NutriBot 🥝 - AI Powered Nutrition Chatbot

Welcome to NutriBot, your go to assistant trained on the nutritional properties of fruits and vegetables. NutriBot excels at answering detailed questions such as “Why does eating potato skin provide more iron than eating potato flesh?” by grounding every response in the source PDF data along with page numbre from that pdf.

🌐 Live Demo

👉 Check out the live version here: 🔗 NutriBot

💬 Try Asking NutriBot

• How does iron deficiency impact human health, including work capacity and immune function?
• Why does eating potato skin provide more iron than eating potato flesh?
• What are some primary dietary sources of iron in the American food supply?
• How are the metabolic fates of copper and iron related?
• What causes fruits to turn brown after we cut them?

NutriBot Data Source

Vicente, A. R., Ortiz, C. M., Sozzi, G. O., Manganaris, G. A., & Crisosto, C. H. (2014). Nutritional properties of fruits and vegetables. In Postharvest Biology and Technology of Fruits, Vegetables, and Flowers (pp. 69–122). Elsevier. https://doi.org/10.1016/B978-0-12-408137-6.00005-3 :contentReference[oaicite:1]{index=1}

Key Features

• AI-Powered Chat: Conversational interface using Groq LLM API integrated via FastAPI, Scalable to n number of PDFs.
• Document Embedding & Search: Uses open source HuggingFace embeddings and Chroma vector store to index and retrieve PDF content.
• RetrievalAugmented Generation (RAG): Combines semantic search from vector database with LLM responses for accurate, context-aware answers.
• Dynamic Loading Animation: Nutrition themed Lottie animations during inference for engaging feedback
• Responsive UI: Built with Next.js and Tailwind CSS, fully mobile responsive and accessible
• Markdown & Links: Bot replies support full Markdown including clickable links via react-markdown and remark-gfm.
• Source Referencing: Every answer includes references with page numbers and direct links to the original document for full transparency.

  Tech Stack

• Backend: Python • FastAPI • Uvicorn • Groq • LangChain • ChromaDB • PyMuPDF
• Frontend: Next.js v15 • React • Tailwind CSS • Framer Motion • Lottie-React
• DevOps: Docker • GitHub • Vercel • Google Cloud Run

🛠️ Project Structure

• Frontend: Next.js deployed on Vercel
• Backend: FastAPI containerized on GCP Cloud Run using Docker
• Vector Database: Chroma for document retrieval
• LLM API: GROQ’s LLaMA 3 model

Getting Started (Local)

Prerequisites

• Python 3.10+
• Node.js ≥ 18 (via NVM or direct install)
• pnpm or npm
• Docker (for containerized setup)

🚀 How to Deploy Locally

1. Clone the Repo

git clone https://github.com/Omkar-Gavali/ai-chatbot.git
cd ai-chatbot

2. Backend Setup

cd backend
python -m venv venv
source venv/bin/activate   # or venv\Scripts\activate on Windows
pip install -r requirements.txt

# Create a .env in backend/:
GROQ_API_KEY=your_groq_api_key_here

3. Run Backend

Place your PDF files in backend/data/ and run:

uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload

Backend: http://localhost:8000

4. Frontend Setup

cd frontend
npm install

# Create .env.local in frontend/:
NEXT_PUBLIC_API_URL=http://localhost:8000
NEXT_PUBLIC_CHAT_ENDPOINT=${NEXT_PUBLIC_API_URL}/chat

5. Run Frontend

npm run dev

Frontend: http://localhost:3000

🐳 Docker & GCP Cloud Run

1. Build Backend Image

cd backend
docker build -t ai-chatbot/backend .

2. Test Locally

cd backend
docker run -p 8000:8000 backend

3. Push to GCR

cd backend
gcloud auth configure-docker
docker tag nutribot-backend gcr.io/your-gcp-project-id/backend
docker push gcr.io/your-gcp-project-id/backend

4. Deploy to Cloud Run

gcloud run deploy nutribot-backend \
  --image gcr.io/your-gcp-project-id/ai-chatbot-backend \
  --platform managed \
  --region europe-west3 \
  --allow-unauthenticated \
  --port 8000

🌍 Vision & Impact

NutriBot is part of a movement:

• Democratizing Knowledge: Free nutritional insights Fruits and Vegetables.
• Responsible AI: Context aware with fall back mechanism to ensure responses are only based on the documents provided .
• Scalable Architecture: Ready for tomorrow’s multi document world.
• Positive AGI Future: AI as a collaborator, not a replacement.

“We must not only imagine a better future with AI. We must build it — with responsibility, passion, and relentless optimism.”

📈 Future Scope

• 📚 Multi-PDF Support: Unlimited document ingestion.

• 🔌 Plugin Ecosystem: Connect to live APIs (recipe databases, health trackers) for real-time personalization.

• 🤖 LLM Choice: Swap models (LLaMA, GPT, Claude) with different parameters(top_k,temperature etc.)

• 🔍 Advanced Retrieval: Hybrid semantic + keyword search.

• 🌐 Localization: Multi-language Q&A.

• ⚙️ Offline LLMs: Onpremise models for privacy.

• 📊 Knowledge Graphs: Enhanced reasoning pipelines.

• 🔐 Authentication: User-specific personalization.

• ⚖️ AI Safety & Ethics: Integrate fairness and consent frameworks,

⚙️ Configuration Files

• requirements.txt: Python deps for backend.
• Dockerfile: Container spec.
• .env.example: Template env vars.

📜 License

MIT License © 2025

AI & AGI Vision

As AI/AGI progresses from pattern recognition to real-world experience learning, systems will autonomously generate data and refine themselves in situations. I foresee NutriBot evolving into a network of domain specific advisors nutrition, skincare, mental health contributing to a future where AI augments human well being at scale, democratizing access to expert guidance.

Contributing

We welcome your ideas, whether it’s adding new PDF loaders, improving UI/UX, or experimenting with embedding models. Please fork, commit, and open a pull request your contributions drive NutriBot toward a healthier, AI-empowered world!

📚 Knowledge Multiplier

"Every PDF you upload expands NutriBot’s ability to provide deeper, more connected nutritional insights. The future of personalized nutrition begins with the documents you share."

Made with 💖 for the Future

If you find this project inspiring, please ⭐ Star on GitHub and share with the world!