Docs
One Command Deploy Embedding Service on Fly.io

One Command Deploy Embedding Service on Fly.io

How to deploy Embedding Service on Fly.io

How to deploy Embedding Service on Zeabur

You could refer to One Click Deploy You Own Hybrid AI Search Engine

How to deploy Embedding Service on Zeabur on Fly.io

Get Fly.io Account (Optional)

  • Go to Fly.io and create an account.

Install Flyctl (Optional)

  • Install flyctl by following the instructions here.

Clone Embedding Code From GitHub

git clone https://github.com/memfreeme/fast-embedding.git

Deploy Embedding Service

fly deploy
 
fly secrets set API_TOKEN={your token}

The Advantages of deploying on Fly.io

You could refer to One Command Deploy Searxng on Fly.io

Why do you need to deploy your own embedding service?

You could refer to Hybrid AI Search 1 -- how to build fast embedding service

The primary reason for deploying your own embedding service is performance. With a local embedding service, you can achieve results in milliseconds.

For multilingual text embedding, I tested and found that the results of the paraphrase-MiniLM-L6-v2 model are better than the openai text-embedding-3-large model.

The last point is that your data privacy is more controllable.