	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Crop Advisory Agent

Crop Advisory Agent

Author: Venkata Sudhakar

India has over 140 million farming households. Access to timely, localised crop advisory is one of the biggest challenges in Indian agriculture. A Crop Advisory Agent can act as a digital agronomist - recommending which crops to sow based on soil type and season, suggesting fertiliser quantities, flagging pest risks, and advising on irrigation schedules. Built with ADK and Gemini, it can be deployed as a WhatsApp chatbot or mobile app for rural farmers.

This tutorial builds a Crop Advisory Agent using Google ADK and Gemini 2.0 Flash. The agent takes the farmer location, soil type, and current season as inputs, recommends suitable crops with expected yield, and provides a cultivation plan.

The below example shows the Crop Advisory Agent advising a farmer in Nashik, Maharashtra on Rabi season crop selection.

from google.adk.agents import Agent
from google.adk.tools import FunctionTool
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.genai.types import Content, Part

# Crop database by soil type and season
CROP_DB = {
    ("black", "rabi"): [
        {"crop": "Wheat", "duration_days": 120, "yield_qtl_per_acre": 18,
         "water_req": "medium", "msp_per_qtl": 2275},
        {"crop": "Chickpea", "duration_days": 100, "yield_qtl_per_acre": 8,
         "water_req": "low", "msp_per_qtl": 5440},
        {"crop": "Mustard", "duration_days": 110, "yield_qtl_per_acre": 7,
         "water_req": "low", "msp_per_qtl": 5650},
    ],
    ("red", "kharif"): [
        {"crop": "Groundnut", "duration_days": 130, "yield_qtl_per_acre": 10,
         "water_req": "medium", "msp_per_qtl": 6377},
        {"crop": "Soybean", "duration_days": 100, "yield_qtl_per_acre": 12,
         "water_req": "medium", "msp_per_qtl": 4892},
    ],
    ("alluvial", "rabi"): [
        {"crop": "Potato", "duration_days": 90, "yield_qtl_per_acre": 80,
         "water_req": "high", "msp_per_qtl": 400},
        {"crop": "Wheat", "duration_days": 120, "yield_qtl_per_acre": 22,
         "water_req": "medium", "msp_per_qtl": 2275},
    ],
}

def get_crop_recommendations(soil_type: str, season: str,
                              land_acres: float, district: str) -> dict:
    """Recommend crops based on soil type and season."""
    key = (soil_type.lower(), season.lower())
    crops = CROP_DB.get(key, [])
    if not crops:
        return {"error": "No data for soil type " + soil_type + " in " + season + " season"}
    recommendations = []
    for c in crops:
        expected_yield = c["yield_qtl_per_acre"] * land_acres
        expected_revenue = expected_yield * c["msp_per_qtl"]
        recommendations.append({
            "crop": c["crop"],
            "duration_days": c["duration_days"],
            "expected_yield_qtl": round(expected_yield, 1),
            "expected_revenue_rs": round(expected_revenue),
            "water_requirement": c["water_req"],
            "msp_per_qtl_rs": c["msp_per_qtl"]
        })
    recommendations.sort(key=lambda x: x["expected_revenue_rs"], reverse=True)
    return {
        "district": district,
        "soil_type": soil_type,
        "season": season,
        "land_acres": land_acres,
        "top_crops": recommendations
    }

