Agroculture

A context-aware agriculture assistant that provides market prices (Agmarknet), weather insights, proactive alerts, Hindi/English voice I/O, and a clean web UI.

1) Prerequisites

• Python 3.10 or 3.11 (recommended)
• pip installed
• Internet connectivity for APIs and model downloads

2) Installation

# From repository root
python -m venv .venv
# Windows PowerShell
. .venv\Scripts\Activate.ps1
# macOS/Linux
# source .venv/bin/activate

pip install -r requirements.txt

Environment variables (create a .env file in repo root):

AGMARKNET_API_KEY=your_data_gov_in_api_key
MISTRAL_API_KEY=your_mistral_api_key

3) Build the local vector DB (RAG)

Run the RAG script once to build the local knowledge base used by the chatbot’s advisory logic.

python rag.py

This will populate/refresh the agri_db/ directory.

4) Start the backend (FastAPI + Uvicorn)

uvicorn main:app --host 127.0.0.1 --port 8000 --reload

• API docs: http://127.0.0.1:8000/docs
• Useful endpoints:
  • GET /status?user_id=<id>
  • GET /alerts?user_id=<id>
  • POST /alerts/run-now?user_id=<id> (generate alerts immediately)
  • POST /voice/transcribe (ASR; defaults to Hindi)
  • POST /voice/ask (audio in → text + base64 MP3 out)
  • POST /tts (text → base64 MP3; auto en-IN/hi-IN)

5) Open the UI

Open index.html directly in your browser. It connects to the backend at http://127.0.0.1:8000 by default.

Optionally serve it from a local static server:

# Python simple server (example)
python -m http.server 5173
# Open: http://127.0.0.1:5173/index.html

6) Typical workflow

1. Install dependencies: pip install -r requirements.txt
2. Create .env with your keys
3. Build RAG DB: python rag.py
4. Start backend: uvicorn main:app --host 127.0.0.1 --port 8000 --reload
5. Open index.html in your browser

7) Notes & Tips

• Voice input defaults to Hindi recognition to prevent Urdu autodetection. For English, pass ?lang=en to /voice/ask.
• TTS uses Edge voices: en-IN-NeerjaNeural (English), hi-IN-SwaraNeural (Hindi).
• Alerts are generated hourly; use POST /alerts/run-now?user_id=<id> to generate on demand.
• If Whisper install has NumPy/Numba issues, faster-whisper is used automatically.