README.md

---
title: Paul Graham Essay Bot
emoji: 🐠
colorFrom: pink
colorTo: pink
sdk: docker
pinned: false
---

# 🔍 Open Source RAG with Hugging Face Enpoints

![Open Source](https://img.shields.io/badge/Open%20Source-Models-brightgreen)
![LangChain](https://img.shields.io/badge/Built%20with-LangChain-blue)
![Chainlit](https://img.shields.io/badge/UI-Chainlit-orange)
![HuggingFace](https://img.shields.io/badge/Deployed%20on-HuggingFace-yellow)

## 📝 About

Welcome to this **Paul Graham Essay Bot** - a friendly AI-powered system that demonstrates the power of Retrieval Augmented Generation using completely open source models! This application leverages modern AI technology to provide intelligent answers to your questions based on a collection of essays by Paul Graham, covering topics such as programming languages, startup culture, spam filtering, design principles, and the philosophy of hacking and innovation.

## ✨ Features

- **Open Source Models**: Powered by NousResearch/Meta-Llama-3.1-8B-Instruct for text generation and Snowflake/snowflake-arctic-embed-m for embeddings
- **HuggingFace Integration**: Models deployed as serving endpoints on HuggingFace
- **Intelligent Retrieval**: Utilizes RAG (Retrieval Augmented Generation) for accurate and contextual responses
- **Fast & Responsive**: Async processing for quick responses even with large document collections
- **Content-Focused**: Explore ideas and concepts from the essays, not just information about the author

## 🔧 How It Works

Behind the scenes, this application:

1. **Loads and Processes Documents**: Breaks down essay content into manageable chunks
2. **Creates Embeddings**: Converts text into numerical representations using Snowflake/snowflake-arctic-embed-m
3. **Builds a Vector Database**: Stores the embeddings in a FAISS vector store for efficient retrieval
4. **Retrieves Relevant Content**: Finds the most relevant essay sections based on your questions
5. **Generates Thoughtful Responses**: Uses Meta-Llama-3.1-8B-Instruct to craft helpful answers based on the retrieved content

## 🤔 Example Questions

- "What are some key strategies for starting a successful startup?"
- "Why is Silicon Valley considered a hub for tech innovation?"
- "How can good design improve user experience in technology products?"

## 🛠️ Technical Details

This application uses:
- **LangChain**: For document processing and orchestrating the RAG pipeline
- **FAISS**: For efficient vector similarity search
- **HuggingFace Endpoints**: 
  - NousResearch/Meta-Llama-3.1-8B-Instruct for text generation
  - Snowflake/snowflake-arctic-embed-m for embeddings
- **Chainlit**: For the interactive chat interface
- **Hugging Face Spaces**: For deployment and hosting

Happy exploring the fascinating content with open source AI! 📚✨

### HuggingFace Endpoint Usage
LLM Endpoint
![LLM Endpoint Usage](llm_endpoint_usage.jpeg)

Embedding Endpoint
![Embedding Endpoint Usage](embedding_endpoint_usage.jpeg)