The tables and RPCs your app actually queries in production.
Neat pulls your live query traffic — every table and RPC your app actually hits, with call and error counts — and fuses it with the schema it already extracts from your code.
Set up NEAT's Supabase connector so you can see which tables and RPCs my app actually queries in production. Steps: 1. Ensure NEAT is running for this repo (`npx neat.is`; start the daemon if it isn't up). 2. Gather: my Supabase PROJECT REF (20-char id in the project URL / dashboard settings); a Supabase MANAGEMENT API TOKEN — create a read-only one at supabase.com/dashboard/account/tokens and export it as SUPABASE_MGMT_TOKEN (never paste it inline); the project HOST (<ref>.supabase.co); and WHICH SERVICE in my graph owns this database (e.g. `api`). 3. Run: npx neat.is connector add supabase --api-project-ref <ref> --node-ref <ref>.supabase.co --service-name <service> --token $SUPABASE_MGMT_TOKEN (Optional: if you have a Postgres connection string, pass it so NEAT also reads pg_stat_statements call counts.) 4. Restart the project's daemon so it begins polling. 5. Verify with NEAT's get_observed_dependencies for that service: you should see the tables/RPCs that actually received traffic, with call and error counts. Report which tables show real traffic, and flag any the code declares but prod never touches. NOTE: this connector reads live query TRAFFIC (edge_logs + pg_stat_statements). It does not fetch DDL schema or RLS policies — don't claim it can see those.
See which 'core' tables get zero production traffic, and which undeclared ones get hammered.