This document provides a comprehensive overview of the MinutesAI system architecture, including components, data flow, and technical infrastructure.
MinutesAI is designed as a modular, scalable system for processing meeting transcripts and generating structured Minutes of Meeting documents. The architecture follows microservices principles with clear separation of concerns.
The heart of the MinutesAI system responsible for transcript processing and content extraction.
- Purpose: Loads and preprocesses meeting transcripts
- Supported Formats: .txt, .doc, .docx, .mp3, .wav
- Features:
- File format detection and validation
- Text preprocessing and cleaning
- Audio transcription (via Whisper/speech-to-text)
- Encoding detection and normalization
- Purpose: Extracts structured meeting minutes sections using AI
- Technology: LangChain + Ollama integration
- Capabilities:
- Natural Language Processing
- Context-aware section identification
- Entity recognition (attendees, action items, decisions)
- Sentiment analysis and key point extraction
- Purpose: Formats extracted data into various output formats
- Output Formats: Text, JSON, HTML, PDF, Markdown
- Features:
- Template-based formatting
- Custom styling and branding
- Multi-language support
- Export customization
- Technology: Python Click framework
- Features:
- Command-line processing
- Batch operations
- Configuration management
- Progress indicators and verbose logging
- Technology: Streamlit framework
- Features:
- File upload interface
- Real-time processing status
- Interactive result viewing
- Configuration panel
- Export functionality
- Format: YAML-based configuration files
- Scope: Application settings, AI model parameters, output templates
- Features: Environment-specific configs, validation schemas
- Technology: File-based caching system
- Purpose: Store intermediate processing results
- Benefits: Faster reprocessing, recovery from failures
- Location: Configurable output directories