🛒 Mini Store QA Demo with Ollama + Chroma + FastAPI

This is a simple proof-of-concept app that demonstrates how to build a question-answering system for a product catalog using:

✅ Ollama (local LLM server)
✅ ChromaDB (vector database)
✅ FastAPI (REST API)
✅ Docker Compose (orchestration)

🚀 What does it do?

• Loads a JSON file with ~50 sample products (clothing, accessories, etc.).
• Indexes product information as embeddings in Chroma.
• Exposes a FastAPI endpoint to receive user questions about the products.
• On each question:
  1. Chroma filters products related to the query.
  2. Context of top results is passed to Ollama.
  3. Ollama generates a natural language answer.

All wrapped in a containerized environment so you can spin it up easily.

📂 Project structure

app/
  ├── chroma_utils.py      # Functions for indexing and searching products in Chroma
  ├── main.py              # FastAPI server with endpoints
  ├── ollama_utils.py      # Code to send prompts to Ollama and handle responses
  └── schemas.py           # Pydantic schemas for FastAPI

docker-compose.yml         # Docker orchestration for Ollama + FastAPI
Dockerfile                 # Build FastAPI app container
requirements.txt           # Python dependencies
products.json             # Product dataset (JSON with 50 products)

🐳 How to run it

1. Clone the repo:

   git clone https://github.com/guduchango/mini-store-qa-demo.git
   cd mini-store-qa-demo

2. Build and start the services:

   docker compose up --build

3. Download ia in the ollama container:

   docker exec -it ollama-chroma-v1-ollama-1 /bin/sh
   ollama pull deepseek-llm
   ollama pull llama-2-7b-chat

4. Once the containers are up, FastAPI will be accessible on:

   http://localhost:8001
   

5. Check health endpoint to confirm everything is running:

   curl http://localhost:8001/health

6. Test the /query endpoint (example with curl):

   curl -X POST -H "Content-Type: application/json"    -d '{"ask": "I need comfortable clothes for lounging at home, what do you suggest?"}'    http://localhost:8001/query

📝 Requirements

The Python environment (inside the Docker container) uses:

chromadb
sentence-transformers
requests
fastapi
uvicorn

🪵 Logs

You can see detailed logs for the app and Ollama directly in your terminal when you run:

docker compose up

Logs include:

• Product indexing info.
• Incoming queries.
• Chroma search results.
• Prompts sent to Ollama.
• Responses generated.

✨ Notes

• Ollama is installed as a service in Docker; it serves your local LLM (e.g., TinyLlama).
• Chroma stores vectors in a persistent volume (app_chroma_data) so you don’t need to re-index products each time.

---