biotracker

An iOS app for people who get bloodwork from more than one provider and want to actually see what's happening over time. Snap a photo of any lab report, and the app extracts the values, normalizes them against reference ranges, and charts the trends.

Why I built this

I track my own bloodwork across multiple providers and the data lives in five different PDFs, two patient portals, and a stack of paper. Nothing on the App Store would ingest a photo from an arbitrary lab and give me a clean longitudinal view. So I built it.

Stack

• Swift 6 · SwiftUI · iOS 26
• SwiftData for local persistence (photos kept out of the store via @Attribute(.externalStorage))
• Apple Charts framework for trend visualization
• Anthropic Claude vision API (claude-sonnet-4-5) for structured extraction from lab photos
• XcodeGen (project.yml) so the project file stays diffable

How it works

• Capture — pick a photo of a lab report from the photo library. The image is JPEG-compressed and sent to Claude with a strict JSON schema prompt; the response is validated against the same LabResultsImport type the JSON importer uses.
• Normalize — every reading is attached to a canonical Biomarker with its own reference range, so values from different labs line up on the same axis.
• Trend — biomarker detail views show every reading over time with in-range / out-of-range zone bands and flag-colored points, so abnormal moves are obvious at a glance.
• Dedup — re-importing the same draw merges into the existing record instead of duplicating it.
• Backup — full JSON export/import of biomarkers and their reading history.

Tabs

• Dashboard — biomarkers grouped by category with quick stats (count, flagged, last draw)
• Trends — multi-biomarker chart view
• Timeline — every blood draw in reverse chronological order
• Settings — import / export, photo capture, API key

Setup

1. Clone, then run xcodegen generate to produce BioTracker.xcodeproj.
2. Open in Xcode 16+, build for an iOS 26 simulator or device.
3. To use photo extraction: Settings → AI Extraction, paste an Anthropic API key. Stored in UserDefaults on-device only.

Status

Personal project, actively iterating. The photo extraction pipeline is the newest piece — works end-to-end against claude-sonnet-4-5 but the prompt is still being tuned for less common lab formats.