Add Google Gemini AI Integration

Overview

Added support for Google's Gemini AI models, enabling users to interact with Gemini's language models through the application's existing LLM interface.

Features

• Integration with Google's Generative AI SDK
• Support for multiple Gemini models:
  • gemini-1.5-flash
  • gemini-1.5-pro
  • gemini-1.5-flash-8b
  • gemini-1.0-pro
• Streaming response support for real-time interactions
• Proper handling of system messages by combining them with the first user message
• Configurable max output tokens
• Error handling with user-friendly messages

Implementation Details

• Added new Gemini configuration options in the config system
• Implemented message streaming using async iterators
• Added proper message history handling with role conversion (assistant → model)
• Integrated with the existing abort mechanism for message cancellation
• Added API key validation and error handling

Configuration

Users can configure Gemini through the following settings:

• API Key: Google API key for authentication
• Model Selection: Choice between available Gemini models
• Max Tokens: Configurable through the default maxTokens setting

Technical Notes

• Messages are properly formatted to match Gemini's expected structure
• System messages are merged with the first user message due to Gemini's message format requirements
• Streaming implementation ensures real-time response delivery
• Error handling includes specific handling of API key validation errors

Testing Required

• Verify streaming functionality works correctly
• Test system message handling
• Confirm error scenarios are properly handled
• Validate max token limits are respected
• Check abort functionality works as expected

Breaking Changes

None. This is an additive feature that maintains compatibility with existing LLM integrations.