""" knowledge_base.py ------------------ A small in-house knowledge base describing generic intralogistics / smart warehouse concepts (ASRS, AGV/AMR, WMS, conveyor sorting, safety, picking strategy). This is original written content (not scraped from any vendor site) used purely as grounding context for the Retrieval-Augmented Generation (RAG) pipeline that powers the AI Assistant tab. Each entry has: id - short unique key title - human readable title text - the passage used for retrieval + generation grounding category - tag used for the retrieval evaluation set """ KNOWLEDGE_BASE = [ { "id": "asrs_overview", "title": "Automated Storage & Retrieval Systems (AS/RS)", "category": "equipment", "text": ( "An Automated Storage and Retrieval System (AS/RS) uses stacker " "cranes, shuttles, or mini-load systems to automatically place " "and retrieve pallets, totes, or cartons from high-density " "racking. AS/RS units are typically monitored for crane cycle " "time, load/unload faults, and rail or lift motor temperature. " "When a crane reports a fault code, the standard response is to " "pause the aisle, dispatch a technician, and reroute retrieval " "jobs to an adjacent aisle if one is available." ), }, { "id": "agv_amr_overview", "title": "AGVs and Autonomous Mobile Robots (AMR)", "category": "equipment", "text": ( "Automated Guided Vehicles (AGV) follow fixed paths such as " "magnetic tape or wires, while Autonomous Mobile Robots (AMR) " "navigate dynamically using LiDAR and SLAM mapping. Both are " "used to move totes between picking stations, buffer zones, and " "packing lines. Fleet management software assigns tasks based " "on battery level, current queue length, and distance to the " "target station. A vehicle blocked for more than a defined " "timeout is automatically re-routed and flagged for review." ), }, { "id": "wms_overview", "title": "Warehouse Management System (WMS)", "category": "software", "text": ( "A Warehouse Management System (WMS) coordinates inbound " "receiving, put-away, inventory tracking, order picking, " "packing, and outbound shipping. It integrates with a " "Warehouse Control System (WCS) that talks directly to " "conveyors, sorters, and AS/RS controllers in real time. Key " "WMS metrics include inventory accuracy, order cycle time, and " "pick rate (lines picked per hour)." ), }, { "id": "conveyor_sorting", "title": "Conveyor and Sortation Systems", "category": "equipment", "text": ( "Sortation systems such as cross-belt, tilt-tray, or shoe " "sorters route totes and parcels to the correct chute based on " "barcode or RFID reads. Conveyor health is typically monitored " "through motor current, belt speed, and vibration sensors. A " "sudden rise in motor temperature combined with increased " "vibration usually indicates bearing wear or belt misalignment " "and should trigger a maintenance work order before a jam or " "unplanned stoppage occurs." ), }, { "id": "picking_strategies", "title": "Order Picking Strategies", "category": "process", "text": ( "Common picking strategies include discrete picking (one order " "at a time), batch picking (multiple orders in one pass), zone " "picking (pickers assigned to fixed areas), and wave picking " "(orders released in scheduled waves aligned with shipping " "cutoffs). Goods-to-person systems, where an AS/RS or AMR " "brings inventory to a stationary operator, generally reduce " "walking time and increase pick rate compared to person-to-goods " "picking." ), }, { "id": "predictive_maintenance", "title": "Predictive Maintenance in Warehouse Equipment", "category": "maintenance", "text": ( "Predictive maintenance uses sensor data such as motor " "temperature, vibration amplitude, and current draw to detect " "abnormal equipment behaviour before a breakdown occurs. " "Machine-learning models such as Isolation Forest or " "autoencoders are commonly trained on historical sensor " "readings to flag anomalies. Catching a deviation early lets " "a technician schedule maintenance during a planned downtime " "window instead of reacting to an unplanned line stoppage." ), }, { "id": "safety_protocol", "title": "Warehouse Safety Protocols", "category": "safety", "text": ( "Safety protocols in automated warehouses cover pedestrian " "separation from AGV/AMR traffic lanes, lockout-tagout (LOTO) " "procedures before entering an AS/RS aisle, and near-miss " "incident reporting. Any near-miss or safety incident should be " "logged immediately with the location, equipment involved, and " "a brief description, and forwarded to the site safety officer " "the same shift." ), }, { "id": "inventory_accuracy", "title": "Inventory Accuracy and Cycle Counting", "category": "process", "text": ( "Inventory accuracy is usually maintained through cycle " "counting, where a subset of SKUs or storage locations is " "counted on a rolling schedule rather than a single annual " "count. Discrepancies between system quantity and physical " "quantity above a defined tolerance trigger a recount and, if " "confirmed, an inventory adjustment transaction in the WMS." ), }, { "id": "kpi_overview", "title": "Core Warehouse KPIs", "category": "process", "text": ( "Frequently tracked warehouse KPIs include order accuracy, " "on-time shipment rate, dock-to-stock time, pick rate (lines " "per hour), equipment uptime, and inventory turns. Automation " "projects are typically justified using expected improvements " "in throughput, labour cost per order, and space utilisation." ), }, { "id": "energy_efficiency", "title": "Energy Efficiency in Automated Warehouses", "category": "sustainability", "text": ( "Energy use in automated warehouses can be reduced through " "regenerative braking on AS/RS cranes and conveyors, " "variable-frequency drives on motors that scale power to load, " "and scheduling high-throughput operations to avoid peak " "electricity tariff windows." ), }, ]