Skip to main content

Oversight and reports

Where can I see what an agent did on a task?​

Open the task, select the Details tab, then Sessions. Audit › Sessions opens on that task's runs only; set Window to All time to include older ones. Each row shows how long the run took, its node and why it ended, and a run that never started a session is under Attempts with no session. To read what the agent did and said, select ⋯ on a row, then Download. You need a role that can view sessions, and your plan must include Session transcripts; downloading needs a role that can download transcripts.

See Follow a task's sessions and How audit works.

Why did an agent's run end before its task was done?​

Its session in Audit says why. In the sidebar, select Audit, then Sessions, widen Window if the run is older than a day, and set End reason; add Node to check one machine. The lines under a row give the recorded cause. For the last output the node's terminal showed, select ⋯ on the row, then Pane log; it is offered only when the node sent one. Viewing Sessions needs a role that can view sessions and a plan with Session transcripts; opening a pane log needs Project Maintainer, Org Admin or Org Owner.

See Find out why a session ended and How audit works.

Why is a compliance gap still not green after its fix merged?​

A gap turns green only once a rescan passes. After Fix all, the gap moves through Fix queued, Fix in progress and PR open to Fix merged — rescan to confirm. Select Rescan on the project's page, or turn on Daily scan on the Settings tab. Compliance is on some plans only, and scanning needs the permission to edit compliance.

See How compliance works and Run a compliance scan.

Where can I see what's failed or stuck across my organization?​

On the command center: select Reports in the sidebar, below Directory. Failed or stuck counts the work that needs a person, and the Risks tab lists each failed, stuck or merge-failed task, with its reason on hover; select a task's name to open it. Only Org Owners and Org Admins see Reports.

See Find what needs attention and Read the command center.

Why didn't a production incident file a task?​

Every connected project records its incidents, but filing tasks happens only when the project's Production QA switch is on, on its Production QA tab. The project also needs the automations that file them: select Create defaults under Automations. At most one open fix task exists per incident, so repeats don't add more. Production QA is a plan feature, and creating the automations is for Org Admins.

See How Production QA works and Turn on Production QA for a project.

Why does a release's Metrics tab say Not measured?​

Jaah measures errors and latency only after a production release of the project, and only when a monitor reports on the project. So the tab reads Not measured: with the reason, such as "not a production release" or "no monitor reports this project". Right after a production release it reads Pending — measured and a time instead: the window is still open, and the numbers appear when it closes. Production QA is a plan feature.

See Release runs and How Production QA works.

Why didn't I get a notification?​

The bell shows only the organization you are working in; switch to another to see its items. Outside the console, each channel has an Off, Brief or Full switch per group under Alerts in your Preferences, and Task progress starts Off by email. A new email address gets nothing until you confirm it. Billing, AI account login and Node failures go only to Org Owners and Org Admins.

See How notifications work and Your profile › Alerts.

Can we accept a compliance gap instead of fixing it?​

Yes, with a waiver: it accepts one failing rule on one repository until a date you choose. On the repository's Overview tab, select ⋯ on the gap, then Waive, and fill in Owner, Expires, Compensating control and Reason. The rule then reads Waived until that date, and fails again once it lapses. A rule in Lock mode, or one that allows no waiver, can't be waived. Waiving needs the permission to edit compliance; Revoke on the Rules tab ends a waiver early.

See Waive a compliance gap and How compliance works.

Where can I see what we spend on AI?​

On the command center's Cost tab: select Reports in the sidebar, pick a range such as 30d, then Cost. Spend is the total, and Spend by project, Spend by agent and Spend by model show where it went; pick a Project to narrow every figure to it. A run posts its spend when it finishes, so tasks still running add nothing yet. Waste is the spend on work that failed. Only Org Owners and Org Admins see Reports.

See See where your AI spend goes and How the command center works.

Can my alerts go to a Slack channel or Telegram?​

Yes. Select your initials, then My Profile, then Preferences and Add channel, then Slack channel, Slack DM or Telegram. Each becomes a channel of its own: pick Off, Brief or Full for each group of alerts it gets. Everyone in a Slack channel or Telegram group sees the alerts you can see, so pick one whose members may read them. Slack channel is offered only to Org Owners and Org Admins, and Slack needs a plan that includes it.

