Activity | Faustine Docs

Activity is the list of runs Faustine performed on a Web Agent: analyses, auto-updates, design reads and component imports. It is the Activity tab of the agent. Conversations with users are in the Monitoring tab (see Read the logs), and published versions are in the Versions tab.

Usage

In the portal, go to Web Agents, select your agent and select the Activity tab. The list is ordered from newest to oldest. To open a run with its log, select its row. Refresh loads the list again.

Columns

Column	Content
When	When the run was created.
What	The kind of run.
Event	What the run was about: the commits an auto-update compared, or the outcome of a design read.
Status	Where the run stands or how it ended.

Kinds of run

The filter shows the kinds you check. With none checked, it shows every kind.

Filter	What	A run of this kind
Auto-updates	Auto-update	Compares the connected branch with the commit the agent was built from and adapts the draft. A push starts one when auto-update is on. Check for updates starts one by hand. Uploading a new ZIP starts one for an agent created from a ZIP.
Analyses	Analysis	Reads the code and writes the modules and actions.
Design reads	Design read	Reads the look of your app from its code.
Component imports	Component import	Looks for the entity cards of your app in its code.

Auto-update statuses

Status	Meaning
Queued	The run waits, because the draft has unpublished changes or an earlier update is waiting.
Running	The run is in progress.
Published v2, with the number of the new version	The run changed the draft and published it as a new version. The release target serves that version.
Draft updated	The run changed the draft and did not publish it. The draft was edited or a version was published while the run worked, or the publish failed. The run's log gives the reason. Publish the draft to serve the changes.
No impact	The new commits change nothing for the agent.
Collapsed	Resolve all merged this queued run into one run that covers all of them.
Skipped	There was nothing to compare with: no analysis has run, or the agent has no published version.
Failed	The run ended with an error. The run's page shows it.

The page of an auto-update shows the two commits, a link to the diff at the code host, what changed for the agent, the cost and the log.

Pending updates

Auto-updates wait while the draft has unpublished changes. The tab then shows the number of pending updates, with two buttons:

Button	What it does
Resolve next	Runs the oldest queued update.
Resolve all	Runs one update that covers every queued one. The queued runs end as Collapsed.

Both need a draft with no unpublished changes. Otherwise the tab shows The draft still has unpublished changes — publish it first.

Analysis statuses

Status	Meaning
Queued	The analysis waits for a free slot.
Running	The analysis is in progress.
Needs input	The analysis found the modules and waits for you to choose which ones to analyze.
Retrying	The analysis is paused and resumes by itself.
Cancelling, Cancelled	Someone cancelled the analysis.
Interrupted	The analysis stopped before it finished and can be resumed.
Completed	The analysis finished.
Failed	The analysis ended with an error.

The row of an analysis that is running, queued, interrupted or waiting for input opens the analysis page, where you drive it. The row of a finished one opens its figures: duration, cost, the modules with the result of each, the application insights and the log.

Design read and component import statuses

Status	Meaning
Queued	The run has not started.
Running	The run is in progress.
Retrying	The run is paused and resumes by itself.
Completed	The run finished. Event says what it found.
Cancelled	Someone cancelled the run.
Failed	The run ended with an error. Event shows it.

Next steps

Read the logs: what users did with the agent.
Keep the agent up to date: turn auto-update on and review what it changes.
The analysis: what an analysis reads and writes.
Publish a version: serve a draft that an auto-update changed.