def get_cultivation_plan(crop: str, land_acres: float) -> dict:
    """Return fertiliser and irrigation plan for a crop."""
    plans = {
        "Wheat": {
            "seed_kg_per_acre": 40,
            "fertiliser": "DAP 50kg/acre at sowing, Urea 30kg/acre at 21 days",
            "irrigations": "4 irrigations: CRI, tillering, jointing, grain filling",
            "pest_watch": "Aphids and yellow rust - spray Mancozeb if spotted"
        },
        "Chickpea": {
            "seed_kg_per_acre": 30,
            "fertiliser": "DAP 25kg/acre at sowing only - chickpea fixes nitrogen",
            "irrigations": "1-2 irrigations only - drought tolerant",
            "pest_watch": "Pod borer - use Bt spray at flowering stage"
        },
        "Mustard": {
            "seed_kg_per_acre": 1.5,
            "fertiliser": "DAP 40kg/acre + Sulphur 10kg/acre at sowing",
            "irrigations": "2 irrigations: branching and pod formation",
            "pest_watch": "Aphids - spray Imidacloprid at stem elongation"
        }
    }
    plan = plans.get(crop, {"fertiliser": "Consult local KVK", "irrigations": "As needed"})
    return {"crop": crop, "land_acres": land_acres, "cultivation_plan": plan}

crop_tool = FunctionTool(func=get_crop_recommendations)
plan_tool = FunctionTool(func=get_cultivation_plan)

crop_agent = Agent(
    model="gemini-2.0-flash",
    name="crop_advisory_agent",
    instruction="""
        You are a Crop Advisory Agent helping Indian farmers.
        When given farmer details:
        1. Get crop recommendations for their soil type and season
        2. Get the cultivation plan for the top recommended crop
        3. Provide a clear advisory report:
           - Top 3 crop options with expected yield and revenue at MSP
           - Best crop recommendation with reason
           - Cultivation plan: seed rate, fertiliser, irrigation, pest watch
           - Simple language - farmer-friendly advice
        Use practical, actionable language. Mention government MSP prices.
    """,
    tools=[crop_tool, plan_tool]
)

session_service = InMemorySessionService()
runner = Runner(agent=crop_agent, app_name="crop_advisory", session_service=session_service)

session = session_service.create_session(app_name="crop_advisory", user_id="farmer")
prompt = """
    Farmer details:
    Name: Ramkisan Patil
    District: Nashik, Maharashtra
    Soil Type: Black (cotton soil)
    Season: Rabi (Oct-Nov sowing)
    Land: 3 acres
    Water source: Borewell (limited supply)
    What should I grow this Rabi season?
"""
message = Content(role="user", parts=[Part(text=prompt)])
for event in runner.run(user_id="farmer", session_id=session.id, new_message=message):
    if event.is_final_response():
        print(event.content.parts[0].text)

It gives the following output,

CROP ADVISORY - Ramkisan Patil, Nashik
Rabi Season | Black Soil | 3 Acres | Limited Water
===============================================

Top Crop Options
Crop      Duration  Yield (3 ac)  Revenue at MSP  Water
Wheat     120 days  54 qtl        Rs 1,22,850     Medium
Chickpea  100 days  24 qtl        Rs 1,30,560     LOW
Mustard   110 days  21 qtl        Rs 1,18,650     LOW

Recommendation: CHICKPEA
Given your borewell water limitation, Chickpea is the best choice.
It needs only 1-2 waterings, fixes its own nitrogen (lower fertiliser
cost), and gives the highest revenue at MSP Rs 5,440/qtl.

Cultivation Plan for Chickpea (3 acres)
  Seed Rate   : 30 kg/acre = 90 kg total
  Fertiliser  : DAP 25 kg/acre at sowing (chickpea fixes nitrogen)
                Total: 75 kg DAP
  Irrigation  : 1-2 waterings only (drought tolerant)
  Pest Watch  : Pod borer - use Bt spray at flowering stage

Expected Income: Rs 1,30,560 at government MSP
Tip: Register on e-NAM portal to get better prices beyond MSP.

The Crop Advisory Agent can be deployed as a WhatsApp chatbot using the Gemini Live API so farmers can ask questions in Hindi or Marathi. Extend it to incorporate real-time weather forecasts from the IMD API, soil health card data, and local mandi prices so the recommendation is always current. This is a high-impact application for agri-fintech startups and state government digital agriculture initiatives.

Send your comments, suggestions or queries regarding this site to [email protected].