Krishi-Sahayak AI (meaning Agricultural Assistant in Sanskrit/Hindi) is a state-of-the-art interactive decision-support system built for the Google Solution Challenge 2026. Designed specifically to bridge the digital divide for smallholder farmers in rural India, the platform translates cutting-edge generative artificial intelligence into actionable, localized agricultural intelligence.
According to agricultural statistics, smallholder farmers represent over 80% of India's farming population, yet they frequently lack access to real-time, language-friendly expert advice regarding volatile weather, soil compatibility, crop selection, and fluctuating market prices. Krishi-Sahayak AI directly solves this by providing a hyper-local, multilingual assistant that can be queried naturally in local dialects and can read answers aloud for maximum accessibility.
This application is built from the ground up to solve critical global issues aligned with the following UN Sustainable Development Goals:
| Goal | Application Focus | Impact Mechanics |
|---|---|---|
| 🟢 SDG 2: Zero Hunger | Sustainable Agriculture & Higher Yields | Prevents crop failure by providing real-time soil compatibility checks, pest diagnostics, and climate-resilient farming techniques. |
| 🔵 SDG 10: Reduced Inequalities | Accessibility via Local Dialects | Combats illiteracy and digital exclusion by supporting 6 major regional Indian languages with full Text-to-Speech (TTS) voice playbacks. |
| 🟠 SDG 13: Climate Action | Weather-Resilient Farm Planning | Delivers 5-day hyper-local forecasts and dynamically generates climate mitigation strategies based on immediate weather warnings. |
- Folium Map Integration: Farmers or local community workers can pinpoint their farm locations directly on an interactive map.
- Metadata Persistence: The application captures the exact GPS coordinates (Latitude & Longitude), soil types, and farm size, storing them locally inside a validated profile system (
Data.csv).
- LangChain Integration: The application coordinates human queries with structured system instructions using
ChatGoogleGenerativeAI. - Zero-Syllable Context Blending: Behind the scenes, the model aggregates farmer-specific metadata (soil type, region) and current local climate metrics before querying the LLM, producing advice tailored specifically to that farm instead of generic tips.
- OpenWeatherMap Integration: Leverages coordinates to pull real-time, 5-day / 3-hour local forecasts.
- Auto-Alert Mechanism: Triggers active alerts for high temperatures, drought signs, strong winds, and heavy rainfall to prompt early harvest or soil shielding.
- Predictive Trends: Provides simulated price forecasting charts per quintal over the upcoming 7 days for key Indian staple crops (Wheat, Rice, Cotton, Tomato, Maize).
- Financial Decision Support: Identifies short-term upward or downward market trends to help farmers decide whether to sell immediately or store their harvest.
- Accessibility Engine: Integrated with
gTTSto translate Gemini's written guidelines into natural spoken audio. - Full Multi-Language Playback: Supports instant audio playback in native accents for all 6 regional languages.
Both the Streamlit User Interface and the Gemini AI response pipeline natively support complete localized translation for:
- 🇬🇧 English (en)
- 🇮🇳 Hindi (hi) (हिंदी)
- 🇮🇳 Tamil (ta) (தமிழ்)
- 🇮🇳 Bengali (bn) (বাংলা)
- 🇮🇳 Telugu (te) (తెలుగు)
- 🇮🇳 Marathi (mr) (मराठी)
%%{init: {'theme': 'base', 'themeVariables': { 'primaryColor': '#E8F5E9', 'edgeLabelBackground':'#FFFFFF', 'tertiaryColor': '#FFF8E1', 'lineColor': '#2E7D32', 'nodeBorder': '#2E7D32' }}}%%
flowchart TD
classDef frontend fill:#E8F5E9,stroke:#2E7D32,stroke-width:2px,color:#1B5E20;
classDef database fill:#E3F2FD,stroke:#1565C0,stroke-width:2px,color:#0D47A1;
classDef ai fill:#EDE7F6,stroke:#6A1B9A,stroke-width:2px,color:#4A148C;
classDef ext fill:#FFF3E0,stroke:#E65100,stroke-width:2px,color:#E65100;
classDef voice fill:#FCE4EC,stroke:#C2185B,stroke-width:2px,color:#880E4F;
A[🌾 Farmer / Local Assistant] -->|Interacts & Submits Query| B(💻 Streamlit Responsive UI)
B -->|Selects Location| C[🗺️ Folium Interactive Map]
B -->|Triggers GPS Coordinates| D[🌤️ OpenWeatherMap API]
B -->|Reads/Writes Profiles| E[(🗄️ Local Profile Data.csv)]
D -->|Real-Time Forecast Feed| F[⚙️ Context Aggregator]
E -->|Active Farmer Profile Context| F
B -->|Retrieves Chat History| F
F -->|Formatted LangChain System Message| G[🧠 Google Gemini AI Engine]
G -->|Tailored Agricultural Advice| B
B -->|Converts Text Advice to Audio| H[🔊 gTTS Audio Synthesizer]
H -->|Plays Regional Speech Accents| A
class A,B,C frontend;
class E,F database;
class G ai;
class D ext;
class H voice;
Make sure you have Python 3.9 to 3.11 installed on your operating system.
-
Clone the Repository:
git clone https://github.com/Advait251206/Google-Solution-Challenge.git cd Google-Solution-Challenge -
Create and Activate a Virtual Environment:
- Windows:
python -m venv venv .\venv\Scripts\Activate.ps1 - macOS/Linux:
python3 -m venv venv source venv/bin/activate
- Windows:
-
Install Core Dependencies:
pip install -r requirements.txt
-
Setup Environment Variables: Create a
.envfile in the root directory:# API Keys Configuration GEMINI_API_KEY=your_google_gemini_api_key_here WEATHER_API_KEY=your_openweathermap_api_key_here # Operational Log Settings LOG_LEVEL=INFO
-
Run the Application:
streamlit run app.py
├── .gitignore # Standardized Python, Streamlit & credential ignore rules
├── README.md # Breathtaking competition submission documentation
├── app.py # Central Python logic containing UI, translations & API orchestration
├── requirements.txt # Standardized libraries declaration (langchain, streamlit, folium, etc.)
├── Data.csv # Local persistent farmer profile registry (Generated on runtime)
└── Log.csv # Historical QA interaction database (Generated on runtime)
- Sensitive Key Exclusions: Standardized Git tracking ignores local
.envand all credential-bearing texts. - Data Privacy: Farmer location mapping coordinates and profiles (
Data.csv) remain completely decentralized and stored locally on the client's host system.