dusk bot
integrationssupabase

dusk botintegrations

a supabase bot for the questions before a release

staging has the new column. does production have the migration that added it? give a dusk bot the references for both supabase projects.

by the dusk bot team · checked · how we made this

it can compare applied migration versions, read backup records and report disk use. keep production read only. the resulting list tells the release engineer where to look before the rollout.

pixel art: a pink pixel bot, a cyan pixel bot and a yellow pixel bot beside the supabase mark on a white tile and a server box, on a grassy meadow in front of sand dunes by the sea, one saying "did production get it"

what a bot can do in Supabase

identify the project before interpreting its records. a clean backup list from the wrong environment answers nothing. project details include the region and organization. confirm those once, then use each project's reference for the recurring reads.

actiondusk bot treats it aswhat you would use it for
Get projectreadidentify the project and its region before comparing environments
List migration historyreadcompare the applied migration versions on both projects
Get a migrationreadinspect a particular entry when one version needs explaining
Get Backup Schedulereadread the production backup schedule
List project database backupsreadlook for actual backup records covering the expected window
Get Disk Utilizationreadbring back disk use for the project under review

composio, the connection service behind dusk bot's apps, lists 235 Supabase actions (version 20261116_00, read 20 november 20263). dusk bot treats 99 as reads, 5 as sends and 131 as changes.

only actions starting with a read word (get, list, search) count as reads, so a read-only bot can't call the rest. a send reaches people or moves money, like accept invite external jit access, and waits for your Allow unless you set a daily amount.

connect Supabase

  1. open Connected apps in the sidebar.
  2. pick Supabase and sign in on Supabase's own page, so your password goes to them and never to a bot.
  3. choose which bots may use it, and whether each may only read or may also change things. your bookkeeper doesn't need your Supabase.
  4. ask in plain words. reading and drafting happen freely; sending asks you first.

label the two project references in chat. connect supabase, pick the bot that handles release reviews and allow only the reads it needs. ask it to return each project's region beside its reference. correct a swapped label before the first scheduled comparison.

routines to copy

use your own operating limits. tell the bot when you expect a backup and the disk threshold your team acts on. make the unit explicit. a percentage and an amount of space need different comparisons.

migration gap before rollout

every weekday at 3, compare the applied migration versions in my staging and production projects. list versions unique to each side. for the newest difference, read its migration entry and summarize its purpose here. if the histories match, give me one line. don't run any sql.

weekdays at 15:00 · versions to reconcile before rollout

backup inventory

each monday at 8, read production's backup schedule and backup list. compare the most recent expected backup with the available records. show the scheduled time and latest record time here. distinguish an empty list from an access error. don't attempt a restore.

mondays at 08:00 · evidence for the operator who owns recovery

disk headroom

daily at 11, read disk use for my production project. compare it with <threshold, e.g. 80%>. bring me the value and the project reference if it crosses that line. if the units don't match, show what came back and ask which limit to use.

daily at 11:00 · a threshold crossing tied to the right database

what it asks you first

the release review needs no permission to apply a migration. give the bot the exact reads above. applying migrations and restoring data change the database. a send approval won't stop those writes. keep both actions out of the release bot's permissions.

what it cannot do

  • prove two schemas match from their migration lists. those lists show the versions each project recorded. changes made outside migrations need a separate comparison.
  • prove a restore will work by listing backups. backup availability depends on the project's plan and setup. a recovery drill tests what a list can't.
  • repair capacity by reporting disk use. the disk action reads utilization. a person still has to choose what to retain or resize.

who should own the release

keep the deployment in the team's release process. if you need a migration applied exactly once in order, use that process. give the bot the narrower question that often holds a release up. which recorded version is missing from which environment?

what it looks like

a made-up staging and production comparison.

before

owen expects the parcel status migration in both projects. staging lists version 202611180915. production's history ends with the previous version.

after the bot

staging only: 202611180915, adds parcel status.
release question: should this version be in production yet?

an illustration with made-up names and numbers.

questions

does the migration comparison need access to customer rows?

no. compare the recorded versions by project reference. the history and migration entry actions supply the release evidence. leave table reads out of this job.

why report a backup schedule and a backup record?

they answer different questions. the schedule tells you when to expect a backup. the list shows what the project actually returns. investigate a gap before relying on that window.

sources

  1. dusk bot docs: connected apps
  2. dusk bot docs: routines
  3. composio, Supabase toolkit (catalog version 20261116_00), checked 20 november 2026

drafted with ai help from the dusk bot docs and the sources listed on this page, then checked against them. if something here is wrong or out of date, write to [email protected].