SourChef is an intelligent, RAG-enhanced chatbot designed for nutrition-focused food recommendations, personalized meal planning, dietary filtering, calorie/macronutrient tracking, and YouTube video recipe suggestions. Built with Gemini Pro (via LangChain), Pinecone, Langfuse, Gradio, and PostgreSQL.

• Information Overload & Confusion: 68% of people report being confused about nutrition advice due to conflicting information
• Lack of Personalization: people often find that generic "one-size-fits-all" nutrition advice, like following a fixed calorie count or macronutrient ratio, doesn't work for them
• Recipe-Nutrition Disconnect: people can't easily connect recipes to their nutritional goals

• Conversational Intelligence: User can get recommends personalized, health-aligned food and recipes using memory-augmented retrieval and reasoning.
• Retrieval-Augmented Generation (RAG) Implementation from vectorized food, nutrition, and recipe knowledge base via Pinecone vectorstore.
• Dietary filtering: Chatbot can act like a helpful "personal diet planner" (e.g. vegan, keto, gluten-free).
• Calorie & macronutrient calculator: Chabot use nutrition and calorie from external source to prepare your nutrition and recipe data before jumping into recommendation.
• YouTube recipe video suggestions: Videos displayed alongside recipe responses.
• LLM monitoring: Traces, tags, and logs every user query and LLM response.

coming-soon

coming-soon

• Frontend: Gradio
• Backend: Python (FastAPI)
• Embedding model: Gemini API/models/embedding-001
• LLM model: Gemini API/models/gemini-2.0-flash
• AI Framework: LangChain
• Relational Database: PostgreSQL
• Vector Database: Pinecone via docker
• LLM Monitoring: Langfuse via docker
• Chat Memory: ConversationBufferMemory
• Containerization: Docker
• Deployment: Gradio
• CI/CD Pipeline: Github actions
• Package Manager: uv
• Python Linting: ruff

• Obtain API key from Google-AI-Studio
• Obtain API key from Pinecone-key
• Obtain API key from YouTube-Data-API-v3
• Obtain API key from Nutritionix-API
• Obtain API key from Langfuse-API

git git@github.com:abliskan/rag-chatbot-backend.git
cd rag-chatbot-backend

conda create --name <environment_name> python=<python_version>
conda activate <env_name>
pip install uv
uv venv
.venv\Scripts\activate
uv pip install python-dotenv langchain langchain-community langchain-pinecone langchain-google-genai langchainhub langfuse pandas gradio requests google-api-python-client
uv pip install --dev pytest
uv pip list

touch .env

# insert the API Key
GOOGLE_API_KEY=your_google_api_key
PINECONE_API_KEY=your_pinecone_key
PINECONE_ENVIRONMENT=your_pinecone_env
PINECONE_INDEX_NAME=your_pinecone_index
YOUTUBE_API_KEY=your_serpapi_key
NUTRITIONIX_APP_ID=your_nutritionix_app_id
NUTRITIONIX_API_KEY=your_nutritionix_key
LANGFUSE_PUBLIC_KEY=your_langfuse_public_key
LANGFUSE_SECRET_KEY=your_langfuse_secret_key
LANGFUSE_HOST=https://app.langfuse.com

Run the gradio from your local computer

python main.py