Supabase launches compute features for long-running agent workloads
Supabase introduced Supabase Compute to run long-running workloads in the same private network as the database, removing previous time limits. The company also released pg-donta, a library that represents database state as files to facilitate agent-driven development.
Supabase announced the launch of Supabase Compute, a serverless product designed to handle long-running workloads such as chunking and embedding. Unlike the company's previous Edge Functions, which were powered by Deno and limited to two-second execution times, Supabase Compute allows these tasks to run in the same private network as the database. The platform supports Rust, Python, and agents, addressing a key limitation for developers building AI applications that require sustained processing power.
The company also released pg-donta, a library that enables declarative schemas. This tool represents the database state as files in a local folder rather than a historical log of migrations. This approach allows coding agents to understand the database shape without running migrations or introspecting a live instance, facilitating faster iteration and establishing a Git-based source of truth for agent-driven development. Supabase stated that while pg-donta can be used with any Postgres database, the company has built specific mechanisms to integrate it with its own platform.