uussoop/qdrant_template

Qdrant template in python

This project shows how to use Qdrant for vector search alongside OpenAI embeddings with docker version of qdrant.

Installation

1. Clone this repo
2. Run the following command to automatically detect your platform and install Docker:

curl -fsSL https://get.docker.com -o get-docker.sh
sh get-docker.sh

3. Install requirements: pip install -r requirements.txt
4. Get an OpenAI API key and add it to .env as OPENAI_API_KEY
5. Start the Qdrant server:

docker compose up -d

This will run a Qdrant server on port 6333 with the provided config.

Usage

The main functions are:

• insert_data - Takes a collection name, payload list, and data list. It embeds the data with OpenAI, structures it into Qdrant points, and inserts them into the collection.

• search_data - Takes a search term, collection name, and result limit. It embeds the search term with OpenAI and searches the Qdrant collection, returning the top matches.

To insert data:

insert_data("test", [{"data":"doc1"}, {"data":"doc2"}], ["Text for doc1", "Text for doc2"]) 

To search:

results = search_data("search text", "test")

The results will contain score, vector, and payload for each match.

See the bottom of the main code sample for example usage.

Configuration

The Qdrant server uses the provided docker-compose.yml file and config. Environment variables are loaded from .env.

The default OpenAI model is text-embedding-ada-002 but this can be changed.

The Qdrant client connects to the server on port 6333 by default.