A sophisticated multi-agent agricultural intelligence system providing hyperlocalized farming insights for Indian agriculture through orchestration and communication between specialized domain expert agents.
BHUMI AI is a agricultural intelligence platform built on Google's Agent Development Kit (ADK) and Agent-2-Agent (A2A) Protocol that orchestrates specialized AI agents to deliver comprehensive farming insights. The system combines real-time market data, weather intelligence, government schemes, and agricultural expertise through an intelligent routing architecture.
Democratize access to precision agricultural intelligence for Indian farmers through AI-powered multi-agent coordination, enabling data-driven farming decisions at hyperlocal levels.
The platform employs a routing-first architecture with an intelligent orchestrator that coordinates specialized agents based on query analysis and agent capability matching.
- Google Agent Development Kit (ADK): Primary agent framework for building, deploying, and managing AI agents
- Agent-to-Agent (A2A) Orchestration: Google's A2A framework for seamless multi-agent coordination and communication
- Gemini AI Models: Advanced language models for natural language processing and multimodal capabilities
- FastAPI: High-performance API framework for agent communication
- Python 3.12+: Modern Python for robust development
- MongoDB: Conversation persistence and context management
- Pinecone: Vector database for advanced RAG (Retrieval-Augmented Generation)
- CEDA Agmarknet API: Real-time agricultural market data
- Open-Meteo API: Comprehensive weather and soil data
- Agent-to-Agent (A2A) Framework: Google's native A2A orchestration for seamless multi-agent workflows
- HTTP/REST APIs: Inter-agent communication protocol
- Agent Cards: Standardized capability and skill definitions
- Remote Agent Manager: Connection pooling and health monitoring
- Conversation Context: Persistent dialogue and personalization
Role: Intelligent routing coordinator and multimodal interface
Capabilities:
- Multimodal Input Processing: Images, text, and voice through Gemini 2.0 Flash
- Smart Agent Routing: Skill-based agent selection and query delegation
- Agent Discovery: Real-time capability assessment and availability monitoring
- Fallback Intelligence: Personal knowledge as backup when routing isn't feasible
Key Features:
- Agent card-based intelligent routing
- Comprehensive agent capability discovery
- Multimodal agricultural query processing
- Context-aware conversation management
Role: Comprehensive agricultural market intelligence specialist
7 Core Skills:
- Commodity Price Intelligence: 453+ commodities across Indian markets
- Market Discovery Intelligence: Geographic market resolution and discovery
- Supply Chain Analysis: Arrival quantities and supply condition monitoring
- Comparative Market Analysis: Multi-dimensional price and trend comparison
- Intelligent Commodity Resolution: Multilingual search with fuzzy matching
- Contextual Market Intelligence: Conversation memory and personalized insights
- Atomic API Orchestration: Dynamic CEDA API workflow coordination
Technical Arsenal: 26+ specialized tools including tier-1 workflows, market discovery, supply analysis, and atomic data tools.
Role: Agricultural weather intelligence and precision farming advisor
7 Core Skills:
- Comprehensive Farm Conditions Analysis: Complete agricultural dashboard
- Advanced Spraying Analysis & Timing: Pesticide application optimization
- Optimal Planting Window Analysis: Crop-specific timing recommendations
- Precision Irrigation Scheduling: Water management optimization
- Crop Disease Risk Assessment: Weather-based disease prevention
- Agricultural Weather Forecasts: 1-16 day farming-focused forecasts
- Historical Weather Analysis: Trend analysis and agricultural planning
Technical Arsenal: 11 specialized tools for weather data, soil analysis, forecasting, and agricultural optimization.
Role: Government schemes and agricultural policy specialist
7 Core Skills:
- Advanced Agricultural Scheme Discovery: 453+ scheme documents and programs
- Comprehensive Eligibility Analysis: Step-by-step application guidance
- Context-Aware Scheme Intelligence: MongoDB-persisted personalized recommendations
- Advanced Subsidy & Benefits Analysis: Financial impact assessment
- Specialized Farmer Support: Category-specific assistance (women, tribal, small/marginal)
- Real-Time Scheme Updates: Latest launches and policy changes
- Advanced RAG Orchestration: Sophisticated Pinecone vector search and synthesis
Technical Foundation: RAG-powered responses with Pinecone vector database and MongoDB conversation persistence.
