Agents from Scratch

This repository contains implementations of LangGraph agents for various business use cases, including email assistance, sales monitoring, and demand forecasting.

Features

• Email Assistant: Intelligent email management with Gmail integration
• Sales Monitor Agent: Zoho Inventory integration for sales tracking and inventory management
• Demand Forecast Agent: Advanced demand forecasting with predictive analytics
• Restock Trigger Agent: Automated supplier ordering and procurement management (NEW!)

Demand Forecast Agent

The Demand Forecast Agent is a sophisticated forecasting system that predicts future demand based on inventory changes in Zoho. It uses the same human-in-the-loop and memory architecture as the email assistant template.

Key Capabilities:

• Multi-method Forecasting: Moving average, exponential smoothing, and hybrid approaches
• Stockout Risk Analysis: Early warning system for inventory shortages
• Seasonal Pattern Detection: Automatic identification of seasonal demand cycles
• Intelligent Reorder Recommendations: Cost-optimized purchase suggestions
• Human-in-the-Loop: Approval workflows for critical forecasting decisions
• Memory Learning: Adapts forecasting methods based on accuracy and user preferences

Forecasting Tools:

• analyze_demand_patterns_tool: Historical demand pattern analysis
• forecast_demand_tool: Future demand prediction with multiple methods
• analyze_stockout_risk_tool: Risk assessment across all inventory items
• generate_reorder_recommendations_tool: Intelligent purchase planning
• seasonal_demand_analysis_tool: Seasonal trend analysis and insights

Usage Example:

from src.email_assistant.demand_forecast_agent_hitl_memory import demand_forecast_agent
from src.email_assistant.demand_forecast_utils import create_stockout_risk_trigger

# Create a stockout risk trigger
trigger = create_stockout_risk_trigger(
    item_name="USB Cable",
    current_stock=5,
    daily_sales_rate=15.2
)

# Run forecasting analysis
result = demand_forecast_agent.invoke({
    "forecast_trigger": trigger
})

See examples/demand_forecast_example.py for comprehensive usage examples.

Restock Trigger Agent

The Restock Trigger Agent automates supplier ordering and procurement workflows, managing the entire purchase order lifecycle from supplier research to order fulfillment. It uses the same human-in-the-loop and memory architecture as the other agents.

Key Capabilities:

• Automated Supplier Sourcing: Research and compare suppliers with pricing, lead times, and ratings
• Purchase Order Management: Create, approve, track, and manage purchase orders end-to-end
• Cost Optimization: Bulk pricing calculations, supplier comparison, and budget management
• Supplier Performance Tracking: Monitor delivery rates, quality metrics, and relationship history
• Human-in-the-Loop: Approval workflows for large orders and critical procurement decisions
• Memory Learning: Adapts supplier selection and approval thresholds based on past performance

Procurement Tools:

• find_suppliers_tool: Research suppliers with pricing and lead times
• create_purchase_order_tool: Generate purchase orders with selected suppliers
• check_order_status_tool: Track purchase order progress and delivery
• approve_purchase_order_tool: Approve orders and send to suppliers
• cancel_purchase_order_tool: Cancel orders with proper notifications
• get_supplier_performance_tool: Evaluate supplier metrics and ratings
• bulk_restock_tool: Plan and cost-optimize multiple item orders

Usage Example:

from src.email_assistant.restock_agent_hitl_memory import restock_trigger_agent
from src.email_assistant.restock_utils import create_stockout_alert_trigger

# Create a critical stockout alert
trigger = create_stockout_alert_trigger(
    item_name="USB Cable",
    current_stock=2,
    reorder_level=25,
    daily_consumption=15.2
)

# Run procurement automation
result = restock_trigger_agent.invoke({
    "restock_trigger": trigger
})

See examples/restock_trigger_example.py for comprehensive usage examples.

Installation

1. Clone the repository
2. Install dependencies: pip install -r requirements.txt
3. Set up environment variables for API integrations (Zoho, Gmail, etc.)
4. Run examples to test functionality

Architecture

All agents follow a consistent pattern:

• Triage System: Classifies and routes incoming requests
• Human-in-the-Loop: Critical decision approval workflows
• Memory System: Learns from user interactions and preferences
• Tool Integration: Modular tools for specific business functions

Contributing

Feel free to contribute additional agents, tools, or improvements following the established patterns.