Enable and pause
A screen or its schedule can be paused from the console, from the bot, or by asking the assistant. Pausing keeps the screen, its versions and its history.
Screens
It keeps workingafter you closethe tab
A screen is a set of conditions plus a cadence. It runs on the server, on closed candles, whether or not anything of yours is open — and what it found is waiting when you come back.
The two decisions in a schedule
Cadence: match it to the candle
Screening an hourly condition every five minutes gives you the same closed candle twelve times. That is not more information; it is the same information with eleven more chances to alert on it. Match the cadence to the interval you are testing, and let the per-symbol cooldown handle the rest.
Closed candles, always
An indicator computed on a forming bar changes every tick. It can cross your threshold and cross back before the bar ends, so a screen reading it fires on a value that never existed at the close and will not be there in the backtest. This was a real defect here — the unclosed candle reached the analyst’s maths whenever the window was left open — and closed-only is now the default rather than an option you can forget.
A run is a record, not a moment
Each run is written down with what it scanned and what it found, and a run that could not evaluate a symbol is marked degraded. Clean runs with no hits and degraded runs with no hits mean opposite things, and a screener that reports them identically is telling you the market is quiet when its data source is down.
Controls
Versions, not edits
Changing a screen creates a new version. History that spans an edit therefore does not silently mix two different rules.
Cooldowns
Per-symbol, so a condition that stays true for six hours does not become six hours of alerts.
Delivery thresholds
A minimum strength before a firing is pushed anywhere, so the channel that matters stays worth reading.
Questions about schedules
What happens if the platform restarts mid-schedule?
Schedules are durable; they are not a timer in a running process. A restart resumes them rather than losing them, which is the difference between a scheduler and a loop.
Can a schedule wake the assistant?
Yes — that is the signal-waiting loop: a screen finds candidates, one assistant turn returns a typed verdict, and a confirmation reaches you. The turn has a daily budget that is checked before it runs, not after.
How many screens can I schedule?
That is one of the things the plan ladder actually rations, along with history depth and strategy searches — because frequency is what the platform is billed for.
Set a cadence and walk away
Free paper accounts get durable schedules against live data.