Generate and store Supabase images with Claude and Cloudflare Workers AI

Quick Overview

This workflow listens for new rows in a Supabase image_jobs table, uses Anthropic Claude to expand each prompt, generates an image with Cloudflare Workers AI (FLUX), uploads the image to Supabase Storage, and writes the public image URL back to the original Supabase record.

How it works

1. Receives a webhook call from a Supabase Database Webhook when a new row is inserted into the image_jobs table.
2. Extracts the inserted record ID and prompt value from the webhook payload and stops early with a JSON response if the prompt is empty.
3. Sends the prompt idea to Anthropic Claude to produce a single, detailed image-generation prompt.
4. Calls Cloudflare Workers AI (FLUX) to generate an image from the enhanced prompt.
5. Converts the returned base64 image into binary data and uploads it to a public Supabase Storage bucket using the job ID as the filename.
6. Builds the public Supabase Storage URL, updates the Supabase row with status = done and image_url, and returns the URL in the webhook response.

Setup

1. Create the image_jobs table in Supabase (for example by running the provided new-workflows-setup.sql) and ensure it includes fields like prompt, status, image_url, and error_message.
2. Create a Supabase Storage bucket named generated-images (or update the bucket name in the workflow) and enable Public access so the public URL works.
3. Add credentials for Supabase (project URL/host and service_role key), Anthropic Claude (HTTP header auth x-api-key), and Cloudflare Workers AI (HTTP header auth Authorization: Bearer <token>) and replace the placeholders in the workflow.
4. In the Config step, set supabaseUrl to your Supabase project URL and confirm the bucket name and Claude model value.
5. In Cloudflare, set your account ID in the Workers AI request URL, then create a Supabase Database Webhook for image_jobs inserts pointing to your n8n webhook URL /webhook/generate-image.