VectorDB-Plugin-for-LM-Studio - Boost Precision by Implementing Retrieval Augmented Generation

Introduction to VectorDB Plugin for LM Studio

The VectorDB Plugin for LM Studio is an innovative tool that boosts the capabilities of large language models (LLMs) by reducing inaccuracies through a technique called retrieval augmented generation (RAG). It allows users to create and search through a vector database that provides context from various documents, resulting in more precise responses from LLMs.

Key Features

Text Extraction: Supports a wide range of documents, including .pdf, .docx, .epub, and more.
Image Summarization: Uses "vision" models to create summaries of image files like .png, .jpg, and others.
Audio Transcription: Converts audio files such as .mp3, .wav, into text for easier searching.
Voice Queries: Allows users to type or speak their questions using a powerful voice recorder.
LLM Responses: Utilizes platforms like LM Studio and other local models for generating responses, with Chat GPT integration coming soon.
Text-to-Speech: Offers options for speech playback of LLM responses through various services like Bark and Google TTS.
Hardware Support: Compatible with CPU and Nvidia GPUs, with plans to expand to AMD and Intel GPUs.

System Requirements

To get started with the VectorDB Plugin, ensure the following prerequisites are met:

Python 3.11
Git and Git LFS
Pandoc
Visual Studio or another compiler with necessary SDKs installed.

Installation Guide

Download and Extract: Get the latest release, extract the contents, and open the src folder.
Set Up Environment: Create a virtual environment within the src folder using:
```
python -m venv .
```
Activate Environment: Activate the virtual environment using:
```
.\Scripts\activate
```
Run Setup Script: For Windows, execute the setup script:
```
python setup_windows.py
```
Enable Ask Jeeves Functionality: Unblock koboldcpp_nocuda.exe in the Assets folder to use this new feature.

Using the Program

To start using the program, remember to activate the virtual environment first. Then, execute python gui.py to launch the application.

Creating and Managing Databases

Download Models: Select and download a vector or embedding model from the Models Tab.
Create Database: In the Create Database tab, select files to add to the database. Available file types include text, images, and audio, which can be converted and summarized before insertion.
Manage Databases: View and delete databases from the Manage Database tab as needed.

Querying the Database

The plugin allows querying the database for relevant information with or without using LM Studio.

Query Without LM Studio: Simply type or record a question and click Submit to get contextual responses.
Query With LM Studio: For more detailed answers, integrate with LM Studio by forwarding the context and question to obtain an LLM-generated response.

Community and Support

For feature requests or to report bugs, users are encouraged to create an issue on the project's GitHub page or contact the developer directly on the LM Studio Discord server.

The VectorDB Plugin for LM Studio is designed to bring a new level of precision to LLM responses by leveraging the power of contextual data retrieval, making it an invaluable tool for users looking to reduce inaccuracies in generated content.