Skip to main content

Intent Graph Demo

This example demonstrates an intelligent StateGraph workflow with advanced query routing, parallel execution, and memory management using the modern declarative graph building system.

📊 Workflow Diagram​

🎯 Core Features​

Intelligent Query Routing:

  • LLM-powered intent classification into: general_qa, short_term_trend, macro_trend, or deep_research
  • Dynamic routing based on detected intent and conversation history
  • Context-aware decision making with market context

Parallel Data Processing:

  • Concurrent data fetching across multiple timeframes (15m, 30m, 1h, 4h, daily, weekly)
  • Real-time cryptocurrency data integration
  • Performance optimization through parallel execution

Advanced Memory Management:

  • Persistent conversation context across sessions
  • Automatic storage of learned patterns and market insights
  • State preservation for analysis results and routing decisions

🚀 Key Capabilities​

  • Declarative Graph Building - GraphTemplate, NodeSpec, EdgeSpec for modular workflows
  • High-Level API Integration - HighLevelGraphAPI for automatic parameter inference
  • LLM Integration - Advanced prompt engineering and response processing
  • Tool Orchestration - Multi-source data integration (PowerData, Tavily, EVM swap)
  • Error Handling - Robust recovery with duplicate log prevention
  • Performance Monitoring - Built-in metrics and execution tracking

📋 Prerequisites​

# Required environment variables
export OPENAI_API_KEY="your-openai-api-key"
export TAVILY_API_KEY="your-tavily-api-key"       # Search engine

🏃 Quick Start​

# Navigate to examples directory
cd spoon-cookbook/example

# Install dependencies
pip install -r requirements.txt

# Run the declarative intent graph demo
python intent_graph_demo.py

🔍 What to Observe​

Architecture:

  • How GraphTemplate and NodeSpec simplify workflow construction
  • HighLevelGraphAPI automatically inferring parameters from queries
  • Modular node implementations with better separation of concerns

Execution Flow:

  • Intelligent routing to appropriate analysis paths based on query intent
  • Parallel data fetching across multiple timeframes
  • Memory loading and updates throughout the process

Performance:

  • Execution times for different routing paths
  • Parallel vs sequential processing performance
  • Memory usage optimization and duplicate log prevention

Advanced Behaviors:

  • LLM-powered routing decisions based on intent analysis
  • Real-time data integration from multiple sources
  • Context maintenance across complex workflows

📁 Source Code​

  • Main Example: intent_graph_demo.py
  • Supporting Modules:
    • spoon_ai/graph/ - Core graph system and declarative builders
    • spoon_ai/graph/builder.py - High-level API and parameter inference
    • Graph System Docs

🎓 Learning Outcomes​

  • Using declarative graph building (GraphTemplate, NodeSpec, EdgeSpec)
  • Leveraging HighLevelGraphAPI for automatic parameter inference
  • Implementing modular, maintainable node functions
  • Advanced LLM integration and prompt engineering
  • Parallel processing for performance optimization
  • Memory management in long-running processes
  • Error handling and recovery strategies

💡 Best Practices​

  • Declarative architecture for improved modularity
  • High-level API usage for automatic parameter inference
  • Scalable design for easy extension
  • Resource-efficient implementation
  • Maintainable, well-documented code