ChatFlex

Instantly turn any dataset into a working chatbot — no code changes needed.

ChatFlex is a lightweight, open-source RAG (Retrieval-Augmented Generation) chatbot that allows users to upload any text or CSV dataset and immediately start asking questions about it. Built for the NYD 2026 Hackathon, it combines vector-based semantic search with LLM-powered response generation for accurate, context-aware answers.

---

What Makes It Different

Most chatbots are hardcoded for one domain. ChatFlex is domain-agnostic — upload a Ramayana dataset, a medical handbook, a product FAQ, or any CSV/text file, and it instantly becomes an expert on that topic. No retraining. No code changes.

---

Features

• Upload Any Dataset — Supports .txt and .csv files up to 10MB+
• Background Processing — Large files are indexed in a background thread; the app stays responsive during upload
• Semantic Search — Vector store retrieves the most relevant context chunks for each query
• LLM-Powered Answers — Integrates with HuggingFace or Gemini API for intelligent response generation
• Flask Web Interface — Simple, clean browser-based UI for upload and Q&A
• Plug-and-Play Dataset Switching — Swap datasets at runtime without restarting the server

---

How It Works

User uploads dataset (CSV / TXT)
        ↓
data_loader.py → splits content into chunks
        ↓
vector_store.py → creates semantic vector index
        ↓
User asks a question
        ↓
vector_store.py → finds top relevant chunks (context)
        ↓
chatbot.py → sends question + context to LLM API
        ↓
Answer displayed in browser

This is a classic RAG (Retrieval-Augmented Generation) pipeline — the same architecture used in enterprise AI assistants.

---

Project Structure

ChatFlex/
├── app.py               # Flask server, routes, background threading
├── chatbot.py           # LLM API integration (HuggingFace / Gemini)
├── data_loader.py       # Text chunking and preprocessing
├── vector_store.py      # Vector index creation and semantic search
├── requirements.txt     # Python dependencies
├── sample_data/         # Example datasets to test with
└── templates/
    └── index.html       # Frontend UI

---

Tech Stack

Layer	Technology
Backend	Python, Flask
AI / LLM	HuggingFace API / Gemini API
Vector Search	Custom vector store (cosine similarity)
Text Processing	Chunking, tokenization
Frontend	HTML, CSS

---

Getting Started

1. Clone the repository

git clone https://github.com/Shivam-Shukla0/ChatFlex.git
cd ChatFlex

2. Install dependencies

pip install -r requirements.txt

3. Add your API key

In chatbot.py, set your HuggingFace or Gemini API key:

API_KEY = "your_api_key_here"

4. Run the app

python app.py

5. Open in browser

http://127.0.0.1:5000

---

Example Use Cases

• Knowledge Base Bot — Upload company docs, get instant Q&A
• Domain Chatbot — Upload Ramayana, legal texts, medical guides
• Study Assistant — Upload lecture notes, ask exam questions
• CSV Data Explorer — Upload structured data, query it in plain English

---

Built For

NYD 2026 Hackathon — Designed to demonstrate rapid AI prototyping with flexible dataset integration.

---

Contributors

• Shivam Shukla — lucifer84670@gmail.com
• Mohit Sharma — msharma42005@gmail.com

---

License

MIT License — free to use, modify, and distribute.