|
|
---
|
|
|
title: PhiRAG
|
|
|
app_file: advanced_rag.py
|
|
|
sdk: gradio
|
|
|
sdk_version: 3.40.0
|
|
|
---
|
|
|
|
|
|
# Advanced RAG System
|
|
|
|
|
|
This repository contains the code for a Gradio web app that demoes a Retrieval-Augmented Generation (RAG) system. This app is designed to allow users to load multiple documents of their choice into a vector database, submit queries, and receive answers generated by a sophisticated RAG system that leverages the latest advancements in natural language processing and information retrieval technologies.
|
|
|
|
|
|
## Features
|
|
|
|
|
|
#### 1. Dynamic Processing
|
|
|
- Users can load multiple source documents of their choice into a vector store in real-time.
|
|
|
- Users can submit queries which are processed in real-time for enhanced retrieval and generation.
|
|
|
|
|
|
#### 2. PDF Integration
|
|
|
- The system allows for the loading of multiple PDF documents into a vector store, enabling the RAG system to retrieve information from a vast corpus.
|
|
|
|
|
|
#### 3. Advanced RAG System
|
|
|
Integrates various components, including:
|
|
|
- **UI**: Allows users to input URLs for documents and then input user queries; displays the LLM response.
|
|
|
- **Document Loader**: Loads documents from URLs.
|
|
|
- **Text Splitter**: Chunks loaded documents.
|
|
|
- **Vector Store**: Embeds text chunks and adds them to a FAISS vector store; embeds user queries.
|
|
|
- **Retrievers**: Uses an ensemble of BM25 and FAISS retrievers, along with a Cohere reranker, to retrieve relevant document chunks based on user queries.
|
|
|
- **Language Model**: Utilizes a Llama 2 large language model for generating responses based on the user query and retrieved context.
|
|
|
|
|
|
#### 4. PDF and Query Error Handling
|
|
|
- Validates PDF URLs and queries to ensure that they are not empty and that they are valid.
|
|
|
- Displays error messages for empty queries or issues with the RAG system.
|
|
|
|
|
|
#### 5. Refresh Mechanism
|
|
|
- Instructs users to refresh the page to clear / reset the RAG system.
|
|
|
|
|
|
## Installation
|
|
|
|
|
|
To run this application, you need to have Python and Gradio installed. Follow these steps:
|
|
|
|
|
|
1. Clone this repository to your local machine.
|
|
|
2. Create and activate a virtual environment of your choice (venv, conda, etc.).
|
|
|
3. Install dependencies from the requirements.txt file by running `pip install -r requirements.txt`.
|
|
|
4. Set up environment variables REPLICATE_API_TOKEN (for a Llama 2 model hosted on replicate.com) and COHERE_API_KEY (for embeddings and reranking service on cohere.com)
|
|
|
4. Start the Gradio app by running `python app.py`.
|
|
|
|
|
|
## Licence
|
|
|
MIT license |
