Private checkpoint and meaningful alerts
Use the host's supported private persistent storage or automation history. Keep only what is needed:
- Local travel date, route and configuration revision.
- Calendar gate result (office, home/leave, ambiguous, unavailable), evidence reference and checked time, without copying full events.
- Usual service identity and normalized facts, nearby practical alternatives, source links, source update times and observation times.
- For structured data: service date, timetable UID and TRUST ID when present, requested and resolved stations, scheduled departure, separate departure/arrival state and forecast source, response generated time, feed-health measurement and age, and individual prediction update time or explicitly unknown. Retain the matching destination arrival event when available. No API keys or credentials in checkpoints.
- Journey stage: planned, user-confirmed aboard, tracked-service arrived, user-confirmed arrival, cancelled or unresolved. Preserve explicit service changes and arrival confirmation; never infer boarding from the clock alone.
- Last user-visible alert's facts, recommendation, timestamp and send status.
- Initial status for each local travel date and configured journey: not sent, confirmed delivered, or delivery uncertain; retain the message reference, checked time and facts. A direct status answer counts when it supplies the same verified information. Reset this marker only for a new travel date or genuinely separate configured journey, not a restart, retry, skill upgrade or train switch.
- Unresolved calendar or live-source coverage failures and whether a warning was already sent.
Reset day-specific service facts at the local date boundary. Retain the last alert and unresolved failures long enough to deduplicate and report recovery. Do not put this data in Tessl or another public package.
Send one initial status on the first eligible check with verified service-specific data, even when on time. Verify departure and destination arrival independently, use current dated data and official disruption notices, and do not describe a timetable-only result, missing destination ETA, stale feed or conflicting incident as an on-time all-clear. Give a concise factual status with check time and source. When an active recurring watch has been verified, add that meaningful changes will be reported within its configured monitoring window. Otherwise do not promise future monitoring.
If the first eligible check reveals disruption or a material coverage gap, send that useful warning instead of a separate reassuring status. Count a successfully delivered warning or direct response to a status request as the day's initial status; later verified recovery may still merit an update. Mark delivery only after successful send. If delivery is uncertain, inspect authorized message history before any retry. A status already sent earlier today prevents a duplicate initial message after an upgrade. Explicit status questions still receive an answer, even if unchanged.
After the initial status, compare against the last observation to notice changes, and against the last alert to decide whether another message is useful. Notify for:
- Explicit cancellation or partial cancellation affecting the journey, a skipped destination or material timetable change.
- Expected departure or arrival delay reaching the configured threshold, or a further material deterioration since the last alert. Suggested default: 5 minutes; skip marginal fluctuations that do not change a decision.
- A verified incident that threatens this route and requires a practical choice, even before an individual prediction appears. State uncertainty.
- A better still-reachable alternative or a change that affects when the traveler must leave.
- Loss of an alternative previously recommended, or a genuine recovery that materially changes a previous warning.
- Missing calendar or live-data coverage that could otherwise leave the traveler relying on an unverified watch.
After the single initial status, do not notify for repeated ordinary on-time status, every platform assignment, tiny estimate changes, stale incidents, unchanged warnings or non-office days without an explicitly confirmed active journey. A user's active-journey confirmation takes precedence over a conflicting calendar location until arrival or the configured cutoff. If persistence fails, use recent authorized alert history when available. If both are unavailable, flag the deduplication limit internally and avoid repeated nonurgent messages; still surface a new urgent cancellation or decision with current evidence. Never retry a send with uncertain delivery without checking its outcome.
Acceptance scenarios
Check these before claiming the watch is ready:
- A weekday is Home and no active journey is confirmed: stop before rail queries and send nothing.
- Office plus a morning outOfOffice event titled Commute: continue; do not mistake the travel block for leave.
- Calendar fails: do not infer Office from the usual weekday. Give at most one material coverage warning.
- The target is missing from a filtered board: inspect unfiltered data; do not announce cancellation without evidence.
- Departure is 10 minutes late but arrival prediction is absent: do not invent the arrival. Label it unknown.
- An earlier train is available but station access time is unknown: make the catchability condition explicit.
- The same 10-minute delay appears on consecutive runs: one alert, not one per run. A further delay or lost alternative can warrant another.
- Only cached snippets or a form are readable: status unavailable, never on time.
- A current cancellation appears on the first eligible check: alert immediately; lack of an earlier baseline is no reason to wait.
- A service is restored after an alert: notify only if the verified recovery changes the useful recommendation.
- The scheduler rejects the cadence: report its supported limit; do not multiply tasks or polling loops.
- A colleague installs the public skill: request their private settings; do not use another traveler's calendar or alert history.
- At the normal arrival time, the tracked train's expected arrival slips by 20 minutes: continue within the configured late-arrival window and report the material change. Do not mark complete at the timetable arrival.
- The traveler confirms being on a different service: verify and follow that service. A previous recommendation alone does not establish they took it.
- An intermediate passing-point delay is observed but the destination forecast is absent: report only what is known, without treating that delay as a destination prediction.
- Fresh data confirms destination arrival or the user says they arrived: end the day's journey alerts silently.
- An API journey contains an on-time departure and a three-minute-late destination arrival: preserve both values. Do not substitute departure status for arrival status.
- A destination forecast has source
live_report: label the genuine arrival ETA unknown; the carried-forward movement delay is not a Darwin prediction.
- A fresh response still labels a past arrival time as forecast: arrival remains unconfirmed. Match a destination activity event with
event=arrival, estimate.type=actual, correct dated train and plausible timestamp before completion.
- An activity row is a departure at the destination, or an arrival at the train's later terminus: neither supplies the missing destination arrival time. A limited activity list omitting the train proves nothing.
- The board reports a future
actual timestamp, wrong service date, stale health measurement or conflicting identity: reject that fact and use a fallback or mark coverage unknown.
- The API resolves a requested station to a group or related substation: verify the exact stop serves the configured destination; do not accept a similarly named or different station silently.
- API health is fresh but individual forecast update time is absent: record that age as unknown, describe the check time accurately, and do not claim the forecast itself was updated seconds ago.
- First eligible check verifies on-time departure and destination arrival, with no conflicting disruption: send one concise initial status and, if the recurring watch is active, reassure the traveler that meaningful changes will be reported.
- The next check is unchanged: remain silent. A restart, retry, skill update or train switch does not reset today's initial-status marker.
- The user already received today's verified status in a direct answer: count it as the initial status; do not send another routine reassurance.
- Initial status delivery is uncertain: check message history before retrying; never assume a failed checkpoint means no message was delivered.
- First check has missing arrival ETA, stale data or a conflicting incident: send the relevant uncertainty or warning, never an on-time reassurance. A genuine later recovery is evaluated under the existing material-change rules.