Check SAP order capacity with an AI prompt

Task: read SAP orders through the MCP Gateway, model them as a project in Rillsoft Project, let Rillsoft calculate the capacity and deliver a report. SAP is read only; in Rillsoft you work on a copy of the pool and restore the initial state at the end. 1. Questions (once, before any call) - Kind: production (production orders with operations) or maintenance (maintenance orders with persons per operation). - Plant, period and optionally work centers or line. Suggest a narrower cut if more than about 300 operations are expected (roughly four Rillsoft calls per operation). - Grid of the analysis: hour, day or week. Bottlenecks within a shift need the hour; on day or week peaks even out. - If the SAP dates lie in the past: shift them by whole weeks? Then keep weekday and time of day and state the shift in every note and in the report. 2. Read SAP (GET tools of the sap-gateway server only; parameters without "$": filter, select, top ≤ 500, skip, orderby) - Production: operations (…ProductionOr_1) with order, operation, text, work center, earliest start and end, standard values 1–3 and their units, quantity; orders (…_A_ProductionOrde) with material and quantity. - Maintenance: operations (…MaintenanceOrderOperat) with order, operation, text, work center, NumberOfCapacities (persons), duration, work, start and end; orders (…_MaintenanceOrder) with text, type, priority, technical object. - Work centers (…A_WorkCenters) and their capacities (…WorkCenterCapacity): capacity category (001 machine, 002 person), number of capacities, shift start/end in seconds, break, factory calendar. - Plant and period in the filter; avoid long OR lists (the iFlow answers HTTP 500), rather read wider and cut yourself. On HTTP 500 retry twice with a pause. - Check the units (minutes or hours) before you calculate. 3. Show the mapping and have it confirmed (create nothing yet) - Order → subproject "Order ‹no› · ‹text or material› · ‹quantity›"; operation → task "‹op no› ‹text› (‹work center›)", note with all SAP values and the original SAP date. - Work center with capacity category 001 → machine type on the task (demand), as many machines of this type in the pool as SAP states capacities (supply). Never attach a concrete machine to overlapping tasks. - Work center with capacity category 002 → role on the task with the number of persons (NumberOfCapacities), as many persons of this role in the pool as SAP states. SAP only delivers the head count: create persons as placeholders ("‹work center› Technician 01") and mark them as Rillsoft data in the report – or use existing persons of the pool if I name them. - Labour time in production (standard value 3) → role from the pool with utilisation = labour time ÷ duration. - Shift → calendar Mon–Fri without a break window. SAP spreads the break evenly: duration = net × shift ÷ (shift − break), in whole minutes. Check the factor against the SAP dates per work center: median of span ÷ duration over the operations within one shift. If it is 1 (SAP plans there without break share), use factor 1 for this work center; state both values in the mapping. - Duration: production = setup + quantity × (machine + labour); maintenance = duration, with duration 0 and work > 0 duration = work ÷ persons (mark it). Duration 0 without work → milestone without machine type/role. - Sequence of operations → finish-to-start dependency only where SAP plans the operations one after another; otherwise starts as in SAP. - State the number of orders, operations, work centers, persons/machines and the expected number of Rillsoft calls. 4. Create in Rillsoft (only after my confirmation) - Save the initial state: rillsoft_resource_pool_get (path, number of persons and roles), rillsoft_project_tree_get. If the open project is not empty or has unsaved changes: ask. - Copy the pool file to a place I name or confirm. rillsoft_project_create without a name (discardChanges: true), rillsoft_resource_pool_file_select on the copy, rillsoft_resource_pool_edit_open, create calendars, teams, roles, machine types, machines, persons, edit_save, edit_close. Take the IDs from the answers (the catalogue shows new entries only after edit_save). - rillsoft_project_update: name with origin (and shift), start/finish, timeStepMinutes 1, durationStepUnit and quantityStepUnit "minute", schedulingMode "deadline_driven", project calendar 7 × 24 h (persons use their own calendar, machines the project calendar – see 6). - Per order rillsoft_subproject_create; per operation rillsoft_task_create (fixedType "duration", note), rillsoft_task_update with the calendarId of the shift, then start and durationHours; machine type via rillsoft_task_machine_role_assignment_set (quantity 1), role via rillsoft_task_role_assignment_set (quantity = persons); dependencies as in 3. - Then per subproject rillsoft_subproject_update with start = earliest operation start and finish = latest operation finish from rillsoft_project_tree_get (like the subproject context command "Take over", which takes the dates from the activities). A new subproject is fixed and otherwise keeps the finish of the project. The operation dates must not change. 5. Reconciliation (before any analysis; if it fails, no report but a message with the figures) - Number of operations SAP = Rillsoft. - Total time: production setup + machine time, maintenance work (duration ÷ factor × persons); deviation at most the minute rounding (½ min × persons per operation). - Start per operation as in SAP (shifted); list deviations individually (e.g. SAP start before the shift). - Start and finish per subproject = earliest start and latest finish of its operations. 6. Let Rillsoft calculate the capacity (Rillsoft calculates, not you) - Save the view first (rillsoft_view_state_get). - Machines: rillsoft_view_state_apply with view {name: "machine_capacity", grouping: ["machine_role"], display: "absolute", capacity: true} and a time filter on the period. Machine supply follows the project calendar: per shift model set the shift calendar as project calendar and read the machine types of this shift from this report. Check that the dates stay the same. - Personnel: view {name: "employee_capacity", grouping: ["role"], display: "absolute", capacity: true}, same time filter. Persons follow their own calendar; one report is enough. - rillsoft_timescale_apply with the chosen grid (for bottlenecks also "hour"), then rillsoft_analysis_report_get. Values are minutes (unit "minute"): divide by 60. Demand = value, supply = sum of offers[].value. - Shortfall = sum of max(0, demand − supply) per time slot, separately for hour, day and week. - Concurrency from the Rillsoft dates: highest number of machines or persons needed at the same time per work center and the periods in which it exceeds the available number. - Afterwards reset the project calendar and the view, do not save. 7. Report (complete) - Header: kind, plant, period, origin of the data (SAP / Rillsoft), shift of the dates, number of calls. - Key findings (three to five, from figures only): largest bottleneck with work center, day and time. - Table per machine type or role: work center, machines/persons from SAP, shift, demand, supply, utilisation, maximum at the same time, shortfall per hour, day and week. - Table of the days or periods with shortfall. State the scale (hour, day or week) with every shortfall; the program's shortfall column calculates to the minute and can exceed the hourly report - say so when you quote it. - Reconciliation with all figures from step 5. - Optional rescheduling: only if I want it, marked as a proposal; Rillsoft has no capacity levelling with rescheduling over MCP – proposal per work center without overlap, never before the SAP start, keep the order sequence, then check in Rillsoft (shortfall 0). Write nothing back to SAP. - Data basis: fields, filters, units, factors, assumptions. 8. Finish - On my request save the project as .rpj next to the pool copy. - rillsoft_project_create without a name, rillsoft_resource_pool_file_select on the original path, check: path, number of persons and roles as before. Same on an abort along the way (edit_close first). - At the end, list which parts were delivered: questions, SAP data, mapping, creation, reconciliation, capacity per report, report with all tables, initial state. What is missing, with the reason.