DesiDiet AI is an enterprise-grade nutritional ecosystem designed specifically for the Bangladeshi population. It leverages a modern GraphRAG (Graph-based Retrieval-Augmented Generation) architecture to deliver highly personalized dietary advice grounded in the National Dietary Guidelines (NDG) Bangladesh 2025.
DesiDiet AI follows the Infinity AI Buildfest 2026 AI-Native Application Blueprint. It consists of three primary layers:
- Data Layer: Neo4j Knowledge Graph (370+ local foods & NDG rules) + Prisma SQL (User profiles & logs).
- Reasoning Layer (FastAPI): Python-based GraphRAG engine that performs semantic retrieval and dietary constraint checking.
- Presentation Layer (React): A premium, bilingual dashboard for real-time AI assistance and health tracking.
For a comprehensive breakdown of our 8-layer enterprise architecture, please see the System Architecture Documentation.
Run the Neo4j graph database locally using Docker:
docker run -d --name neo4j-khadok \
-p 7474:7474 -p 7687:7687 \
-e NEO4J_AUTH=neo4j/khadok2025 \
neo4j:5.12cd graphRAG
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# Ingest data into Neo4j
python3 preprocessing/clean_csv.py
python3 build_graph.py
# Start the API server
uvicorn app.main:app --reloadThe frontend is a modern React application.
cd frontend
npm install
npm run devDetailed instructions are in the Frontend README.
The system utilizes a hybrid model to link nutritional science with user data.
- IBW & Macro Targeting: Automatic calculation of Ideal Body Weight and caloric goals.
- Constraint Propagation: The Graph engine traverses from
MedicalCondition→DietaryRule→FoodGroupto filter unsafe items. - LLM Reasoning: Integrates with local/remote LLMs to synthesize culturally accurate meal suggestions.
DesiDiet AI is an AI-powered assistant and not a medical device. Full details are available in our Conditions Page.
Developed with for the people of Bangladesh.