AUSCrawl

20 years of AUS course data, one SQLite file.

---

Warning

Do not run the crawler unless you know what you are doing. The crawler makes tens of thousands of requests to AUS Banner and can easily overwhelm the server if misconfigured, which can result in service disruption and get you in trouble with the university. A pre-built database (aus_courses.db) is already included in this repository with a complete snapshot of all course data since 2005 — use that instead.

What is this?

AUSCrawl is a fast, async web crawler that scrapes AUS Banner for course data across every semester since 2005 and stores it in an SQLite database. But more importantly, it ships a ready-to-use database (as a release download) so you never have to run the crawler yourself.

Written in Python. Single file. ~15 minutes for a full crawl of 74,000+ course sections, catalog descriptions, prerequisites, and more.

---

The Database

Every release ships aus_courses.db (gzipped, ~16 MB), a complete SQLite database containing every course, instructor, prerequisite, and catalog description from AUS Banner since Fall 2005. Just download it, gunzip, and start building. (It's distributed as a release asset rather than committed to the repo so clones stay small.)

Table	Records	Description
courses	75,467	Every course section ever offered
course_dependencies	156,512	Prerequisite/corequisite links with minimum grades
section_details	73,778	Prerequisites, corequisites, restrictions, waitlist, fees (+ structured JSON)
section_instructors	72,476	Every instructor per section (incl. co-taught), with primary flag
catalog	3,046	Course descriptions, credit/lecture/lab hours
catalog_detail	3,532	Course-level attributes (degree-requirement tags), schedule types, levels
instructors	1,987	All instructors with emails and first appearance
semesters	101	Every term from Fall 2005 to the present
subjects	98	All subject codes (COE, ENG, MTH, etc.)
attributes	231	Course attributes
levels	9	Academic levels (Undergraduate, Graduate, etc.)

---

Build Something Cool

This dataset is a goldmine for AUS students. Use it to help your fellow students or sharpen your own skills:

• Prerequisite visualizer — build an interactive graph of course dependencies for your major
• Schedule planner — help students find open sections that fit their timetable
• Instructor tracker — see which professors teach what, and how their assignments changed over the years
• Course trend analysis — which courses are offered less frequently? Which departments are growing?
• Grade requirement explorer — find every course that requires a minimum grade of C- or higher
• Data science projects — 20 years of course data across 98 subjects is a great dataset for learning SQL, pandas, or building dashboards

If you build something with this data, open an issue and let us know — we'd love to see it.

---

Getting Started

# Download the latest database (compressed, ~16 MB) from Releases
curl -L -o aus_courses.db.gz \
  https://github.com/DeadPackets/AUSCrawl/releases/latest/download/aus_courses.db.gz
gunzip aus_courses.db.gz

# Open it with sqlite3
sqlite3 aus_courses.db

# Or use Python
python3 -c "
import sqlite3
conn = sqlite3.connect('aus_courses.db')
for row in conn.execute('SELECT term_name, COUNT(*) FROM courses JOIN semesters ON courses.term_id = semesters.term_id GROUP BY courses.term_id ORDER BY courses.term_id DESC LIMIT 5'):
    print(row)
"

Example Queries

-- All courses taught by a specific instructor
SELECT term_id, subject, course_number, title, days, start_time, end_time
FROM courses WHERE instructor_name LIKE '%Smith%'
ORDER BY term_id DESC;

-- Courses with prerequisites and minimum grades
SELECT d.subject, d.course_number, d.dep_type, d.minimum_grade,
       sd.prerequisites
FROM course_dependencies d
JOIN section_details sd ON sd.crn = d.crn AND sd.term_id = d.term_id
WHERE d.dep_type = 'prerequisite'
GROUP BY d.subject, d.course_number;

-- How many sections per semester
SELECT s.term_name, COUNT(*) as sections
FROM courses c JOIN semesters s ON c.term_id = s.term_id
GROUP BY c.term_id ORDER BY c.term_id;

-- Course catalog with hours breakdown
SELECT subject, course_number, description, credit_hours, lecture_hours, lab_hours
FROM catalog WHERE subject = 'COE';

-- Find all prerequisites for a specific course
SELECT d.subject, d.course_number, d.minimum_grade
FROM course_dependencies d
JOIN courses c ON c.crn = d.crn AND c.term_id = d.term_id
WHERE c.subject = 'COE' AND c.course_number = '390'
GROUP BY d.subject, d.course_number;

---

Database Schema

The SQLite database contains 13 normalized tables with proper indexes:

Core tables:

• semesters — term ID and name (e.g. 202620, Spring 2026)
• subjects — subject codes and full names (e.g. COE, Computer Engineering)
• courses — every course section with schedule, instructor, classroom, etc.
• instructors — deduplicated instructor names and emails with first_seen
• levels — academic levels (Undergraduate, Graduate, etc.)
• attributes — course attributes with first_seen

