Process multiple requests in FIFO using OpenAI Batch API and Supabase/Postgres

How it works

You provide a list of prompts and a system instruction, the workflow batches them into a single OpenAI Batch API request. The batch job is tracked in a Supabase openai_batches table. A cron job polls OpenAI every 5 minutes, and once the batch completes, the results are decoded and stored back in Supabase.

Set up steps

Create the openai_batches table in Supabase. Schema is in the yellow sticky note. Add your OpenAI and Supabase/Postgres credentials to the workflow. Replace the mock data with your actual prompts and you're ready to go!