ConvoRAG

Conversation intelligence chatbot powered by local-first RAG. Extract topics, detect personality, and chat with conversation history — no external APIs.

Live Demo: https://web-production-fbb03.up.railway.app/ Video

kostack.demo.mp4

Features

• Topic Detection — Automatic semantic topic segmentation using TF-IDF
• Persona Extraction — Habits, personality traits, communication style from message patterns
• Dual-Layer Retrieval — Topic context + individual message evidence
• No External APIs — Runs entirely locally with scikit-learn

Quick Start

# 1. Setup
python -m venv venv
venv\Scripts\activate  # Windows
pip install -r requirements.txt

# 2. Run
python app.py

# 3. Open browser
http://localhost:5000

Upload a CSV with columns: timestamp, sender, message (names flexible).

CSV Format

timestamp,sender,message
2024-01-15 08:05:00,Alice,Good morning!
2024-01-15 09:30:00,Bob,Hey Alice!

Column names auto-detected: date/time, from/author, text/content/body, etc.

Tech Stack

• Backend: Flask + Python
• NLP: scikit-learn (TF-IDF, cosine similarity)
• Frontend: Vanilla HTML/CSS/JS
• Deployment: Gunicorn on Railway/Render

Project Structure

src/
├── processor.py   # CSV loading, topic detection, checkpoints
├── persona.py     # Persona extraction from message patterns
├── rag.py         # Retrieval and answer synthesis
app.py             # Flask API
templates/
└── index.html     # Chat UI

Deployment

Deployed on Railway. Auto-deploys on GitHub push.

To self-host:

gunicorn app:app --bind 0.0.0.0:$PORT --timeout 120

├── Procfile # Heroku/Railway └── render.yaml # Render.com

---

## API Reference

| Method | Endpoint | Description |
|--------|----------|-------------|
| `POST` | `/api/process` | Upload CSV (multipart) or `{"csv":"path"}` |
| `POST` | `/api/query` | `{"query":"..."}` → RAG answer |
| `GET` | `/api/topics` | All topic checkpoints |
| `GET` | `/api/checkpoints` | All 100-msg checkpoints |
| `GET` | `/api/persona` | Persona JSON |
| `GET` | `/api/status` | Processing status |