Multi-Agent Email Assistant
Published:
Overview
An intelligent multi-agent system that automates email writing and management based on natural language instructions. The system uses LangGraph to orchestrate multiple AI agents that monitor, categorize, and respond to emails automatically.
Motivation
This project was created to learn:
- LangGraph & LangChain: For developing AI agent workflows
- LangServe: For simplified API development & deployment using FastAPI
- Gmail API: For building workflow and productivity tools
How It Works
Email Workflow Pipeline
- Email Monitoring
- Constantly checks for new emails using Gmail API
- Triggers processing pipeline for each new message
- Email Categorization
- AI agents sort emails into predefined categories
- Identifies email type (complaint, feedback, question, etc.)
- Response Generation
- Complaints/Feedback: Drafts tailored responses quickly
- Service Questions: Uses RAG (Retrieval Augmented Generation) to pull accurate information from agency documents
- Quality Assurance
- AI quality checks for tone and content
- Formatting validation
- Ensures professional communication standards
- Sending
- Approved emails sent automatically
- Ensures timely communication with clients
Technical Architecture
Multi-Agent System
- Monitor Agent: Watches inbox for new emails
- Classifier Agent: Categorizes email types
- Response Agent: Generates appropriate responses
- QA Agent: Validates quality and formatting
- Sender Agent: Handles email dispatch
Technology Stack
- Framework: LangGraph for agent orchestration
- LLM: Google Gemini API for text generation
- Embeddings: Gemini embeddings for RAG
- API: Gmail API for email operations
- Deployment: FastAPI via LangServe
- Development: Python
Features
- Automatic email monitoring and processing
- Intelligent categorization of email types
- Context-aware response generation
- RAG-based information retrieval from documents
- Quality assurance checks
- Seamless integration with Gmail
Prerequisites
- Python 3.7+
- Groq API key
- Google Gemini API key (for embeddings)
- Gmail API credentials
Project Status
Work in progress - continuously improving agent capabilities and adding new features.
Links
- GitHub Repository
- Topics: Multi-Agent Systems, Gemini API, LangChain, LangServe, LangGraph