edge-runtime VS vecs

- Functions have a critical bug that prevent us from using in development so we decided to move to our own API servers: https://github.com/supabase/edge-runtime/issues/212

source: https://github.com/supabase/edge-runtime

we needed this to get feature parity across local, the platform, and self-hosting

Refactoring the module loader fixes a few other bugs, such edge functions erroring out when an deno.lock file is already present in the project.

Cold-starts are the time it takes for the “initial load” of an Edge Function. Because the model needs to be downloaded to the Edge Function, could starts can take anywhere from ~2-6s (based on the model). Loading the initial model and building the pipeline usually contributes to it. We are experimenting with the idea of attaching a “read-only disk” of models to our Edge Runtime which mitigate any download penalties. We’ll share more details about these optimizations in a future blog post.

[Supabase engineer & Author of the blog post]

Hi Mark, hosted Supabase Edge Functions would still run on Deno Deploy, so those limitations would still exist. However, we plan to introduce file system access via integration with Supabase Storage. This is still a rough idea stage, maybe we'll have a solid answer in a couple of months :)

For dynamic imports, we haven't looked into it since Supabase users haven't requested it. If you can open an issue on Edge Runtime repo [0] and explain how you intend to use them, we can probably work on a solution.

[0] https://github.com/supabase/edge-runtime/issues

supabase/vecs is our Python client for vector projects. Today, we're releasing a new feature for vecs called "adapters".

hey hn, supabase ceo here.

This is an iteration of several things that we’ve developed to support Hugging Face.

1. Python adapters. The first is our python lib, Vecs [0], which now supports “adapters”. Adapters are simply a pipeline of transformations that an input goes through (for example: split the text into chunks, then turn it into an embedding). We’ve included support for Hugging Face sentence transforms to automatically download the models on first run, and then cache them for subsequent runs.

2. Deno support. I want to start by pointing out that most of the hard work to support Hugging Face in Javascript is through Joshua’s continued efforts on Transformers.js (Joshua is an employee at HF). To support this effort we’re rolling out support across all of the Supabase Edge Functions. We’ve started with sentence embeddings for now since that’s the most common use-case for pgvector.

At the moment 98% of Supabase customers use OpenAI to create embeddings in pgvector. With today’s release we’re hoping to add more options, especially for situations where fewer dimensions are viable (for better pgvector performance[2])

A few of the supabase engineers will be in the comments to answer any questions, as well as Joshua from HF (@xenova on HN)

[0] Vecs: https://github.com/supabase/vecs/

[1] Transformers.js: https://huggingface.co/docs/transformers.js/index

[2] pgvector performance with lower dimensions: https://supabase.com/blog/fewer-dimensions-are-better-pgvect...

Vecs is a python client storing and searching vectors backed by Postgres and pgvector. It gives a familiar collection-like interface to upserting and searching.