🧀 Cheese Chatbot Agent

A conversational AI agent for answering questions about cheese products, powered by LLMs, MongoDB, Pinecone vector search, and Streamlit.

Features

• Conversational Search: Ask natural language questions about cheese products.
• Hybrid Search: Combines MongoDB (structured) and Pinecone (semantic vector) search for robust results.
• LLM Reasoning: Uses OpenAI's GPT models for query understanding, MongoDB query generation, and response synthesis.
• Rich Product Data: Supports detailed cheese product information, including brand, price, category, and more.
• Transparent Reasoning: View the agent's step-by-step reasoning process in the UI.

Demo

Getting Started

1. Clone the Repository

git clone <your-repo-url>
cd cheese-chatbot-agent

2. Install Dependencies

pip install -r requirements.txt

3. Prepare Environment Variables

Create a .env file in the project root with the following keys:

MONGO_URI=your_mongodb_connection_string
DB_NAME=your_db_name
COLLECTION_NAME=your_collection_name
PINECONE_API_KEY=your_pinecone_api_key
PINECONE_INDEX_NAME=your_pinecone_index_name
OPENAI_API_KEY=your_openai_api_key

4. Prepare Cheese Product Data

• Place your raw cheese product data in data/cheese_raw.json (array of product objects).

• For Pinecone: Run the Pinecone ingestion script to enrich and embed data:

  python ingestion/pinecone_ingest.py

• For MongoDB: Convert/enrich your data to data/cheese_docs_monggo.jsonl (one JSON per line), then run:

  python ingestion/monggo_ingest.py

5. Run the App

streamlit run streamlit_app.py

The app will be available at http://localhost:8501.

Usage

• Ask questions like:
  • "Show me all sliced cheese under $10"
  • "Which cheeses are most popular?"
  • "How many brands of cheddar do you have?"
  • "Recommend a cheese for a party"
• The agent will display answers, product details, and its reasoning process.

Project Structure

agent/
  langgraph_agent.py      # Main agent logic and workflow
  mongo_search.py         # MongoDB search logic
  pinecone_search.py      # Pinecone vector search logic
data/
  cheese_raw.json         # Raw cheese product data
  cheese_docs_monggo.jsonl # Enriched data for MongoDB
  cheese_docs_pinecone.jsonl # Enriched data for Pinecone
ingestion/
  monggo_ingest.py        # Script to ingest data into MongoDB
  pinecone_ingest.py      # Script to ingest data into Pinecone
streamlit_app.py          # Streamlit web app
requirements.txt          # Python dependencies
graph.png                 # Workflow diagram

Dependencies

• Python 3.8+
• Streamlit
• OpenAI Python SDK
• Pinecone
• PyMongo
• langgraph
• python-dotenv
• tqdm

Install all dependencies with:

pip install -r requirements.txt

Notes

• You need valid API keys for OpenAI and Pinecone, and access to a MongoDB instance.
• The ingestion scripts require your data in the expected format (see data/cheese_raw.json for an example).
• The app expects the Pinecone index and MongoDB collection to be pre-populated.

License

MIT License