This project uses CrewAI + OpenAI + Weaviate to dynamically search a sales database.

Instructions

git clone https://github.com/Anchor-Co-Pilot/agentic-rag-poc.git

pip install -r requirements.txt

Create a .env file

OPENAI_API_KEY=your-openai-api-key
WEAVIATE_URL=your-weaviate-instance-url
WEAVIATE_API_KEY=your-weaviate-api-key

Vector Database Schema (Weaviate)

Collection name: sales_rag_docs

Field	Type
description	text
invoice_no	text
quantity	text
unit_price	text
country	text
return_status	text

🚀 How to Run

1. Run python src/preprocessing/pandas_preprocessing.py to upload data to the vector database and create embeddings

2. Edit tests/test_crew.py

Change the line:

user_input = "Find all electronics in Germany"

to any query you want. Note that for this poc, it only allows one filter at a time

3. Run the crew:

PYTHONPATH=src python tests/test_crew.py

---

✅ Example Output

[
  {
    "description": "Headphones",
    "invoice_no": "231932",
    "return_status": "Returned",
    "unit_price": "29.11",
    "quantity": "49",
    "country": "Germany"
  },
  ...
]

---