Krishi Sahayak

AI Agriculture Assistant for Indian Farmers

Live Demo → llm-agri-bot.onrender.com

An AI-powered agriculture chatbot that helps Indian farmers with crop advice, pest control, soil health, weather, and government schemes — using text, voice, and image analysis.

---

Features

Feature	Description
Text Chat	Ask any agriculture question and get expert answers
Image Diagnosis	Upload a crop photo → AI identifies diseases, pests & deficiencies
Voice Input	Speak in English, Hindi, or Hinglish via Groq Whisper
Voice Output	Bot reads answers aloud using Groq Orpheus TTS
Conversation Memory	Remembers your chat context (Redis, with in-memory fallback)
Prompt Caching	50% cost savings — cached prefixes across requests
Dark/Light Theme	Glassmorphism UI with one-click theme toggle
Multilingual	Responds in English, Hindi, or Hinglish

---

Tech Stack

Backend	Python 3.11+, Flask, Groq SDK
Frontend	HTML5, CSS3 (Glassmorphism), jQuery
LLM	openai/gpt-oss-120b (text), meta-llama/llama-4-scout-17b-16e-instruct (vision)
Speech	Groq Whisper whisper-large-v3-turbo (STT), Groq Orpheus canopylabs/orpheus-v1-english (TTS)
Memory	Redis (with automatic in-memory fallback)
Deploy	Render, Docker, Gunicorn

---

Quick Start

Prerequisites

• Python 3.11+
• uv package manager
• Groq API key (free tier available)

1. Clone & Install

git clone https://github.com/mohammed97ashraf/LLM_Agri_Bot.git
cd LLM_Agri_Bot
uv sync

2. Configure

cp .env.example LLM_Agri_Bot/.env

Edit LLM_Agri_Bot/.env and add your Groq API key:

GROQ_API_KEY=gsk_your_key_here

3. Run

uv run python LLM_Agri_Bot/run.py

Open http://127.0.0.1:5000

---

Environment Variables

Copy .env.example to LLM_Agri_Bot/.env and configure:

Variable	Required	Default	Description
GROQ_API_KEY	Yes	—	Your Groq API key (get one)
LLM_MODEL	No	openai/gpt-oss-120b	Text LLM model
LLM_VISION_MODEL	No	meta-llama/llama-4-scout-17b-16e-instruct	Vision LLM model
LLM_TEMPERATURE	No	0.3	Model temperature (0–2)
LLM_MAX_TOKENS	No	2048	Max response tokens
STT_MODEL	No	whisper-large-v3-turbo	Speech-to-text model
TTS_MODEL	No	canopylabs/orpheus-v1-english	Text-to-speech model
TTS_VOICE	No	autumn	TTS voice name
REDIS_HOST	No	localhost	Redis host (optional — falls back to memory)
REDIS_PORT	No	6379	Redis port
REDIS_SSL	No	false	Enable Redis SSL
FLASK_SECRET_KEY	Yes*	dev-secret-key	Flask session secret (*required in production)
FLASK_DEBUG	No	true	Enable debug mode

---

Project Structure

LLM_Agri_Bot/
├── app/
│   ├── __init__.py             # App factory (create_app)
│   ├── config.py               # Environment-based configuration
│   ├── routes/
│   │   ├── main.py             # Index page, robots.txt, sitemap, llms.txt
│   │   └── chat.py             # Chat API (text, voice, image)
│   ├── services/
│   │   ├── llm_service.py      # Groq LLM + vision + prompt caching
│   │   ├── memory_service.py   # Redis + in-memory fallback
│   │   ├── stt_service.py      # Groq Whisper STT
│   │   ├── tts_service.py      # Groq Orpheus TTS
│   │   └── prompt_manager.py   # XML + CoT system prompt
│   ├── static/
│   │   ├── css/style.css       # Glassmorphism UI (dark/light)
│   │   ├── js/chat.js          # Chat logic, image upload, voice
│   │   └── images/             # Favicon
│   └── templates/
│       └── index.html          # Main template (SEO + JSON-LD)
├── Sample_image/               # Screenshots for README
├── llms.txt                    # AI crawler disclosure
├── .env.example                # Environment template
├── gunicorn.conf.py            # Production Gunicorn config
├── Dockerfile                  # Docker deployment
├── render.yaml                 # Render blueprint
├── run.py                      # Dev entry point
├── pyproject.toml              # uv / project config
└── requirements.txt            # pip fallback

---

Deployment

Render (Recommended)

1. Push your code to GitHub
2. Go to render.com → New → Web Service
3. Connect your GitHub repo
4. Render auto-detects render.yaml and Dockerfile
5. Add your GROQ_API_KEY (and other env vars) in the Render dashboard
6. Click Deploy

Docker

docker build -t krishi-sahayak .
docker run -p 10000:10000 --env-file LLM_Agri_Bot/.env krishi-sahayak

---

API Reference

Method	Endpoint	Description
GET	/	Chat interface
POST	/chat	Send text, image, or audio — returns AI response
POST	/chat/clear	Clear conversation history
GET	/health	Health check (Redis status)
GET	/robots.txt	Search engine crawl rules
GET	/sitemap.xml	XML sitemap
GET	/llms.txt	AI crawler disclosure
GET	/.well-known/llms.txt	AI crawler disclosure (well-known path)

---

How It Works

User sends message (text / image / voice)
        │
        ▼
┌─────────────────────────────────────────┐
│  Flask Backend                          │
│  ├── Text? → Groq LLM (gpt-oss-120b)  │
│  ├── Image? → Llama 4 Scout (vision)   │
│  └── Voice? → Whisper STT → LLM → TTS  │
│                                         │
│  Memory: Redis (or in-memory fallback)  │
│  Cache:  Groq automatic prompt caching  │
└─────────────────────────────────────────┘
        │
        ▼
Response with text + optional voice audio

---

Contributing

Contributions welcome! See CONTRIBUTING.md.

git checkout -b feature/your-feature
uv sync
# make changes
git commit -m "Add your feature"
git push origin feature/your-feature

---

License

MIT License — see LICENSE

---

Built with care for Indian farmers

By Mohammed Ashraf · GitHub · LinkedIn