Great Learning Agent

Automates downloading weekly Jupyter notebooks from a Great Learning course, converting them to clean .py scripts, and generating .md summaries. Run it manually any time — it detects and processes any Thursday release that has not been handled yet.

---

What it does

1. Reads config/schedule.yaml to find the most recent unprocessed Thursday release.
2. Opens a real (non-headless) Chromium browser using your saved session.
3. Navigates to your course page, locates the notebook for that week, and downloads it.
4. Converts the .ipynb to a clean Python script (no cell markers, no magic commands).
5. Generates a Markdown summary with an overview, per-cell walkthrough, and TODO exercises.
6. Records the week as processed so it is never downloaded again.

---

Directory layout

greatlearning-agent/
  config/
    schedule.yaml          # course release schedule — edit to match your program
  data/
    auth/                  # Playwright session (gitignored — created by login_once.py)
    downloads/raw_ipynb/   # raw .ipynb files downloaded from the course
    scripts/py/            # converted .py scripts  ← OUTPUT
    summaries/             # generated .md summaries ← OUTPUT
    logs/                  # debug screenshots and HTML dumps
    state/
      processed.json       # tracks which weeks have been processed
  src/
    run_release_check.py   # main entry point
    login_once.py          # one-time login to save a browser session
    config.py
    auth.py
    discovery.py
    downloader.py
    converter.py
    analyzer.py
    state_manager.py
    utils.py

Output files land in:

Type	Path	Naming convention
Raw notebook	data/downloads/raw_ipynb/	original filename from course
Python script	data/scripts/py/	YYYY-MM-DD_week<N>_<topic-slug>.py
MD summary	data/summaries/	YYYY-MM-DD_week<N>_<topic-slug>.md

---

Prerequisites

• Python 3.12
• A Great Learning account with access to the course
• macOS (tested) or Linux

---

Setup

1 — Clone and create a virtual environment

git clone https://github.com/YOUR_USERNAME/greatlearning-agent.git
cd greatlearning-agent
python3.12 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
playwright install chromium

2 — Create .env

Create a file named .env in the project root:

COURSE_URL=https://olympus.mygreatlearning.com/courses/XXXXX?th=YYYY&pb_id=ZZZZ

How to find your URL:

1. Log into olympus.mygreatlearning.com.
2. Open your course (e.g. "Certificate Program in Agentic AI").
3. Copy the full URL from the browser address bar. It looks like:

   https://olympus.mygreatlearning.com/courses/148723?th=dpep&pb_id=20415
   

   Paste that exact URL as the value of COURSE_URL.

3 — Update config/schedule.yaml

The file ships with the April 2026 Agentic AI cohort schedule. If your cohort has different dates, edit the release_date fields (format: YYYY-MM-DD) to match your Thursdays. Mark break weeks with break: true and project weeks that share a date with a content week with skip: true.

4 — Log in once

source .venv/bin/activate
python src/login_once.py

A browser window opens. Log in with your Great Learning credentials. Once redirected away from the login page the session is saved automatically to data/auth/storage_state.json. You will not need to do this again unless the session expires (typically after a few weeks).

---

Running

source .venv/bin/activate
python src/run_release_check.py

Run this any day. The script finds the most recent Thursday release that has not been processed yet, downloads it, and writes the outputs. If everything is up to date it exits immediately with a log message.

Session expired?

If you see Session expired — redirected to login, re-run the login step:

python src/login_once.py

---

Output examples

Python script (data/scripts/py/2026-04-22_week2_prompt-engineering-fundamentals.py):

• Plain .py with no # %% cell markers, no %/! magic lines
• Header block with source title, week number, and download timestamp
• TODO exercises preserved verbatim

Markdown summary (data/summaries/2026-04-22_week2_prompt-engineering-fundamentals.md):

• Overview and learning objectives
• Per-cell walkthrough table
• List of TODO exercises with line references
• Required pip packages
• VS Code run instructions

---

Troubleshooting

Symptom	Fix
Browser opens and immediately closes	Session file is missing — run python src/login_once.py
No release scheduled for today	No unprocessed Thursday found — check data/state/processed.json and config/schedule.yaml
Notebook not found on course page	Content may not be released yet; re-run after the Thursday release time
403 or page blocked	The script uses a real (non-headless) browser with a genuine user-agent — this should not happen normally