
    • Update paths in settings if needed
    • Some MCP servers may need to be reinstalled for LLxprt Code

Getting Help