Chat-Based Data Analysis on Multiple CSVs

How the system works

User can input multiple messages regarding a query and input csvs. Assistant can respond in 3 modes: text, visualization and tables. Gemini's api is used to get response in structured output format. Assistant can return response in tables mode with an expression for pd.Dataframe.query() or it can return details required to build a chart/graph using plotly in visualization mode.

How to set up and run the project

cd interview-chatbot-problem
python -m venv .deploy
source .deploy/bin/activate
pip install -r requirments.txt
streamlit run frontend.py

Libraries used and why

Libraries	Why
pandas, numpy	Dataframe management, cleaning, analysis
plotly	Generate charts
streamlit	Frontend
google-genai	Communicating with Gemini

How evaluation could be performed

Write integration and Regression tests for:

• Testing functionality with improper csvs: null values, incorrect dtypes, different formatting for datetime
• Introduce SQL in output and check against output's response with desired sql qury.