agrichat-annam

🌱 AgriChat-Annam: Agricultural RAG Chatbot

A Retrieval-Augmented Generation (RAG) chatbot for answering agricultural queries using local LLMs (Ollama), ChromaDB for vector search, and a FastAPI web interface.

---

Table of Contents

• Features
• Project Structure
• Setup Instructions
  • 1. Install Ollama
  • 2. Download LLM and Embedding Models
  • 3. Install Python Dependencies
  • 4. Prepare Data
  • 5. Build the Chroma Vector Database
  • 6. Start the Web Application
• Code Explanation
  • Creating the Vector Database
  • Query Handling Logic
  • Web Interface (FastAPI)
• Usage
• Troubleshooting
• Contributing

---

Features

• Local LLMs: Uses Ollama to run models like Gemma, Llama locally
• Vector Search: Stores and retrieves agricultural Q&A using ChromaDB and embeddings
• Web Interface: User-friendly chat interface built with FastAPI and Jinja2 templates
• Privacy-Preserving: All data and models run locally

---

Project Structure

agrichat-annam/
├── .env                        # Environment config (used in backend)
├── docker-compose.yml         # Runs frontend and backend together
├── render.yaml                # Render deployment settings
├── README.md
│
├── agrichat-backend/          # FastAPI Backend with RAG pipelines
│   ├── app.py
│   ├── backendRequirements.txt
│   └── RAGpipelinev3/         # Main RAG pipeline (embedding, retrieval, LLM)
│       ├── main.py
│       ├── creating_database.py
│       ├── ollama_embedding.py
│       ├── chromaDb/
│       ├── Data/
│       └── Gemini_based_processing/   # Gemini-based RAG variant
│           ├── main.py
│           ├── creating_database.py
│           └── chromaDb/
│
├── agrichat-frontend/         # Static frontend (HTML, JS, CSS)
│   ├── index.html
│   ├── script.js
│   ├── style.css
│   └── Dockerfile
│
├── nginx/                     # NGINX config for containerized frontend
│   └── default.conf

Setup Instructions

1. Install Ollama

Linux: curl -fsSL https://ollama.com/install.sh | sh

text

Windows/Mac:
Download installer from ollama.ai

Verify Installation: ollama --version

text

---

2. Download LLM and Embedding Models

ollama pull gemma:1b ollama pull nomic-embed-text

text

---

3. Install Python Dependencies

pip install langchain chromadb pandas fastapi uvicorn python-multipart markdown

text

---

4. Prepare Data

1. Place your Q&A CSV in the data/ directory
2. Ensure CSV has columns: questions and answers

---

5. Build the Chroma Vector Database

python creating_database.py

text

---

6. Start the Web Application

uvicorn app:app --reload --port 8000

text Visit http://localhost:8000

---

Code Explanation

Creating the Vector Database (creating_database.py)

• Converts CSV Q&A data into vector embeddings
• Uses OllamaEmbeddings wrapper for text embeddings
• Stores documents in ChromaDB using ChromaDBBuilder

Query Handling Logic (main.py)

• ChromaQueryHandler class handles:
  • Vector similarity search
  • Result reranking
  • LLM prompt construction
  • Response generation via Ollama API

Web Interface (FastAPI) (app.py)

• FastAPI server with two endpoints:
  • GET /: Renders chat interface
  • POST /query: Processes user questions
• Converts markdown responses to HTML

---

Usage

1. Start Ollama server: ollama serve

text

2. Launch application(from agrichat-backend directory): uvicorn app:app --reload --port 8000

text

3. Access chat interface at http://localhost:8000

---

Troubleshooting

Issue	Solution
Ollama not running	Run ollama serve in separate terminal
Model not found	Verify with ollama list and pull required models
ChromaDB path errors	Use absolute paths for Windows systems
Low memory errors	Use smaller models like gemma:1b

---

Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Open a Pull Request

---