Exam Practice Scraper

Scrapes any examcademy.com exam into a local Streamlit practice app. Change one value in config.json to switch exams (DP-600, AZ-305, etc.).

Current state: DP-600 fully loaded — 178/178 questions with correct answers.

---

Prerequisites

Tool	Purpose
Python 3.11+	runs the scraper and app
Microsoft Edge	authenticated browser for scraping
Claude Code personal subscription	analyzes image-only question answers

---

Setup (once)

pip install -r requirements.txt
playwright install msedge

config.json — the only file you ever change:

{
  "exam_code": "dp-600"
}

To switch to AZ-305: change "dp-600" to "az-305". That's it. Everything else is automatic.

---

Full pipeline

Run these steps in order when scraping a new exam. For DP-600 only Step 4 and 5 are needed (Steps 1–3 are already done and image_fixups.json is committed).

Step 1 — Scrape

python scraper.py

Opens your real Edge profile (already logged in to examcademy.com). Clicks "Show Answer" on every question, saves HTML + images to brons/.

Step 2 — Build silver layer

python pipeline.py

Merges all bronze pages into silver/<exam-slug>_questions.json.

Step 3 — Extract image-based answers

HOTSPOT and DRAG DROP questions reveal their answers as screenshots — there is no text in the HTML. Claude Code reads those images natively and writes the structured answers to image_fixups.json.

Skip this step for DP-600 — image_fixups.json is already in the repo with all 58 image questions solved.

When you scrape a new exam, open Claude Code in this folder and run:

/analyze-exam-images

Claude Code reads every answer image, identifies what is selected/placed in the UI screenshot, and writes image_fixups.json. Takes about 2 minutes per exam.

Why Claude Code instead of OCR? You already have the subscription — no extra API key or cost. It reads images natively, understands dropdowns/checkboxes/drag-slots visually, and produces accurate structured answers. No Tesseract, no heavy ML packages, no slow processing.

Step 4 — Load database

python preload_db.py

Reads the silver layer + image_fixups.json, merges them, and writes all questions to Practice Exam/exam_app.db. Prints a summary like:

Loaded 178 questions
With correct_answer: 178/178

Step 5 — Start the app

cd "Practice Exam"
streamlit run app.py

Open http://localhost:8501.

---

Switching exams (or re-scraping the current one)

python switch_exam.py az-305   # switch to a different exam
python switch_exam.py dp-600   # re-scrape the current exam from scratch

Updates config.json, wipes stale brons/ and silver/ data, and prints the next steps. Works for both switching to a new exam and forcing a clean re-scrape of the current one. Then run Steps 1–4 as usual. The app detects the exam change and clears the old question bank automatically.

---

Project layout

config.json                ← exam settings (gitignored)
config.example.json        ← template
switch_exam.py             ← switches to a new exam (updates config, wipes stale data)
image_fixups.json          ← image-question answers (written by /analyze-exam-images)
brons/
  raw/page_N.html          ← saved raw HTML per page
  page_N.json              ← parsed questions per page
  images/                  ← downloaded question and answer images
silver/
  <exam-slug>_questions.json
Practice Exam/
  app.py                   ← Streamlit practice app
  exam_app.db              ← SQLite session/answer store (gitignored)