From the source
Supabase is launching Supabase Compute in private alpha, letting developers run agents, background jobs and APIs in the same project as their data.
- The company says developers told it some work does not fit in an Edge Function, because it needs more memory or CPU, runs longer than 400 seconds, or is written in another language, and that such work usually means a second platform with its own deploys and logs.
- Each service gets a full Linux environment and can run an AI agent, an API or web service configured with a Dockerfile, a durable job processor pulling tasks from Postgres, a data sync or embedding backfill, or a WebSocket service.
- Services are created with the Supabase CLI, version 2.120.0 or later, by setting compute = true under [experimental] in config.toml, picking a runtime of Node, Deno or a Dockerfile, and choosing a size of 2 GB with 1 vCPU or 4 GB with 2 vCPUs.
- Dreambase moved its data agent onto Compute to test it, and its Postgres queries took a median of 2.9 ms against 65 ms from another region.
- During the alpha Compute is limited to the us-west-2 region, is for evaluation rather than production, and may change or end; services can restart at any time, get all of the project's secrets, and have no autoscaling, GPU or persistent disk.
