A Flow Lab demo built without a client. Walkwren and Kettleby Dog Walks are fictional. A review of your app costs $350 and takes 3 days. How it was built and what it costs
Kettleby Dog Walks took walk bookings through Walkwren, a React and Supabase app made with an AI app builder. All ten findings from the review are now fixed. Seven were confirmed on a real PostgreSQL database, and three were found by code review. Each fix was then checked again.
Each finding shows who could do what, the fix itself and the same check before and after it.
Each action below works for the people it is meant for. These actions were performed on the fixed database as the people who need them. The fix was also applied to data the old app could have left behind.
The fix also gives the owner a list of past bookings to review: bookings priced differently from the price list or filed under someone else’s customer record. Only the server and the Supabase dashboard can read the list.
The owner still needs to decide how late a walk can be canceled, which hours bookings are open, whether 10 upcoming walks per customer is the right limit and whether the app should allow edits other than canceling. For now, those edits go through the Supabase dashboard.
The database checks run on real PostgreSQL (PostgreSQL, in memory). Supabase’s anon, authenticated and service_role roles, the permissions Supabase gives new tables by default and its auth.uid() and auth.jwt() functions are recreated, so the row level security rules work the same way as in real Supabase. The HTTP layer (PostgREST) is not recreated, and the test database has one connection, so the “one booking at a time” lock is checked as a database rule, not with real simultaneous requests. “Before” is the database built from the migration the app builder generated. “After” is the same database with the fix applied on top of existing data, which is how it would go live. The old data may hold values the fix cannot convert safely, such as a time slot typed as “tomorrow 10am,” a date that does not exist or a walker already booked for two walks at once. If so, the fix stops, names those records and changes nothing. The three findings in the app interface (F2, F8, F9) were checked by code review, not by running the app. The “before” code was written for this demo to show problems that are common in Supabase apps made with AI builders.
Flow Lab reads the code and the Supabase setup, tests every finding hands-on and sends a report like this one. You choose what to fix.
Flow Lab fixes the database with migrations that keep your data, updates the code and re-runs every check until all of them pass and the app works as before.
Flow Lab replaces any exposed keys, moves server-only work to edge functions and launches the app while you watch.
The app builder keeps generating code, so each month every check runs again on what changed. New findings go into a short report.