RAGServe

A simple chatbot API for your website using RAG and langchain.

Features

LLM-Powered Conversations: Uses advanced models like Llama-3.2 and BAAI's embedding model for intelligent and contextual responses. Vector-Based Contextual Search: Enhances relevance using FAISS to search similar contexts in preprocessed documents. Customizable Personality: Embodies the personality of Gargi, a wise and knowledgeable assistant with a touch of humor. Modular Design: Easy-to-understand codebase with clear separation of concerns.

Prerequisites

Node.js (v16 or later) TypeScript Hugging Face API key (for accessing models) FAISS for vector-based similarity search

Usage

1. Clone the repository
2. Add your additional information in data/doc.txt arranged in different sections so that chunking does not effect RAG performance much
3. Create a .env file with HF_API_TOKEN inside
4. Run npm install
5. Run npm run build
6. Run npm start to start the node.js server

Customization

• Change the Assistant's Personality: Modify the systemPrompt in assistant/config.ts to customize Gargi's character.
• Switch Models: Update the model field in embeddingConfig and generationConfig in assistant/config.ts to use different Hugging Face models.
• Add More Data: Add additional documents to the data/ folder and update doc.txt as needed.

How It Works

1. Document Processing: The application reads doc.txt and preprocesses it into smaller chunks for efficient similarity search using FAISS.
2. User Query Handling: When a query is sent to /assistant/ask, the server: Searches the vector store for relevant context. Combines the query and context into a prompt. Sends the prompt to Hugging Face's language model for response generation.
3. Error Handling: If any issue occurs, the server returns a witty and friendly error message.