Personal Knowledge Database Assistant

       

This is a personal knowledge base assistant created using LLMs. The project has been designed and developed meticulously to ensure efficient management and retrieval of vast and intricate information, making it a powerful tool for information acquisition.

This project refers to this tutorial.

Start-up

Hardware Requirements

• CPU: Intel Core i5 or higher
• RAM: 4GB or higher
• OS: Windows, macOS, Linux

Settings

• Clone the repo:

  git clone https://github.com/Zoooooone/LLM_project.git
  cd LLM_project

• Create a Conda environment and install dependencies:

  • Python version 3.9 or higher

    conda create -n llm-project python=3.10.11
    conda activate llm-project
    pip install -r requirements.text

• Run the project:

  ./run.sh

  or

  python -m project.serve.main

Development

Architecture

This project is based on Langchain, a framework designed to simplify the creation of applications using LLMs. This project's architecture mainly consists of these parts:

• LLM Layer: Encapsulates API calls to OpenAI APIs.
• Data Layer: Includes source data of the personal knowledge database and embedding API.
• Database Layer: A vector database built on the source data, used Chroma.
• Application Layer: The top-level encapsulation of the core features. Further, encapsulated the retrieval Q&A chain provided by Langchain.
• Service Layer: Implementation of service access for this project: a demo with Gradio.

More details

Following diagrams illustrates the interaction flow between the user and this assistant.