Supabase adds edge AI hooks
Supabase Edge Functions now accept webhook receivers (Stripe, GitHub), support on‑demand image generation, and run small AI inference tasks — effectively enabling low‑latency, privacy‑preserving AI at the edge. That edge‑first pattern reduces inference latency and provides a practical path to orchestration of external LLMs near clients. (supabase.com)
Supabase documented a built‑in Edge Runtime inference API on April 16, 2024 and published example code that instantiates Supabase.ai.Session('gte-small') to generate embeddings directly in an Edge Function. (supabase.com)) Since Edge Runtime v1.36.0 the platform can run the gte‑small model natively inside Edge Functions without external dependencies, enabling on‑node embedding generation rather than calling external inference endpoints. (github.com)) Embedding inference in the Edge Runtime uses the ONNX runtime and a native extension built via the Rust wrapper Ort to reduce WASM cold‑boot costs, and Supabase notes embeddings are lightweight enough to run on CPU without GPU acceleration. (supabase.com)) Supabase provides production examples and a Stripe webhook template in its examples repo (including commands like supabase functions deploy and supabase secrets set), showing how third‑party webhook receivers can be wired into Edge Functions for event‑driven embedding or processing pipelines. (github.com)) On‑demand image and Open Graph generation is listed in the Edge Functions guide and community examples demonstrate calling OpenAI’s image APIs from Edge Functions for image generation workflows. (supabase.com)) A complete demo (“Semantic Search with pgvector and Supabase Edge Functions”) links database webhooks to Edge Functions that generate embeddings and store vectors in Postgres (pgvector), enabling similarity search via RPC from the database. (github.com)) Supabase’s Edge Functions run on a Deno‑compatible runtime, are globally distributed for low latency, and the docs explicitly advise treating Postgres as a remote, pooled service and keeping credentials in project secrets for serverless safety. (supabase.com))
