APMS.ai

Quality deep-dive: which quality PARAMETERS went out of range (by lost time and event count), which MACHINES deviate most, and each machine's first-pass yield from produced-vs-scrap. Answers 'quality problems', 'which parameter is out of spec', 'first pass yield', 'which machine has the worst quality'. Optional: top, critical_only (excludes reserve-band deviations), machine, date_range or start_date/end_date.

quality_analytics

Product SETUP / changeover time — which products or machines take the longest (or shortest) to set up, plant setup totals, and how setup time is trending. Answers 'which products take longest to set up', 'setup time by machine', 'is our changeover time improving', 'setup trend'. This is a DIFFERENT thing from downtime (setup is planned configuration time, not a failure). Optional: by (product|machine, default product), machine (filter one), top, bottom (true for shortest/fastest), trend (true for a time series), bucket (hour|day|week|month).

setup_analytics

Shared-health pools — components shared across several machines, where one counter is consumed by many contributors. Returns each pool, its member machines, which member is consuming it fastest, and which members are paused. Answers 'shared health pools', 'which machine is wearing out the shared component', 'pool status'. Optional: machine, date_range or start_date/end_date.

shared_health_analytics

Benchmark every machine against the plant's OWN median and best performer, with the gap to best, the distance from median, and automatic outlier flagging scaled to the fleet's spread. Answers 'which machines are underperforming', 'how far behind is X', 'what's the gap to our best machine', 'where is the biggest improvement opportunity'. Optional: metric (oee|availability|performance|quality|produced, default oee), machine, date_range or start_date/end_date. If OEE is unavailable (needs ideal cycle time + scrap capture), use metric='availability'.

machine_benchmark

WHY the PRR score is what it is: per-KPI attainment across every scored cycle, ordered WEAKEST FIRST, with average actual vs target. Use for 'why is our score low', 'which KPI is dragging us down', 'where are we losing points', 'KPI performance breakdown'. Optional: top (default 15), machine, date_range or start_date/end_date. Scores only — carries no reward points.

prr_kpi_breakdown

Generate a polished, downloadable PDF from data you ALREADY gathered (MIS / production / OEE / downtime / maintenance / attendance / summary). A branded cover, header and footer are added automatically — you only compose body 'sections', each {type, ...}: 'kpi'{title?, items:[{label,value,sub?}]} (headline cards — lead with these); 'chart'{title?, chart:{type:bar|line|pie, labels, series:[{name,values}]}} (reuse a chart JSON an analytics tool returned — add one for any ranking/trend); 'table'{title?, columns, rows}; 'heading'/'text'/'note'{text}; 'keyvalue'{title?, items:{label:value}}. Show quantities with their unit. Then give the user the 'download_path' as a Markdown link.

generate_pdf_report

Machine reliability: MTBF (mean time between failures), MTTR (mean time to repair), failure count, downtime and availability per machine, ordered LEAST RELIABLE FIRST. Answers 'which machine breaks down most', 'MTBF/MTTR', 'reliability report', 'what should maintenance prioritise'. Machine-raised alarms are excluded so the figures describe reliability, not scheduling. Optional: top, machine, date_range or start_date/end_date.

reliability_metrics

List the machines this user can see with their CURRENT live status (running/idle/stopped/off/maintenance/etc.) and current product. Use for 'what are my machines', 'which are running right now'. Optional 'query' filters by machine name. For rankings/trends/OEE use the analytics capabilities instead.

get_machines

Operator HMI login attendance over a period — total logins and how many were LATE (after the shift's scheduled start time), with each operator's phone number and employee id so the user can follow up/call them. Answers 'who logged in late this month', 'who was late and give me their number', punctuality/attendance. Optional: date_range, grace_minutes (default 5), late_only (default true), top.

operator_attendance

List the operators assigned to this user's machines and whether each is currently logged in. Use for 'who are my operators', 'who is on shift now'.

get_operators

Output PACE — units per hour per machine and plant-wide, plus performance vs the ideal cycle time where configured. Answers 'are we on track today', 'production rate', 'how fast are we running'. Optional: machine, date_range (default today), top.

production_pace

The most recent stoppages (latest first) with machine, reason, duration and time. Answers 'what are the latest stops', 'recent breakdowns', 'why did things stop today'. Optional: machine (name/id; omit for all), date_range (default this_week), top.

recent_stops

Compare THIS period against the immediately-preceding equal-length period — production, run hours, downtime and availability, with the change and % change. Answers 'is this month better than last', 'how does this week compare', period-over-period trend. Optional: date_range (the current period).

period_compare

Performance split by SHIFT over a period — output, run hours, availability, downtime and stops per shift. Answers 'which shift performs best', 'compare shifts', 'shift-wise production/downtime'. Optional: date_range.

