sb2bg/8bit-auto-embeddings

8Bit Auto Embeddings

Description

The 8Bit Auto Embeddings repository contains the backend and machine learning embeddings components used in the 8Bit Auto web application. It uses FastAPI for creating a RESTful API and ChromaDB for managing and querying embedded data.

Features

• FastAPI for efficient and easy-to-document API routes.
• ChromaDB integration for persistent storage of embedding vectors.
• Sentence-Transformer models for generating embeddings from text data.

Getting Started

These instructions will guide you through setting up the project locally for development and testing.

Prerequisites

• Python 3.8+
• FastAPI
• Uvicorn (ASGI server)
• ChromaDB
• Sentence Transformers

Installation

1. Clone the repository:

   git clone https://github.com/sb2bg/8bit-auto-embeddings.git

2. Navigate to the project directory:

   cd 8bit-auto-embeddings

3. Install the required Python packages:

   pip install -r requirements.txt

Running the API Server

1. Start the Uvicorn server:

   uvicorn main:app --reload

   The server will run on http://127.0.0.1:8000 and is accessible via browser or API testing tools like Postman.

Generating Embeddings

To generate embeddings using chroma_embedder.py, follow these steps:

1. Ensure data.csv is in the repository root, formatted with any columns, as long as it includes excerpt, which is the text data to be embedded.
2. Run chroma_embedder.py:

   python chroma_embedder.py

   This script uses the SentenceTransformer model to convert titles from data.csv into embeddings, storing them in a ChromaDB collection named cars which is persisted in the chroma.db folder.

API Documentation

Once the server is running, you can access the API documentation automatically generated by FastAPI at http://127.0.0.1:8000/docs. This documentation provides interactive endpoints where you can test the API functionalities directly.

However, the following endpoints are available:

• GET /chat/{car_str} or /chat/{car_str}?n_results={n_results}
  • Description: Retrieve the n_results (default 5) nearest embedding vector for a given car string.
  • Query Parameters:
    • car_str (str): The car description for which to retrieve the embeddings.
    • n_results (int): The number of nearest embeddings to return (default: 5).
  • Returns: - [{current_bid_formatted: string, excerpt: string,thumbnail_url: string, title: string}]: A list of the nearest car vector matches for the given car string.

Usage

• Use the API to retrieve, update, embedding records.
• Interact with the ChromaDB to query embedding vectors based on textual inputs.
• Generate embeddings from text data using the SentenceTransformer model.

License

This project is licensed under the MIT License - see the LICENSE file for details.