bili-shadowreplay/.cursor/rules/ai-features.mdc

# AI Features and LangChain Integration

## AI Components

- **LangChain Integration**: Uses `@langchain/core`, `@langchain/deepseek`,
  `@langchain/langgraph`, `@langchain/ollama`
- **Whisper Transcription**: Local and online transcription via `whisper-rs` in
  Rust backend
- **AI Agent**: Located in [src/lib/agent/](mdc:src/lib/agent/) directory

## Frontend AI Features

- **AI Page**: [src/page/AI.svelte](mdc:src/page/AI.svelte) - Main AI interface
- **Agent Logic**: [src/lib/agent/](mdc:src/lib/agent/) - AI agent implementation
- **Interface**: [src/lib/interface.ts](mdc:src/lib/interface.ts)
  \- AI communication layer

## Backend AI Features

- **Subtitle Generation**:
  [src-tauri/src/subtitle_generator/](mdc:src-tauri/src/subtitle_generator/) -
  AI-powered subtitle creation
- **Whisper Integration**:
  [src-tauri/src/subtitle_generator.rs](mdc:src-tauri/src/subtitle_generator.rs)
  \- Speech-to-text processing
- **CUDA Support**: Optional CUDA acceleration for Whisper via feature flag

## AI Workflows

- **Live Transcription**: Real-time speech-to-text during live streams
- **Content Summarization**: AI-powered content analysis and summarization
- **Smart Editing**: AI-assisted video editing and clip generation
- **Danmaku Processing**: AI analysis of danmaku (bullet comments) streams

## Configuration

- **LLM Settings**: Configure AI models in [src-tauri/config.example.toml](mdc:src-tauri/config.example.toml)
- **Whisper Models**: Local model configuration for offline transcription
- **API Keys**: External AI service configuration for online features

## Development Notes

- AI features require proper model configuration
- CUDA feature enables GPU acceleration for Whisper
- LangChain integration supports multiple AI providers
- AI agent can work with both local and cloud-based models
  description:
  globs:
  alwaysApply: true

---