AgriSakha AI is a voice-first prototype that helps farmers access agricultural advice, credit scoring, and soil analysis — all in Hindi. Designed to work on basic feature phones without requiring smartphones or digital literacy.

• Hindi Speech Recognition using Vosk ASR (vosk-model-small-hi-0.22)
• Hindi Text-to-Speech using Coqui TTS (tts_models/hi/indic-tts)
• Handles common farmer queries: pest control, loans, soil testing
• Responds in Hindi via keyword matching with actionable advice
• USSD-style menu: 1.कर्ज 2.कीट 3.मिट्टी

• engine.py — Satellite-data-weighted credit scoring engine
  • Factors: Crop health (60%), Soil moisture (20%), Acreage (20%)
  • Mocks ISRO Bhuvan satellite API for field data
  • Returns score + Hindi voice response (e.g., "क्रेडिट स्कोर: 78")
• blockchain.py — Custom SHA-256 blockchain for tamper-proof credit records
  • Genesis block initialization, block addition, and chain validation
  • Each credit score event is stored as an immutable on-chain record

• soil.py — Soil parameter analysis engine
  • Inputs: Nitrogen, Phosphorus, Potassium, pH, Organic Carbon
  • Weighted model calculates fertilizer recommendation score
  • Generates Hindi report with nutrient status (कम / सामान्य / अधिक)
  • Recommends Urea dosage in kg/acre

• Interactive UI to input soil parameters and generate instant Hindi reports
• Run locally to demo the soil analysis pipeline
• Outputs Hindi fertilizer recommendations and numeric recommendation score

AgriSakha/
├── streamlit_app.py          # Streamlit web UI for soil analysis
├── config.py                 # Central config (TTS model, ASR model, thresholds)
├── voice_ai/
│   ├── core.py               # VoiceProcessor: record → transcribe → respond → speak
│   └── requirements.txt      # TTS, Vosk, sounddevice, pydub, librosa
├── credit_scoring/
│   ├── engine.py             # CreditEngine: satellite-based scoring + blockchain storage
│   └── blockchain.py         # AgriBlockchain: SHA-256 chain with validation
├── edge_computing/
│   ├── soil.py               # SoilAnalyzer: NPK/pH analysis + Hindi report generation
│   └── requirements.txt      # numpy, etc.
├── data/
│   ├── sample_queries.json   # Hindi/English test queries + USSD commands
│   ├── soil_samples.json     # Sample soil data for testing
│   ├── NFHS_5_Factsheets_Data.xls
│   └── RS_Session_246_AU1132_1.1.csv
├── docs/
│   ├── ARCHITECTURE.md
│   ├── INSTALL.md
│   └── DEMO.md
└── tests/
    ├── test_voice_ai.py
    └── test_credit_scoring.py

• Python 3.8+

git clone https://github.com/ankush850/AgriSakha.git
cd AgriSakha

cd voice_ai
pip install -r requirements.txt

cd ..
pip install streamlit numpy
streamlit run streamlit_app.py

python credit_scoring/engine.py

python voice_ai/core.py

python -m pytest tests/

For detailed setup, see docs/INSTALL.md.

class Config:
    TTS_MODEL = "tts_models/hi/indic-tts"
    ASR_MODEL = "models/vosk-model-small-hi-0.22"
    NODE_ADDRESS = "127.0.0.1:5000"      # Blockchain node
    SOIL_THRESHOLDS = {
        'nitrogen':   (30, 50),           # mg/kg
        'phosphorus': (15, 40),
        'potassium':  (20, 50)
    }

• AI4Bharat for Hindi NLP tools and IndicTTS support.
• Alphacephei (Vosk) for lightweight offline Hindi ASR.
• ISRO Bhuvan for satellite data API inspiration.
• The/Nudge Institute for the Pragati AI for Impact Hackathon platform.