This repository is a simple ReAct-style analytics agent for CSV data. The agent accepts a natural-language business question, chooses tools, records an action log, runs the selected analysis, and returns a final answer.
The current version is intentionally lightweight and deterministic so it is easy to demo in a portfolio. It does not require external APIs to work.
- Load a sales CSV dataset
- Use tools for revenue summaries and product analysis
- Produce a ReAct-style reasoning trace
- Return a clear final answer based on tool outputs
- Run as a CLI demo or Streamlit app
react_data_agent/
├── app.py
├── requirements.txt
├── data/
│ └── sales_data.csv
├── src/
│ ├── agent.py
│ ├── planner.py
│ ├── tools.py
│ └── workflow.py
└── tests/
Install dependencies:
pip install -r requirements.txtRun a CLI demo:
python -m src.workflowRun the app:
streamlit run app.py- Which region has the highest revenue?
- What are the top 5 products by revenue?
- What is the average revenue per order?
- Show revenue by category.