See Add chats as alert channels and How notifications work.

How do I find out who changed something?​

Look in Audit's Activity tab. Select Audit in the sidebar, type a task number, a project or a person in Search or filter activity and pick a suggestion, and narrow the time with From and To. Select the row: the Event tab shows When, the Actor and the Action, and Payload shows the values that changed. All writes in this session lists everything else that person changed in the same sign-in. You need a role that may view audit activity, and a plan that includes Audit.

See Find an event in the audit log and How audit works.

How do I see what my team delivered this week?​

On the command center's Delivery tab. Select Reports in the sidebar, pick 7d (and a Project to narrow it), then select Delivery. Tasks by state counts the tasks in each state; select a count to open them on the board. Created vs done shows whether work arrives faster than it finishes, and Open epics shows the stage each epic has reached. Scroll down for merged pull requests and lines changed. Only Org Owners and Org Admins see Reports.

See Check what your team delivered and Read the command center.

Can my alerts go to a second email address?​

Yes, as a channel of its own. Select your initials, then My Profile, then Preferences, Add channel and Email. Type the address and select Add. It reads Unverified and gets nothing until you confirm it: select the link in the email Jaah sends, or type its 6-digit code in Code and select Confirm. Then select its row and pick Off, Brief or Full for each group of alerts. The address you sign in with keeps its own choices.

See Send alerts to another email address and How notifications work.

Where can I see what the Architectural Guard blocked?​

In the sidebar, select Compliance, then the Guard activity tab. It lists the changes rules in Gate or Lock mode stopped, newest first, on the projects you can see. Filter by Project, Rule and Decision: Denied, Escalated, Error or Waived. Each row names the rule, the file and line, and the task, which opens what the agent was doing. The tab shows only where your organization has the Architectural Guard.

See Review what the guard blocked and How compliance works.

Can Jaah revert a release that breaks production?​

Yes, with Production QA on. On the project's Production QA tab, set Revert policy under Releases: Draft files a draft revert task for someone to start, Auto-file files one that starts on its own, Off files nothing. The project also needs an automation that files tasks from regressions; Create defaults adds one. The revert is an ordinary task: its pull request goes through review, and Jaah never merges it by itself. At most one revert is filed per incident.

See Revert a release that breaks production and How Production QA works.

How do I find out when my task reaches Production?​

Jaah tells you with a Live notice, named after the task, once a release carries it to Production. It shows in the bell with Release as its kind. To choose how it reaches your channels, open My Profile, then Preferences, a channel's row, and the Releases group. Set Live, Approval needed and Release failed each to Off, Brief or Full. The bell gets release notices either way.

See Follow your project's releases and How notifications work.

Why are my pull requests sitting in the merge lane without merging?​

The lane says why. Select Audit, then Merge lanes, and set State to Open: each row says why it last moved. Select the project's name, then point at the dot at the end of the lane's tabs. Held names the run holding the lane: answer it if it is asking, otherwise select Start or Complete. Off says why, with a link to the setting that turns it back on. A red cross named for the branch means the integration branch is red and queued pull requests wait for its fix task.

See Find out why the merge lane is stuck and Use the merge lane.

How do I check that a fix resolved a production incident?​

Start from the incident. Select Incidents, pick the Project, set State to Resolved and open the incident. On Overview, compare Last seen with Resolved, the time it stopped. The Tasks tab lists each task filed for it, under Triage, Fix or Revert; open one to read what it changed. On History, a Reopened above Resolved means it came back. If it reopened, follow the Suspects tab to the pull request behind it. Production QA is a plan feature.

See Check that a fix resolved an incident and Incidents.

Why did one of my alert channels stop getting alerts?​

Its row says why. Select your initials at the bottom of the sidebar, then My Profile, then Preferences, and hover the chip on the channel's row. Unverified: confirm the address with Verify or Resend. Not linked: remove the Telegram chat and add it again. Reconnect Slack: an Org Owner or Admin connects your Slack workspace again. Wrong workspace: reconnect the Slack DM, or remove and add the Slack channel again. No chip: check that its groups aren't set to Off.

See Fix a channel that stopped getting alerts and How notifications work.