genpark-inventory-forecasting-skill
GenPark AI Agent Skill -- # Inventory Demand Forecasting Skill
This repository contains the Inventory Demand Forecasting Skill — an agent skill configuration interface (skill.json), developer Python SDK (forecaster.py), and executable validation scripts designed to predict product stock-out dates, compute reorder trigger windows, and auto-draft supplier purchase orders.
🚀 Capabilities
- Smoothed Velocity Engine: Calculates average daily sales velocity from 30-day history while automatically filtering out promotional spike outliers (e.g. flash sale days above 3x average).
- Stockout & Reorder Projection: Maps current stock to a projected depletion date, then back-calculates the last safe purchase order date based on supplier lead time + safety buffer.
- Automated PO Drafting: Generates structured ERP-ready purchase order payloads (PO number, SKU, quantity, estimated delivery date, unit cost, total cost) for immediate use in procurement systems.
🛠️ Setup & Installation
-
Install dependencies:
pip install -r requirements.txt -
Configuration: Set your API environment variable if connecting to a live forecasting backend. Otherwise, the client auto-activates in mock mode:
- PowerShell:
$env:FORECAST_API_KEY="your_api_key" - bash:
export FORECAST_API_KEY="your_api_key"
- PowerShell:
💻 SDK Usage Reference
from forecaster import InventoryForecastingClient
client = InventoryForecastingClient()
sales_history = [
{"date": "2026-06-01", "units_sold": 20},
{"date": "2026-06-02", "units_sold": 18},
# ... 30 days of data
]
result = client.forecast(
sku="SKU-001",
current_stock=450,
sales_history=sales_history,
lead_time_days=14,
safety_stock_days=7
)
print(f"Stockout Date: {result['stockout_date']}")
print(f"Reorder By: {result['reorder_trigger_date']}")
print(f"Order Qty: {result['suggested_order_quantity']} units")
# Generate purchase order draft
po = client.build_purchase_order(
sku="SKU-001",
qty=result['suggested_order_quantity'],
reorder_date=result['reorder_trigger_date'],
lead_time_days=14,
unit_cost=12.50
)
📜 License
This project is licensed under the MIT License.