📧 Personal Email RAG Chatbot

A secure, multi-user, and local RAG (Retrieval-Augmented Generation) system for your emails. Chat with your Gmail history using high-performance vector search (pgvector) and local LLMs (Ollama).

🚀 Features

• Strict User Isolation: Multi-user support with encrypted credentials and separate metadata-filtered search.
• Hybrid Indexing: Optimized with B-Tree (relational) and HNSW (vector) indexes for millisecond retrieval.
• Real-time Streaming: Token-by-token response rendering for a snappier user experience.
• Conversation Memory: Remembers the last 6 messages (sliding window) to handle follow-up questions.
• Privacy First: Everything runs locally on your machine using Ollama and PostgreSQL.

🛠 Project Structure

• app.py: Streamlit-based web UI.
• core/: Core application logic.
  • rag.py: RAG Engine with streaming and memory support.
  • database.py: PostgreSQL & pgvector management.
  • sync_emails.py: Email ingestion and vectorization orchestration.
  • config.py: Centralized configuration and path management.
  • utils/: Security and embedding helper utilities.
• ARCHITECTURE.md: Technical system design and component breakdown.
• EVALUATION.md: Performance and accuracy assessment.

📦 Setup & Installation

1. Prerequisites

• Python 3.10+
• Ollama: Download and run from ollama.com.
• PostgreSQL: With pgvector extension installed.

2. Environment Setup

pip install -r requirements.txt

3. Database Initialization

Ensure your PostgreSQL server is running, then provide your connection details in credentials.yaml:

dbname: emails_rag
db_user: postgres
db_password: your_password
db_host: localhost
db_port: 5432

4. Pull Models

ollama pull qwen2.5-coder:3b
ollama pull nomic-embed-text

🎮 How to Run

Start the Chatbot:

streamlit run app.py

🧑‍💻 Technical Assessment Deliverables

• Architecture Document: ARCHITECTURE.md
• Evaluation Report: EVALUATION.md
• Codebase: Fully modularized and documented.
• Bonus: Robust multi-user isolation and HNSW indexing implemented.