Why a Sewing Pattern Manager?
Crafty sewists in the US and Canada number around 30 million, dwarfing niche markets like 3D‑printing. Managing a growing stash of paper and digital patterns quickly becomes a logistical nightmare.
From Filament Spools to Fabric Patterns – The Migration Challenge
The first hurdle was repurposing an existing 3D‑printing app. Every UI label, data model, and file path referenced spools of filament. A systematic rename‑and‑refactor effort was needed, but hidden dependencies caused the migration to stall.
- Copied project folder and renamed it.
- Replaced all domain‑specific terms (spool, filament) with sewing‑related terminology.
- Used Claude Code in the terminal to locate stuck processes and dangling references.
Running Claude Code outside Xcode consumed 91 % of the token cap, leading to a three‑hour work stoppage. The lesson: background agents can silently drain AI budgets.
AI‑Assisted Refactor – Getting Back on Track
After resetting the environment so no agents ran in the background, Claude cleaned up the migration mess in about 20 minutes. The AI could read the project, understand the domain shift, and apply the necessary code changes without human oversight.
Building the Scanner: High‑Quality Image Capture and Auto‑Crop
The app needed to turn a phone camera into a scanner for pattern envelopes. AI handled three steps:
- Detect the envelope within the photo.
- Straighten and rotate so corners form right angles.
- Crop to the exact bounds.
This automated workflow eliminated manual editing for users.
Extracting Vendor and Pattern Numbers with OCR
Most paper patterns are indexed by vendor name (e.g., Simplicity, McCalls) and pattern number. The app uses OCR to pull these fields from the scanned image.
During testing the AI confused barcode digits with pattern numbers. By training a small classifier, Claude helped the app filter out barcode sequences, ensuring only the true pattern number is stored.
Bonus Features Enabled by OCR
- Full‑text searchable OCR field for any text on the front or back cover.
- Storage of the raw barcode number and image for future reference.
AI as a Build‑and‑Test Partner
Claude can invoke Xcode builds, detect compilation errors, and apply fixes autonomously. This allowed the developer to take a walk in the park while the AI completed image‑orientation logic.
Hands‑Free Development with Dictation
Because the developer works one‑handed (a Yorkipoo on the shoulder), dictation via Wispr Flow makes coding feasible without a keyboard.
Key Takeaways for Indie Developers
- Integrate AI tightly with the IDE to avoid token‑draining background agents.
- Use AI‑driven refactoring for large domain migrations, but monitor usage caps.
- Leverage OCR and simple ML models to add powerful data extraction without building complex pipelines.
- Hands‑free dictation can dramatically improve accessibility and productivity.
Discussion Prompt
Have you tried agentic or vibe‑coding workflows? What worked, what broke, and how do you manage token usage when AI lives inside your IDE? Share your thoughts below.