What It Does

MRP AI adds four advanced manufacturing capabilities to Odoo, each implemented as real working software. The finite-capacity scheduler forward-loads your work orders against each work centre's working-time calendar and parallel capacity using a critical-ratio or priority dispatch rule, so planned start/finish dates never exceed real capacity. The forecasting engine trains genuine statistical models on your sales or production history per product. The analytics layer computes OEE and reliability metrics directly from work-order durations and downtime logs. The IoT endpoint ingests and processes machine telemetry end to end.

Finite-Capacity Scheduler

Forward finite-loading packs operations into the real availability windows of each work centre, respecting the number of parallel lanes. Two dispatch rules are built in: Critical Ratio (time-remaining / processing-time, so jobs at risk of being late go first) and Priority. Each run writes planned start/finish back to the work orders and produces a per-work-centre load and bottleneck report. Scope a run to one plant or run it across all sites.

Statistical Demand Forecasting

A dependency-free forecasting engine with three real models: linear-trend regression (ordinary least squares), simple exponential smoothing with automatic smoothing-constant selection, and additive Holt-Winters triple exponential smoothing for seasonal demand. History is aggregated into monthly buckets from your sale.order.line or mrp.production data per product; the engine picks the right model for the data, trains it, projects a horizon and reports MAE / RMSE / MAPE accuracy.

OEE / MTBF / MTTR Analytics

Overall Equipment Effectiveness — Availability × Performance × Quality — is computed per work centre from work-order durations and mrp.workcenter.productivity logs. Reliability metrics MTBF (mean time between failures) and MTTR (mean time to repair) are derived from the failure/repair intervals, and from maintenance.request when the Maintenance app is installed.

Shop-floor IoT Integration

A real HTTP JSON endpoint — POST /mrp_iot/telemetry — ingests machine telemetry into an mrp.iot.reading log and processes it end to end: a fault reading blocks the work centre and opens a downtime record; a recovery reading unblocks it and closes the record. The endpoint is secured by an optional shared-key header, and an MQTT broker can forward messages to it via a thin bridge. Physical hardware is out of scope; the software ingestion and processing pipeline is complete and testable with a simulated payload.

Key Features

• Finite-capacity forward-loading scheduler (critical-ratio / priority rules)
• Per-work-centre parallel capacity and ideal cycle time
• Work-centre load & bottleneck report per scheduling run
• Holt-Winters, linear-regression and exponential-smoothing demand forecasting
• Forecast accuracy metrics (MAE / RMSE / MAPE)
• OEE = Availability × Performance × Quality from real logs
• MTBF / MTTR reliability metrics
• HTTP JSON shop-floor IoT telemetry endpoint with fault/recovery handling
• Multi-plant / multi-warehouse scoping and cross-plant capacity view
• Pure-Python, fully unit-tested calculation engines (no scikit-learn dependency)

Requirements & Notes

Built on the standard Odoo mrp and stock apps; no Enterprise dependency. Maintenance integration activates automatically when the Maintenance app is installed. The forecasting maths use only the Python standard library. Compatible with Odoo 18 & 19.