From 0548e4ba8c435440f2700168e66f4ba1e7695c59 Mon Sep 17 00:00:00 2001
From: Cursor Agent <cursoragent@cursor.com>
Date: Tue, 16 Jun 2026 23:10:00 +0000
Subject: [PATCH] Add dev environment setup: requirements.txt, AGENTS.md,
 .gitignore

Co-authored-by: Abhishek Pandya <abhishekp106@users.noreply.github.com>
---
 .gitignore       |  4 ++++
 AGENTS.md        | 33 +++++++++++++++++++++++++++++++++
 requirements.txt | 34 ++++++++++++++++++++++++++++++++++
 3 files changed, 71 insertions(+)
 create mode 100644 .gitignore
 create mode 100644 AGENTS.md
 create mode 100644 requirements.txt

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..e3e31ca
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,4 @@
+.ipynb_checkpoints/
+__pycache__/
+*.pyc
+.venv/
diff --git a/AGENTS.md b/AGENTS.md
new file mode 100644
index 0000000..d169b6b
--- /dev/null
+++ b/AGENTS.md
@@ -0,0 +1,33 @@
+# AGENTS.md
+
+This repository is a collection of standalone data-science / ML **Jupyter notebooks**
+(originally authored for Google Colab). There is no server, database, or build step —
+the "application" is JupyterLab plus the scientific Python stack.
+
+Notebooks:
+- `CIS545FinalProject.ipynb` — EDA + sklearn models on the US Accidents dataset (pandas, geopandas, sklearn).
+- `Transformer_Exercise.ipynb` — implement GPT-2 from scratch (torch, transformers, einops).
+- `WAFChallenge.ipynb` — TF-IDF + KMeans clustering on a YouTube videos CSV (sklearn).
+
+## Cursor Cloud specific instructions
+
+- Dependencies are installed with `pip install --user -r requirements.txt` (handled by the
+  startup update script). Packages land in `~/.local`.
+- The Jupyter CLIs (`jupyter`, `jupyter-lab`) live in `~/.local/bin`, which is **not on PATH**
+  by default. Prefix commands with `export PATH="$HOME/.local/bin:$PATH"` or invoke via
+  `python3 -m jupyterlab` / `python3 -m jupyter`.
+- Run JupyterLab with:
+  `jupyter lab --no-browser --ip=0.0.0.0 --port=8888 --ServerApp.token=devtoken`
+  then open `http://localhost:8888/lab?token=devtoken`.
+- Execute a notebook headlessly (good for CI-style checks / quick validation):
+  `jupyter nbconvert --to notebook --execute --ExecutePreprocessor.timeout=120 <file>.ipynb`
+- `requirements.txt` intentionally **omits** the Colab-only pieces that cannot run outside
+  Colab: `google.colab`, `PyDrive`, `oauth2client`, `google_drive_downloader`, and the GitHub
+  research packages `easy_transformer` / `pysvelte`. Cells that import those (used only to pull
+  input CSVs from Google Drive, or for the interpretability visualizations) will fail locally;
+  this is expected. Supply the input data files locally to run the data-dependent cells.
+- The notebooks target newer library versions than Colab pinned; some legacy calls (e.g.
+  `from sklearn.externals.six import StringIO` in `CIS545FinalProject.ipynb`) are removed in
+  current sklearn and will error. This is a notebook code issue, not an environment problem.
+- No GPU is available; PyTorch is the CPU build (`torch==*+cpu`). The Transformer notebook runs
+  on CPU but slowly.
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..542f097
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,34 @@
+# Python dependencies for the Jupyter notebooks in this repository.
+# These notebooks were authored for Google Colab; this file lets you run the
+# scientific/ML portions locally in JupyterLab. Google Colab-only pieces
+# (google.colab, PyDrive, Google Drive data ingestion) are intentionally
+# omitted since they only work inside Colab. See AGENTS.md for details.
+
+# Use CPU-only PyTorch wheels (no GPU in the dev environment).
+--extra-index-url https://download.pytorch.org/whl/cpu
+
+# Jupyter
+jupyterlab
+notebook
+ipykernel
+
+# Data-science core (all notebooks)
+numpy
+pandas
+matplotlib
+seaborn
+scikit-learn
+
+# Geospatial / decision-tree viz (CIS545FinalProject.ipynb)
+geopandas
+shapely
+pydotplus
+
+# Transformer exercise (Transformer_Exercise.ipynb)
+torch
+transformers
+datasets
+einops
+fancy_einsum
+tqdm
+plotly