Wakes, runs & activity

A wake is the moment Brix goes to work; a run is the work itself. This page is how both behave — schedules, self-checks, watching a run live, and the notifications that come back.

Wakes

Every Brix has three ways to decide when to work:

• Its schedule — the interval you set (15 minutes to 7 days). Each run writes the next occurrence before it starts, so one bad run can never stop the chain.
• Its own judgement — after a run, Brix can decide it should check back: "the PR is open — look at review in a day." Self-wakes are clamped to a sane window: at least 5 minutes away, at most 30 days.
• You, on demand — ask it in chat to do, check or look at something and Brix starts a run right away (within about a minute — the scheduler is nudged, not asked to wait). The request runs as an ordinary run and its report lands back in the thread. Scheduled wakes are for "later" and "keep doing this" — not for "do it now".

A scheduler claims due wakes one at a time, roughly once a minute — work is serialized, so a busy account's wakes run in order rather than all at once.

Daily cap. Each Brix has its own cap of 24 wakes per rolling day by default, on top of the session limits every cloud run draws from (see Safety, limits & billing). A due wake beyond the cap is deliberately slipped to the next day and marked — never dropped silently, and never a surprise gap.

Runs

Each fired wake becomes a run — one cloud session pod:

• The repository is cloned, the agent works headless against the wake brief, and the pod is deleted when it exits. Wall clock: up to 90 minutes per run; most finish far sooner.
• In write mode the deliverable is git: commits land on the session branch and a pull request opens against the base branch. A run that produced no commits reports exactly that.
• In monitor mode there is no branch and no PR — the report is the deliverable.

Watch a run live. A run is an ordinary cloud session, so the live viewer works exactly as it does for nex cloud: app.nexrall.com/live/cloud-<session-id>. After it finishes, the log's final tail is captured and survives — read it back with nex cloud logs <id>. Stop a run mid-flight with nex cloud stop <id>, or with the Stop button on the run's sheet in the Brix panel (a fast report-only run cannot be stopped — it ends on its own within about 90 seconds).

Open any run. Every run row and every report in the thread opens a run sheet: the brief it worked from, the full report, what it actually did (for a fast run, the tools it used round by round), the cost, and — for a write run — the branch and the pull request. Runs still in flight refresh on their own while the sheet stays open.

When a run can't start. Runs draw from the same session limits as everything else on your account — one session running, up to 3 queued, 10 per rolling 24 hours. If a wake collides with those limits, both the wake and the run record why, and you get a notification — Brix never piles up invisible work.

The settle step

When the pod finishes — success or failure — Brix does one short settling pass:

1. Reads the run's final log.
2. Writes notes for the next run (see Notes & memory).
3. Decides whether to schedule a self-wake.
4. Sends the notification.

Honest failure is built in: if the summary can't be produced, the report says so ("finished — a summary was not available") rather than inventing one, and a failed run's report says how far it got.

Notifications

When a run settles, the result reaches you on all three channels — push, email, and in-app:

Situation	What you see
Monitor finished	Brix report: <repo> — the report text as the body
Write run finished	Brix finished: <repo> — the summary, and where the PR is
Run failed	Brix could not finish: <repo> — why, and how far it got
Run couldn't start	Brix could not start: <repo> — the reason, e.g. another session was running

Cancelling

• A scheduled wake can be cancelled any time before it fires, and a recurring task can be paused without cancelling it.
• A running wake belongs to its in-flight run — stop the session instead (nex cloud stop <id>, or the Stop button on the run's sheet); the wake and the run both record what happened.