InsightForge is a portfolio-ready AI/ML agent project that turns a raw CSV/XLSX dataset and a natural-language business question into a structured analysis report.
The project demonstrates agent-style decomposition: dataset profiling, cleaning, analysis planning, insight generation, visualization, and report writing.
- Upload or pass a CSV/XLSX dataset
- Automatic data profiling: rows, columns, missing values, duplicates, numeric/categorical/date fields
- Data cleaning: normalized columns, duplicate removal, missing value handling, date inference
- Natural-language question planning
- Automated metric and dimension selection
- Trend analysis over time
- Segment/driver analysis
- Chart generation with matplotlib
- Markdown executive report export
- CLI and Streamlit app entry points
- Unit tests for planner, cleaner, and end-to-end agent flow
User Question + Dataset
|
v
Dataset Profiler -> Data Cleaner -> Analysis Planner
| |
v v
Insight Generator -> Chart Builder -> Report Writer
|
v
Markdown Report + PNG Charts + Follow-up Ready Insights
- Python
- pandas / numpy
- matplotlib
- Streamlit
- pytest
Optional extension points:
- OpenAI/Claude/Gemini for narrative generation
- FAISS/ChromaDB for RAG over uploaded reports
- FastAPI for API deployment
- Docker for reproducible execution
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txtRun the CLI:
python -m insightforge.cli data/sample_sales.csv --question "What are the main drivers of revenue performance over time?"Run the Streamlit UI:
streamlit run app.pyRun tests:
pytestFor the included sample sales dataset, InsightForge produces:
- Dataset profile
- Revenue overview
- Top-performing region/product/channel
- Monthly revenue trend
- Driver insight
- Markdown report in
reports/insight_report.md - PNG charts in
reports/