Split Test Different Agent Prompts with Supabase and OpenAI

n8n template #2992

Summary

Split Test Agent Prompts with Supabase and OpenAI Use Case Oftentimes, it's useful to test different settings for a large language model in production against various metrics. Split testing is a good method for doing this. What it Does This workflow randomly assigns chat sessions to one of two prompts, the baseline and the alternative. The agent will use the same prompt for all interactions in that chat session. How it Works When messages arrive, a table containing information regarding session ID and which prompt to use is checked to see if the chat already exists If it does not, the session ID is added to the table and a prompt is randomly assigned These values are then used to generate a

How to set it up

Open template #2992 on n8n.io and click Use for free, or copy the template JSON and paste it onto an empty canvas in n8n.
Create 3 credentials under Credentials → Add credential: Supabase, OpenAI Chat Model, Postgres Chat Memory. Then open each of those nodes and pick the credential.
Check the model in the AI Agent node. The template may point at a model your account can't use; swap it for one you have access to.
Click Execute workflow and step through each node's output to confirm the data looks right.
Toggle the workflow to Active so its trigger runs on its own.

Hand off to your agent

Prompt

Help me set up the n8n workflow "Split Test Different Agent Prompts with Supabase and OpenAI" (https://n8n.io/workflows/2992). It uses: Supabase, AI Agent, OpenAI Chat Model, Postgres Chat Memory. Import the template JSON into my n8n instance, list every credential I need to create, and walk me through testing it.

Paste into Claude Code and it will do the rest.

Apps and nodes

Supabase AI Agent OpenAI Chat Model Postgres Chat Memory

Questions

Yes. Chris Carr published it free on n8n.io as a verified creator, and it has 7,697 views. It runs on n8n Cloud or a self-hosted n8n instance.

Supabase, OpenAI Chat Model, Postgres Chat Memory. The other 1 node are built into n8n.

Similar workflows

NameViews