Running and watching a run

The right panel of the Explore tab is the run panel. It starts an automation through the Robot, follows the run folder the Robot writes and lists what ran before. The Studio does not execute anything itself. It is the caller, not the runner.

#Starting a run

Open an automation and press Run. The Studio posts to /api/run with the automation id. The server spawns the Robot as a child process with a fresh run id. The headless checkbox controls whether the browser is visible. A green dot beside "demo app" tells you the demo application is up on port 4100, which the shipped examples need.

One run at a time. If a run is already active the server refuses a second one with a message naming the active run, so the panel never has to guess which run your log lines belong to. Stop terminates the whole Robot process tree.

#Watching a run

While a run is live the Studio polls its state.json and logs.txt every second. The status pill moves through starting and running and then settles on one terminal state: success, completed-with-errors, failed or stopped. Beside the pill you get the current transaction, the current phase and the run id, plus a row of totals: transactions, succeeded, business exceptions, system exceptions, retries and skipped.

When the state reaches a terminal value the Studio fetches the summary and the artifact list once, then refreshes the run list. A run that ends completed-with-errors is not a failure of the run itself, so read the transaction list rather than the exit alone.

#The run list and the logs

The run list shows recent runs across every automation, newest first, each with its status, its start time and its automation id. Click one to load a finished run: its state, its summary, its artifacts and its logs, without starting anything new.

The bottom of the window is the log panel. It shows the tail of the run's logs.txt, the human-readable stream, up to the last 400 lines, with a live marker while the run is going. A finished run shows its full tail.

#Screenshots, traces and the run folder

Every run gets its own folder under artifacts/<automationId>/<runId>/. The Studio lists the files it finds there in the Artifacts section, with their sizes. The pieces you will reach for:

• state.json is the machine-readable status the panel polls.
• summary.json holds the per-transaction outcomes, which the panel shows as a collapsible transaction list.
• screenshots/ holds the captures the framework takes on business and system failures. The Debugger tab renders them inline for a finished run.
• trace.zip is the full Playwright trace. The Trace button opens it in the Playwright trace viewer in a separate window. You can also download the zip from /api/run/<id>/trace.

The Debugger tab does more with the same folder. It reads the run's evidence and explains the outcome, marking each claim as a fact or as something it could only infer. It is read-only and it never replays a run.

#Things that catch people out

• The Run button is only live when an automation file is open. Non-automation files and the other tabs do not enable it.
• If the Robot is killed rather than finishing, the server patches its state so a run cannot sit on "running" forever.
• summary.json is written at the end. Selecting a live run can show a not-written-yet answer for the summary while the logs and state are already there.
• The log panel reads logs.txt. The structured logs.jsonl is on disk beside it, but it is not what the panel shows.