rag-llama3 0.1.2.dev2

This repository contains a project that implements a Retrieval-Augmented Generation (RAG) system using the LLaMA3 model. The project focuses on creating embeddings for instructions of a professional bioinformatic software to help users conduct biology research.

RAG-LLaMA3 AI Project

We utilize Retrieval Augmented Generation on the LLaMA3 model to create an AI agent that can answer questions about bioinformatics software DNALinux. It helps users navigate through a large range of bioinformatics tools. Additionally, you will be able to create a simple RAG AI agent with your own resources.

You can check out our code at our repo.

Quick Start with Pre-installed AI Agent

1. Install the package:

   pip install rag-llama3
   

   If you are unsure whether you have installed this package, you can use the following command:

   pip show rag_llama3
   

2. Ask a question to our pre-installed DNALinux AI agent:

   rag-llama3 "your question goes here"
   

How to Construct Your Own RAG AI Agent

Make sure that you have already installed the package. There is a Instruction.ipynb you can use to test the code.

1. Configuration:

   • Set up the directory paths for storing your data and Chroma database. You do not need to create them manually; just specify where they should be, and they will be automatically created:
     • input_dir: Directory for PDF, HTML files, and URLs.
     • urls_path: Path to a file named urls.txt where you put all the URLs.
     • output_dir: Directory where TextExtractor will store all .txt files (for debugging purposes).
     • chroma_db_dir: Directory where your Chroma database will be stored.
     • chroma_db_name: Collection name for your Chroma database.
   • Note: The embedding model defaults to 'mxbai-embed-large'. Feel free to choose your preferred Ollama embedding model.

   import RAG as rag
   import TextExtractor as te
   import VectorDB as vdb
   

2. Directory Setup:

   • Open a Jupyter notebook and run the following code to ensure that your directory is created:

     test_vector_db = vdb.VectorDB(input_dir, output_dir, urls_path, chroma_db_dir, chroma_db_name)
     

   • This will create an object that you can use to manipulate your Chroma vector database. It will automatically create all the directories and an empty Chroma database. If everything is already created, it will not overwrite existing files.

3. Add Files:

   • Place all PDF and HTML files in the input directory. List all URLs in the urls.txt file, each on a new line.

4. Load Data:

   • Use the test_vector_db object to load files into the vector database:

     test_vector_db.load_data()
     

   • Alternatively, load different types of files individually:

     test_vector_db.load_url()
     test_vector_db.load_pdf()
     test_vector_db.load_html()
     

   • Loading might take some time. After loading, check if the vector database has been populated successfully:

     test_vector_db.peek()
     test_vector_db.show_sources()
     

   • Query data from a specific source:

     test_vector_db.query_sources(source_name)
     

   • To delete data from a source:

     test_vector_db.delete_source(source_name)
     

   • Or to clear the entire database (be cautious as this is destructive):

     test_vector_db.clear_database()
     

5. Generate Answers:

   • In a Jupyter notebook, use:

     testRAG = rag.RAG(input, output, chroma_db, collection_name)
     print(testRAG.generate_answer("Your question goes here"))