Extended tables:

• catalog — course descriptions, credit/lecture/lab hours, department
• catalog_detail — course-level attributes (degree-requirement tags), schedule types, levels, and catalog-level prerequisites/corequisites/restrictions
• section_details — prerequisites, corequisites, restrictions, waitlist, fees per section, plus structured prerequisites_json / corequisites_json (boolean AND/OR expression trees) and restrictions_json (typed include/exclude groups)
• section_instructors — every instructor on each section, including co-taught ones, with an is_primary flag
• course_dependencies — flat prerequisite/corequisite links with minimum grade requirements

---

Banner Technical Details

AUS uses Ellucian Banner, a student information system widely deployed across universities. The public-facing schedule search is served at banner.aus.edu behind Cloudflare, exposing several OWA (Oracle Web Agent) endpoints:

Endpoint	Method	Purpose
/axp3b21h/owa/bwckschd.p_disp_dyn_sched	GET	Semester dropdown — returns all available term IDs
/axp3b21h/owa/bwckgens.p_proc_term_date	POST	Subject listing — returns available subjects for a given term
/axp3b21h/owa/bwckschd.p_get_crse_unsec	POST	Course search — returns HTML tables of all matching sections
/axp3b21h/owa/bwckctlg.p_display_courses	GET	Course catalog — returns descriptions, credit hours, department
/axp3b21h/owa/bwckctlg.p_disp_course_detail	GET	Course detail — returns course-level attributes, schedule types, prerequisites
/axp3b21h/owa/bwckschd.p_disp_detail_sched	GET	Section detail — returns prerequisites, corequisites, restrictions, waitlist, fees

The course search endpoint accepts all subject codes in a single POST body (up to ~4,500 bytes before the WAF rejects it), returning a large HTML page with <table class="datadisplaytable"> rows. Instructor emails are obfuscated using Cloudflare's email protection (XOR encoding with the first byte as key). The crawler paces the GET endpoints with a global token-bucket rate limiter (AIMD around ~18–25 req/s); in practice 429 responses begin around ~30 req/s, well below the documented stream limits (~10,000 HTTP/2 streams per connection). Actual enrollment/seat counts are not exposed by AUS Banner (only waitlist figures), even for completed terms.

---

Crawler Documentation

Caution

Only run the crawler if you need fresher data than what's in the included database. Be aware that aggressive crawling can take down AUS Banner and result in your IP being banned. The default settings are tuned to be safe, but modifying worker counts or running multiple instances simultaneously can cause problems.

Click to expand crawler docs

Requirements

Python 3.13+ and uv.

Usage

uv run python crawl.py [options]

Flag	Description
-o, --output	SQLite output path (default: aus_data.db)
-t, --terms	Only crawl specific term IDs (e.g. 202620 202510)
-w, --workers	Max concurrent requests (default: 50)
--rate	Target GET requests/sec; AIMD ceiling that paces the catalog/detail phases (default: 18, backs off on 429s). Raise to go faster, lower for extra safety
--delay	Extra seconds to pause before each request (default: 0; pacing is normally handled by --rate)
--latest	Only crawl the most recent semester
--resume	Skip semesters already in the database
--force	Drop and recreate all tables
--no-catalog	Skip catalog description scraping
--no-details	Skip section detail scraping
-v, --verbose	Debug-level logging

How It Works

The crawler runs in 5 phases:

1. Semester discovery — fetches the list of all available terms from Banner's dropdown
2. Subject catalog — fetches subject codes from every semester and deduplicates (the dropdown varies per term)
3. Course scraping — POSTs to the schedule search endpoint for every semester with all subjects in a single batch, then parses the HTML response with lxml (50 concurrent workers)
4. Catalog scraping — GETs course catalog pages for a sample of 6 evenly-spaced terms to collect descriptions, hours, and departments (10 concurrent workers)
5. Detail scraping — GETs the section detail page for every unique CRN/term pair to extract prerequisites, corequisites, restrictions, waitlist info, and fees (10 concurrent workers)

Technical Details

• Async HTTP/2 via httpx with connection pooling and automatic retry with exponential backoff
• lxml for HTML parsing (12x faster than BeautifulSoup)
• ThreadPoolExecutor offloads CPU-bound parsing from the async event loop
• Catalog sampling reduces catalog requests by ~80% while maintaining full course coverage
• Cloudflare email protection decoding (XOR-obfuscated instructor emails)
• Crash resilience — each phase saves to DB immediately; detail phase does periodic batch saves every 5,000 entries; --resume skips completed work
• Rate-limit aware — respects server 429 responses with exponential backoff; GET endpoints capped at 10 workers to avoid triggering bans

---

_{Built for AUS students, by an AUS student.}
MIT License