Generate Embeddings

Set up Supabase locally

Make sure you have the latest version of the Supabase CLI installed.

Initialize Supabase in the root directory of your app and start your local stack.

supabase init
supabase start

Create Edge Function

Create an Edge Function that we will use to generate embeddings. We'll call this embed (you can name this anything you like).

This will create a new TypeScript file called index.ts under ./supabase/functions/embed.

supabase functions new embed

Setup Inference Session

Create a new inference session to use for the lifetime of this function. Multiple requests can use the same inference session.

Currently, only the gte-small (https://huggingface.co/Supabase/gte-small) text embedding model is supported in Supabase's Edge Runtime.

const session = new Supabase.ai.Session('gte-small');

Implement request handler

Modify our request handler to accept an input string from the POST request JSON body.

Then generate the embedding by calling session.run(input).

Deno.serve(async (req) => {
  // Extract input string from JSON body
  const { input } = await req.json();

  // Generate the embedding from the user input
  const embedding = await session.run(input, {
    mean_pool: true,
    normalize: true,
  });

  // Return the embedding
  return new Response(
    JSON.stringify({ embedding }),
    { headers: { 'Content-Type': 'application/json' } }
  );
});

Note the two options we pass to session.run():

• mean_pool: The first option sets pooling to mean. Pooling refers to how token-level embedding representations are compressed into a single sentence embedding that reflects the meaning of the entire sentence. Average pooling is the most common type of pooling for sentence embeddings.
• normalize: The second option normalizes the embedding vector so that it can be used with distance measures like dot product. A normalized vector means its length (magnitude) is 1 - also referred to as a unit vector. A vector is normalized by dividing each element by the vector's length (magnitude), which maintains its direction but changes its length to 1.

Test it!

To test the Edge Function, first start a local functions server.

supabase functions serve

Then in a new shell, create an HTTP request using cURL and pass in your input in the JSON body.

curl --request POST 'http://localhost:54321/functions/v1/embed' \
  --header 'Content-Type: application/json' \
  --header 'apikey: SUPABASE_PUBLISHABLE_KEY' \
  --data '{ "input": "hello world" }'

Be sure to replace SUPABASE_PUBLISHABLE_KEY with your project's publishable key. You can get this key by running supabase status.