LaptopWiki - RAG Pipeline with Ollama and ChromaDB

This project implements a Retrieval Augmented Generation (RAG) pipeline for answering questions about laptop issues. It leverages the Laptop Wiki community as a knowledge base and uses powerful open-source language models from Ollama for information extraction and question answering.

How it Works

1. Data Extraction and Refinement:

   • Data is extracted from the Laptop Wiki community.
   • The extracted data is rephrased and refined using the llama3:8b LLM from Ollama to ensure high-quality language and consistency.

2. Vector Database Creation:

   • The refined data is split into chunks and embedded using the nomic-embed-text model from Ollama.
   • These embeddings are stored in a ChromaDB vector database for efficient similarity search.

3. Question Answering:

   • User queries are embedded using the same nomic-embed-text model.
   • The ChromaDB database is queried for the most relevant chunks based on similarity to the query embedding.
   • The retrieved chunks, along with the original query, are fed into the llama3:8b LLM with a specific prompt to generate a comprehensive and helpful answer.

Files

• create_database.py: This script handles the entire pipeline for creating the vector database:

  • Downloads the required Ollama models (llama3:8b, nomic-embed-text).
  • Loads data from text files in the sample_data directory.
  • Rephrases the loaded data using the llama3:8b LLM.
  • Splits the data into chunks, embeds them, and stores them in the ChromaDB database at the specified path.

• query_database.py: This script loads the created database and answers user queries:

  • Downloads the required Ollama models.
  • Loads the ChromaDB database from the specified path.
  • Takes a user query as input, retrieves relevant information from the database, and feeds it to the llama3:8b LLM to generate a response.

Instructions for Running

1. Install Dependencies:

   pip install -r requirements.txt

2. Download Ollama Models:

   ollama pull llama3:8b
   ollama pull nomic-embed-text 

3. Prepare your data:

   • Place your Laptop Wiki data in text files within the sample_data directory.

4. Create the Database:

   python create_database.py

5. Run the Question Answering System:

   python query_database.py

   You can then input your laptop-related queries.

Notes:

• Ensure that the paths to your data (DATA_PATH) and database (CHROMA_PATH) are correctly set in the scripts.
• ChromaDB persistency can sometimes be finicky. If you encounter issues, try deleting the existing database directory and recreating it.
• This project is a starting point, and you can further customize it by:
  • Adding more data sources.
  • Fine-tuning the LLMs for your specific use case.
  • Experimenting with different embedding models and prompt engineering for better results.