Run your Flow

Now that you’ve defined and compiled your flow, it’s time to execute it!

In this guide, we’ll set up an Edge Worker to process your flow tasks, trigger your first flow, and observe its execution.

Prerequisites

Before starting, make sure you have completed:

• Create your first flow guide
• Compile flow to SQL guide

1. Create a worker function

   Create a new Edge Function that will process tasks for your flow:

   npx supabase functions new greet_user_worker

   Replace contents of index.ts file with the following:

   supabase/functions/greet_user_worker/index.ts

   import { EdgeWorker } from "jsr:@pgflow/edge-worker";
   import GreetUser from '../_flows/greet_user.ts';
   
   // Pass the flow definition to the Edge Worker
   EdgeWorker.start(GreetUser);

   The Edge Worker will:

   1. Poll for tasks on the greet_user queue
   2. Process tasks based on the step slug
   3. Return the appropriate result for each step
   4. Handle retries if any step fails

2. Disable JWT verification

   Edit supabase/config.toml to disable JWT verification for local development:

   [functions.greet_user_worker]
   enabled = true
   verify_jwt = true
   verify_jwt = false
   import_map = "./functions/greet_user_worker/deno.json"

3. Start the Edge Runtime

   Start the Edge Runtime to make your worker function available:

   npx supabase functions serve

   This will start the Edge Runtime server but not yet start your worker. You should see similar output in your terminal:

   Setting up Edge Functions runtime...
   Serving functions on http://127.0.0.1:54321/functions/v1/<function-name>
   Using supabase-edge-runtime-1.67.4 (compatible with Deno v1.45.2)
   serving the request with supabase/functions/greet_user_worker

4. Start your worker

   In a new terminal, send an HTTP request to start your worker:

   curl http://localhost:54321/functions/v1/greet_user_worker

   You should see output in your Edge Runtime terminal indicating the worker has started:

   [Info] [INFO] worker_id=unknown module=DenoAdapter DenoAdapter logger instance created and working.
   
   [Info] [INFO] worker_id=unknown module=DenoAdapter HTTP Request: null

   Caution

   If you have troubles starting the worker, make sure you disabled JWT verification in the previous step.

5. Trigger your first flow

   Now let’s start a flow run! Using Supabase Studio:

   1. Open Supabase Studio in your browser (typically at http://localhost:54323)
   2. Navigate to the SQL Editor
   3. Execute this SQL to start your flow:

   SELECT * FROM pgflow.start_flow(
     flow_slug => 'greet_user',
     input => '{"first_name": "Alice", "last_name": "Smith"}'::jsonb
   );

   The output should look like:

   run_id        | flow_slug    | status  | input                                         | output | remaining_steps
   --------------+--------------+---------+-----------------------------------------------+--------+-----------------
   <run_id UUID> | greet_user   | started | {"first_name": "Alice", "last_name": "Smith"} | null   | 2

6. Monitor execution

   Your worker should start processing tasks immediately. You should see log output in the Edge Runtime terminal as each step executes:

   [Info] [INFO] worker_id=<long UUID> module=ExecutionController Scheduling execution of task 1
   
   [Info] [INFO] worker_id=<long UUID> module=ExecutionController Scheduling execution of task 2

   You can check the run status using Supabase Studio:

   In the SQL Editor, run this query:

   SELECT * FROM pgflow.runs
   WHERE flow_slug = 'greet_user'
   ORDER BY started_at DESC
   LIMIT 1;

   As steps complete, you’ll see the remaining_steps count decrease. When it reaches 0, the run will be marked as completed:

   run_id        | flow_slug    | status    | input                                         | output                              | remaining_steps
   --------------+--------------+-----------+-----------------------------------------------+-------------------------------------+-----------------
   <run_id UUID> | greet_user   | completed | {"first_name": "Alice", "last_name": "Smith"} | {"greeting": "Hello, Alice Smith!"} | 0

Where next?

How pgflow works Understand how pgflow executes flows behind the scenes

Ship to Supabase Deploy your flows to production on Supabase's hosted platform

Make flows resilient Configure retry behavior, timeouts, and error handling for your flows

Observe runs with SQL Learn how to monitor and debug your flow execution with SQL queries