ragify v1.0.10

📌 Ragify

Ragify is a CLI tool that allows you to upload PDF documents and store their embeddings in a vector database (Pinecone or ChromaDB) for Retrieval-Augmented Generation (RAG) applications. It also provides a function to retrieve relevant responses from the vector database and pass them through an LLM to generate answers based on the uploaded document.

🚀 Features

✅ Supports Pinecone and ChromaDB as vector databases.
✅ Splits PDF documents into chunks using LangChain.
✅ Generates embeddings using OpenAI's text-embedding-3-large model.
✅ Stores embeddings in the selected vector database for efficient retrieval.

📦 Installation

Install Ragify using npm:

npm i ragify

🛠️ What This Library Provides

This package provides two key functions:

• uploadFile(filePath): Uploads a PDF file, generates embeddings, and stores them in the selected vector database.
• askQuestion(query): Retrieves relevant information from the stored embeddings and uses an LLM to generate a response.

Currently, Pinecone and ChromaDB are the supported vector databases.

🌎 Environment Variables

Before using the library, set up your .env file with the required credentials.

For Pinecone

DB_TYPE=pinecone
PINECONE_API_KEY=your_pinecone_api_key
PINECONE_INDEX_NAME=your_index_name
PINECONE_ENV=your_pinecone_environment
OPENAI_API_KEY=your_open_ai_api_key
OPENAI_MODEL=your_desired_model by default the 'gpt-4' model will be used
OPENAI_EMBEDDING_MODEL=your_desired_model by default the text-embedding-3-large will be used

For ChromaDB

DB_TYPE=chroma
CHROMA_DB_URL=http://localhost:8000
COLLECTION_NAME=pdf_embeddings
OPENAI_API_KEY=your_open_ai_api_key
OPENAI_MODEL=your_desired_model by default the 'gpt-4' model will be used

🚀 Usage

To run the CLI tool:

node cli.js

Follow the prompts to select a database and provide the necessary details.

Alternatively, you can use the functions in your Node.js project:

import { uploadFile, askQuestion } from "ragify";

// Upload a PDF file
await uploadFile("./documents/example.pdf");

// Ask a question based on the document
const response = await askQuestion("What is the summary of the document?");
console.log(response);

📝 How It Works

1️⃣ User selects a vector database (Pinecone/ChromaDB).
2️⃣ User provides the necessary database details.
3️⃣ PDF file is loaded and split into chunks using LangChain.
4️⃣ Embeddings are generated using the OpenAI API.
5️⃣ Embeddings are stored in the selected vector database.
6️⃣ When a query is made, relevant embeddings are retrieved and passed through an LLM to generate a response.

🔍 Debugging Tips

If embeddings are not being stored correctly in Pinecone:

1️⃣ Check API Key

curl -X GET "https://api.pinecone.io/v1/whoami" -H "Api-Key: ${PINECONE_API_KEY}"

2️⃣ Check if Pinecone index exists

curl -X GET "https://controller.${PINECONE_ENV}.pinecone.io/databases" -H "Api-Key: ${PINECONE_API_KEY}"

3️⃣ Print Loaded Document Chunks

Modify uploadFile() to inspect document chunks:

console.log(allSplits[0]);

🤝 Contributing

Contributions are welcome! Feel free to submit issues and pull requests to improve this library.

📜 License

This project is licensed under the MIT License.