Skip to main content
October 21, 2026 · v0.5.157

New Features

A Notifications Page and Detailed Notification Settings

Notifications now have their own page, and settings go well beyond turning a kind on or off:
  • One page for everything - /notifications lists every notification, read or not, grouped by where it came from: an app, a category within it, and the kind of notification.
  • Categories declared by apps - A custom app can group its notifications into categories (for example operations/food-safety) in oxy-app.json, giving each kind a label, an urgency level and up to three actions. Apps that don’t declare categories keep working; their notifications appear under General.
  • Mute by app, category or kind - Turn a channel (push, email, Slack) on or off at any level. A narrower setting wins, so you can mute a whole app but keep one kind you care about.
  • Pause and quiet hours - Pause all notifications until a set time, or set quiet hours by day and timezone. Pushes held during quiet hours arrive as a single summary when they end, and you can let critical or time-sensitive notifications through.
  • Urgency floors and per-device controls - Send only notifications above a chosen urgency to a channel or a device; everything else still reaches the inbox.

Notifications That Reach People

  • Slack DMs reach unlinked members - A Slack notification used to reach only people who had already linked their Slack account, so most were never delivered. Oxygen now finds org members in Slack by their email address. Platform notices such as failed runs, billing issues and anomalies are sent to Slack too, and every DM says which app sent it.
  • Browser push you can turn on from the inbox - A Get notifications on this device prompt appears in the notification bell, the mobile inbox and the notifications page, instead of only deep in Settings. Pushes now name the app and the kind of notification, not just “You have a new notification”.
  • One alert per failing streak - A pipeline or automation that keeps failing used to send a notice on every failed run, which could mean an email every hour. You now get one alert when a source starts failing and at most one reminder a day while it keeps failing.
  • Test notifications from the console - The custom-app admin console can send a test notification as the app, delivered only to you.

Custom-App Promote and Rollback

Moving a custom app to production now follows one safe path:
  • Production only serves what staging has served - oxyc promote moves a build that has been verified in staging (or promotes a sandbox build with --from), and oxyc rollback returns production to an earlier build.
  • Migrations before every move - Database migrations bundled with an app run before production switches to the new build, whether you promote, roll back or publish, so new code never meets tables that don’t exist yet.
  • Schedules, webhooks and Run now outside production - A function that runs on a schedule or a webhook can now be exercised in staging and in sandboxes, so timing, retries and signature checks are tested before production.
  • Secrets per environment - The admin console’s Secrets pane has an environment picker for production, staging and sandboxes. Values can be copied between environments without leaving the server, every secret change is audited, and the Promote, Publish and Roll back confirmations warn when the build needs a secret that isn’t set.

Live Function Logs

oxyc logs --follow streams a custom app’s function output as it happens, for any environment. Before, logs from scheduled runs, slow functions or calls made from the app’s own page arrived about 30 seconds late, or not at all on deployments without trace capture.

Faster Custom-App Development

  • Run a function locally - oxyc fn dev <name> runs one function from your app’s source on your machine against a test context built from oxy-app.json, with --watch to rerun on save. Changing one line no longer means a full build, upload and remote call.
  • One-step first publish - Publishing a new app is a single oxyc publish, and oxy-app.json is checked against one shared definition by the CLI, the Vite plugin and the server, so they no longer disagree on which fields are valid.
  • Open tabs pick up new builds - A custom app left open on a tablet or kiosk now notices a new build within about five minutes. If a page fails to load after a publish, the tab reloads once onto the new build instead of breaking. Apps can opt in to reloading automatically at a safe moment.

Sign-In for Agents Without a Browser

An agent in a cloud container, a CI job or on a remote machine can now get its own credential. Run oxyc login, oxyc tokens create --agent or --sandbox-agent with --no-browser: it prints a link and an eight-letter code for a person to approve. Name the approver and Oxygen notifies them by push, email, Slack and the inbox, so they can approve from their phone.

Platform Improvements

Analytics

  • Readable dates in rollup answers - When the analytics agent answered from a pre-aggregation, dates and timestamps came back as raw internal values, so charts had unreadable axes and the agent spent its answer converting them. They are now standard dates and times.

Custom Apps

  • GitHub webhooks are accepted - A GitHub webhook to a custom-app function was always rejected with a signature error. GitHub’s signature format is now recognized.
  • Faster function calls - Functions now start from a prepared snapshot, cache their bundles and settings, and need fewer database queries per call, so each invocation starts sooner.
  • Live logs need app admin access - Streaming an app’s logs requires admin access to that app in every environment.

Reliability and Errors

  • Clearer database connection errors - When a Postgres server refuses a connection, for example because an account is over quota, the error now shows the server’s message instead of a generic “error connecting to server”.
  • Staff limited to some organizations - The admin user directory and user status pages now respect a staff member’s organization scope.