The platform leverages Google's Agent-to-Agent (A2A) framework for sophisticated multi-agent coordination, enabling seamless communication, task delegation, and collaborative intelligence across specialized agents.
# Agent-to-Agent Communication Flow via A2A Framework
orchestrator -> agent_discovery() -> capability_matching()
orchestrator -> a2a_send_message(agent_name, query)
specialized_agent -> a2a_process_query() -> return_response()
orchestrator -> synthesize_response() -> user_delivery()- Seamless Integration: Native ADK integration for agent-to-agent communication
- Intelligent Routing: Built-in query routing and agent selection capabilities
- Context Preservation: Automatic conversation context sharing across agents
- Error Handling: Robust failure recovery and graceful degradation
- Performance Optimization: Efficient message passing and resource management
- Agent Definition: YAML-based agent configuration with skill declarations
- Tool Integration: Seamless tool binding and execution
- Conversation Management: Built-in context persistence and memory
- Model Integration: Direct Gemini model integration with advanced capabilities
- Skill Matching: Query analysis against agent skill sets
- Capability Assessment: Real-time agent capability evaluation
- Load Balancing: Intelligent query distribution across agents
- Context Preservation: Conversation continuity across agent interactions
- Real-Time APIs: Live integration with government and weather data sources
- Vector Search: Advanced semantic search through Pinecone
- Conversation Persistence: MongoDB-backed dialogue history
- Caching Strategy: Optimized data retrieval and response times
- Multimodal Support: Process text, images, and voice inputs
- Contextual Understanding: Maintain conversation context across interactions
- Skill-Based Routing: Automatically route queries to appropriate specialists
- Comprehensive Responses: Synthesize insights from multiple data sources
- Real-Time Market Data: Live commodity prices and market conditions
- Precision Weather Intelligence: Hyperlocal weather and soil conditions
- Government Scheme Intelligence: Comprehensive policy and subsidy information
- Historical Analysis: Trend analysis and predictive insights
- RESTful APIs: Standard HTTP/REST interfaces for all agents
- Scalable Architecture: Microservices-based design for horizontal scaling
- Modular Design: Independent agent deployment and updates
- Robust Communication: Reliable inter-agent messaging and coordination
- Python 3.12+
- Google ADK Framework
- MongoDB instance
- Pinecone account
- API keys for
# Start all agents in separate terminals
cd schemes_rag_agent && python __main__.py # Port 10004
cd weather_agent_adk && python __main__.py # Port 10005
cd market_agent_adk && python __main__.py # Port 10006
cd agricultural_orchestrator && python __main__.py # Port 10007- Orchestrator:
http://localhost:10007- Main entry point - Market Agent:
http://localhost:10006- Market intelligence - Weather Agent:
http://localhost:10005- Weather intelligence - Schemes Agent:
http://localhost:10004- Government schemes
- Dynamic Agent Discovery: Real-time capability assessment and routing
- Skill-Based Intelligence: Query matching against detailed agent capabilities
- Multimodal Processing: Image, text, and voice input handling
- Context-Aware Routing: Intelligent agent selection based on conversation history
- Hyperlocal Intelligence: City and district-level agricultural insights
- Multi-Source Integration: Market, weather, and policy data synthesis
- Farmer-Centric Design: Tools and insights tailored for Indian agricultural practices
- Real-Time Decision Support: Live data for immediate farming decisions
- Microservices Architecture: Scalable, maintainable, and resilient system design
- Advanced RAG Implementation: Sophisticated document retrieval and synthesis
- Robust Error Handling: Comprehensive failure recovery and user feedback
- Performance Optimization: Efficient data processing and response generation
- ** Regional Expansion**: Support for additional regional languages and crops
- ** Mobile Integration**: Native mobile applications for field use
- ** Advanced AI Models**: Integration of newer Gemini models and capabilities
- ** Analytics Dashboard**: Comprehensive farming analytics and insights
- ** IoT Integration**: Sensor data integration for precision agriculture
- ** Educational Modules**: Training and educational content for farmers
Project: BHUMI AI Agricultural Intelligence Platform
Framework: Google Agent Development Kit (ADK)
Architecture: Multi-Agent Intelligent Orchestration
Focus: Indian Agricultural Intelligence & Precision Farming
Built with ❤️ for Indian farmers using open-source AI technology.