drilldown
ChatGPTThe individual records behind a number — each row carries name, deliverables (what was asked for, e.g. Water Bottle, Towel), area/asset, guest, crew, status, scheduled/SLA, timings and last note. subject = tasks or sops. kind examples — tasks: all, completed, onTime, late, open, pending, overdue, escalated, incidents, guest, status, group, priority, type, area, name, deliverable, guestName; sops: all, completed, open, abandoned, flagged, status, group, area, name. value is the bucket for status/group/priority/type/area/name.
fetch
ChatGPTFetch one record by id (task ids look like 2026-08-25-0001, SOP submission ids likewise). Returns the full record as text.
get_google_reviews
ChatGPTGoogle rating, review count, star split, rating trend and the latest reviews for a site (only when the Google Reviews add-on is enabled).
get_guest_voice
ChatGPTGuest requests for a site and period (KPIs, per day/hour, by department, top requests/deliverables/rooms, repeat requests) plus the repeat guests currently in-house with their visit history summary and requests in the period.
get_members_live
ChatGPTCrew right now: headcount, by department, and who is clocked in (name, department, designation, since when).
get_overview
ChatGPTHeadline totals for tasks and SOP submissions in a period: total, completed, late, open, overdue, escalated, completion/on-time rates. Org-wide or one site.
get_repeat_guest_history
ChatGPTA returning guest's full stay history (every stay with room and dates) and the requests raised in each stay. Identify by name or mobile.
get_sop_submission
ChatGPTOne SOP submission in full: checkpoints with answers and flags, activity trail, timings.
get_sops_summary
ChatGPTSOP compliance for a site and period: submissions, completed, open, abandoned, flagged, by group, top SOPs, top flagged, top abandoned, repeat flags.
get_task
ChatGPTOne task in full: details, assignees, timings, deliverables and the progress-note trail. Use the id from a drilldown or search.
get_tasks_summary
ChatGPTTask analytics for a site and period: KPIs, by group / priority / type, top requests, top incidents, top maintenance, and repeat incidents (same area + same issue 2+ times). Honours the area scope.
list_sites
ChatGPTList the properties (sites) of this organisation with their ids. Call this first to get a siteId.
query_task_history
ChatGPTTask history analytics (from 2026-07-01 onward, refreshed hourly). PREFER THIS over get_tasks_summary/drilldown for ANY question about patterns over time, at any date range, including ranges under 3 months. It is the ONLY tool that can answer: which areas/rooms/assets have REPEATED problems (minCount), and WHEN each was LAST reported (last_seen) and FIRST reported (first_seen). get_tasks_summary/drilldown cannot count recurrence or return a last-reported date, and cap at 92 days. Trigger words: regularly, repeatedly, recurring, again, keeps happening, frequent, repeat offenders, chronic, over the last N months, since <month>, trend, history, when was it last. For complaints/problems/faults filter taskType:"incident" (taskType:"service" is routine requested work like amenities and cleaning, which will otherwise dominate the counts). IMPORTANT - many task names are CONTAINERS and the real request is in the DELIVERABLE: "Housekeeping Amenities" alone covers 52 different items (Water Bottle, Towel, Dental Kit, Slippers...). So whenever the question is about WHAT guests or staff asked for, or about a named item, add "deliverable" to groupBy (e.g. groupBy ["area","deliverable"]) or filter deliverable:"Towel" - grouping by task alone will report one meaningless row. Deliverable groups count items, not tasks, and exclude tasks that name no item. Omit groupBy to list individual task records; set groupBy to aggregate. Example — "rooms with issues regularly since August, and when it last happened": start "2026-08-01", groupBy ["area","task"], filters {taskType:"incident"}, minCount 3 -> read last_seen for the most recent date. Add stillOpen true for "still unresolved". For repeat-guest questions group by guest_id: it is an OPAQUE hashed key (never a phone number or email) that still groups the same guest exactly across stays, and each row carries a readable guest name to quote in the answer - report the name, never the guest_id itself. Phone/email are only recorded for tasks raised from 2026-10-17 onward, so for earlier periods group by guest instead and say that repeat counts are name-matched and may merge namesakes.
search
ChatGPTSearch tasks and SOP submissions by text (task/SOP name, guest name, area, group) in a period. Returns ids for fetch.