shift_performance

CROSS-MODULE analysis — the high-value 'so-what' tool. Joins, per machine over a period, health & quality out-of-range time, downtime + MTBF/MTTR, availability, run load and output, AND computes how they correlate across machines. Use whenever a question links modules: 'how is health affecting downtime/production', 'does running machines harder hurt their health', 'what's driving losses', 'how can we improve OEE', root-cause / improvement questions. Interpret the correlations and the per-machine rows into clear cause-effect and concrete, prioritized improvement actions — never quote raw coefficients. Optional: machine (focus one), date_range, top.

correlation_insights

Export plant data as ROWS for a spreadsheet, BI tool or further analysis — fixed columns plus records, optionally as CSV text. Use this when the user wants the underlying data rather than a summary: 'export', 'give me the raw numbers', 'as CSV', 'download the data', 'I want to put this in Excel'. Required: dataset — one of production, production_trend, oee, downtime, stops, reliability, attendance, machines. Optional: format (json|csv, default json), group_by (for 'production': machine|product|operator|shift), bucket (for 'production_trend': hour|day|week|month), limit, machine, date_range or start_date/end_date. Always check 'truncated' before presenting the result as complete.

export_dataset

WHEN downtime happens — stop minutes and event counts by hour of day or by day of week, with the worst slot called out. Answers 'when do we lose most time', 'is there a pattern to our stoppages', 'which hour/day is worst', 'do stops cluster around shift change'. Optional: grain (hour|weekday, default hour), machine, date_range or start_date/end_date.

stop_pattern_analysis

Forecasts likely stoppages for what is running NOW or planned for TOMORROW (or a date), from each machine's own history: expected stops and downtime in the planned hours, the most likely stop reasons with the chance of each, the hours it usually stops, health parameters out of range now or recently, and maintenance due or overdue, with a high/medium/low risk per machine. Answers 'what stoppages can I expect today/tomorrow', 'which machines are at risk', 'how do I avoid stoppages in the next run'. Use the evidence to recommend preventive actions. Optional: when (now|tomorrow|YYYY-MM-DD, default now), hours (for now, default 8), machine, lookback_days (7-90, default 30), top.

stoppage_forecast

Deep-dive on ONE machine: its current live status + product, and over a period its output, run hours, availability, full time breakdown and most recent stops. Use for 'how is Tube Mill-3 doing', 'status of CNC-12', a single-machine snapshot. Required: machine (name or id); optional: date_range, top (recent stops).

machine_detail

Look up plant configuration / master data: units (plants), shifts with timings, the product catalog, the machine catalog (name/location/brand/type), and the alarm/reason catalog. Answers 'list our units/shifts/products/machines', 'what are the shift timings'. Required: kind (units|shifts|products|machines|alarms); optional: search. NOTE: APMS has no 'production line' or 'department' entity — those map to a machine's location / a unit.

reference_lookup

List the products configured on this user's machines. Optional 'search' filters by product name. Use for 'what products do we run'.

list_products

Machine HEALTH PARAMETERS of every kind — analog sensors (temperature, pressure…), running time since service, manual readings, on/off status signals, counters since reset, and energy/flow meters. view=now (default): each parameter's current value with its unit, its safe/reserve/critical limits, the zone it is in now and since when, and for manual readings when the next entry is due. view=history: over a period, min/avg/max of sensors, running hours used, counts, time in each status, meter consumption, minutes and events in reserve/critical, and resets. Answers 'what is the temperature of machine X now', 'which parameters are out of range', 'show health parameters of X', 'how did spindle load behave this week', 'when was the counter reset'. Optional: machine, parameter (name contains), view (now|history), date_range or start_date/end_date (history, default this_week).

machine_health

Executive plant summary over a period — production output and run hours, live machine status mix, plant availability, time breakdown, and the top products + top downtime causes. Answers 'how is the plant doing', daily/weekly/monthly summary, executive report. Optional: date_range.

summary_analytics

Scrap and raw-material use over a period, with SCRAP PERCENTAGE: scrap ÷ raw material used × 100 in the raw material's unit (e.g. 0.2 kg scrapped of 1 kg used = 20%). Returns the overall scrap %, scrap % per machine or per product (worst first), the top scrap types and raw materials used with their units. A group whose units cannot be compared is reported as not measured, never as 0%. Answers 'what is our scrap %', 'which machine/product has the most scrap', 'scrap by type', 'raw material consumed'. Optional: by (machine|product|scrap_type|raw_material, default machine), machine, top, date_range or start_date/end_date.

material_scrap_analytics

PRR performance score as a timeline — one point per day, week or month. Use for 'is our score improving', 'score trend', 'how did performance move this quarter'. Optional: bucket (day|week|month, default day), machine, date_range or start_date/end_date. Scores only — carries no reward points.

prr_score_trend