
Build a Python script that auto-generates architecture decision records from your codebase
You'll have a working script that analyzes your code structure and generates ADR drafts for your team to review and finalize, saving hours of documentation overhead.
What you'll make
ADR Generator Script
- scan_codebase() — file structure and dependency analysis: what to include
- extract_patterns() — AI identification of architectural choices: what to include
- generate_adr_draft() — structured ADR template population: what to include
- main() — orchestration with human review checkpoints: what to include
Paste your complete Python script with example output from running it on your actual codebase
The 3 parts
Each part ends with a check you can see.
Map Hidden Decisions
Scan your codebase by hand to identify 5-7 architectural choices that aren't documented anywhere but are visible in the code structure
Done when
A list of 5-7 specific architectural decisions with the file/folder evidence that shows each choice was made
Build the Pattern Scanner
Create the scan_codebase() and extract_patterns() functions that use AI to identify architectural patterns in your code structure
Done when
Working functions that can scan your codebase and identify at least 3 of your hand-mapped decisions from Task 1
Draft ADR Templates
Build generate_adr_draft() and main() functions that convert detected patterns into structured ADR drafts with human review checkpoints
Done when
A complete script that generates ADR drafts for detected patterns and includes explicit prompts for human review and editing
Capstone: you put together your ADR Generator Script.
Where this helps
An elective: a single lesson that sits outside rtwo's job certificates and doesn't count toward one.
About this lesson
What you'll be able to do
- Run a working Python script that scans your real codebase and generates ADR drafts with architectural decisions extracted from code patterns
- Identify the 3-5 types of architectural choices where AI gets it right vs where human context is required
- Have a documented workflow for keeping ADRs current as your codebase evolves
- Know when to trust AI's architectural analysis vs when to override with team knowledge
Who it's for
- Development teams drowning in undocumented architectural decisions
- Tech leads who want to automate ADR creation but need human oversight on the output
- Anyone maintaining a Python codebase who can read and modify scripts
What it won't cover
- Writing ADRs from scratch without existing code to analyze
- Complex enterprise deployment or CI/CD integration
- Learning ADR format fundamentals — assumes you know what good ADRs look like
- Multi-language codebases beyond Python
Skills you'll practise
- Code structure analysis with AI
- Architectural pattern recognition
- Human-AI judgment calibration
- Documentation automation workflow design