Set up local dev environment for the notebooks

[abhishekp106]

Summary

This repo is a collection of standalone data-science / ML Jupyter notebooks (authored for Google Colab). It had no dependency manifest, build, or services. This PR adds the minimal scaffolding needed to set up and run the notebooks locally in JupyterLab.

• requirements.txt — the scientific/ML stack the notebooks import (JupyterLab, numpy, pandas, matplotlib, seaborn, scikit-learn, geopandas, shapely, pydotplus, plotly, einops, fancy_einsum, torch (CPU), transformers, datasets). Colab-only ingestion pieces (google.colab, PyDrive, easy_transformer/pysvelte) are intentionally omitted since they only work inside Colab.
• AGENTS.md — ## Cursor Cloud specific instructions documenting how to install deps, the ~/.local/bin PATH caveat, how to launch JupyterLab / run notebooks headlessly, and which cells are expected to fail locally.
• .gitignore — ignore .ipynb_checkpoints/ and Python caches.

How the environment runs

• Install: pip install --user -r requirements.txt (the configured startup update script).
• Run: jupyter lab --no-browser --ip=0.0.0.0 --port=8888 --ServerApp.token=devtoken → http://localhost:8888/lab?token=devtoken.

Verification

Created a throwaway hello-world notebook (not committed) exercising the core techniques from each repo notebook and ran it in JupyterLab end-to-end: TF-IDF + KMeans clustering (WAFChallenge), a sklearn decision tree at 0.978 accuracy (CIS545FinalProject), a matplotlib figure, and a torch forward pass (Transformer_Exercise). All cells executed with no errors.

jupyterlab_hello_world_demo.mp4

JupyterLab executed outputs (top)
JupyterLab executed outputs (bottom)

_{To show artifacts inline, enable in settings.}