patw/vectorstream

Automatic embedding using mongo changestreams for Atlas Vector Search

Vector Stream

This tool automatically fetches and updates vector embeddings for your Atlas Vector Search use cases.

• Supports OpenAI (text-ada-002), Mistral (mistral-embed) and VectorService (https://github.com/patw/VectorService)
• Configure the .env file to point to your atlas instance, database, collection and which field to vectorize and which field to store the embedding
• Configure optional OpenAI or Mistral keys for hosted services
• Automatically performs embedding for all records with missing keys
• Listens to changestreams for insert, replace and update events. Calls the embedding model to update embedding field.

Installation

pip install -r requirements.txt

Running App

Copy sample.env to .env and modify with connection string to your Atlas instance

python main.py