vectorMatch

Text Embedding and Search with PostgreSQL and Hugging Face in Docker

This project demonstrates a Python script that embeds text using a model from Hugging Face, stores the embeddings in PostgreSQL with the pgvector extension, and allows searching the database using regular text queries by comparing embeddings. After the data is retrieved an llm is used to generate a response with ollama. The Project is run with Docker Compose

Features

• Embeddings: Use Hugging Face's transformers to embed input text.
• PostgreSQL with pgvector: Store embeddings in a PostgreSQL database using the pgvector extension to perform vector-based searches.
• Search Functionality: Retrieve database entries by comparing the input text's embedding to the stored embeddings.
• Docker Support: Run the whole application with Docker compose
• Ollama: Generate response based on local llm

Prerequisites

Make sure you have the following installed:

• Docker

Setup

Get the project directory

git clone https://github.com/comhendrik/vectorMatch.git

Start docker and go into the project directory and run the compose file

docker compose up

Wait for the script to be done, this can take a few minutes and then attach yourself to the vectorMatch container

docker attach vectormatch-vector-match-1