Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.bowerlabs.ai/llms.txt

Use this file to discover all available pages before exploring further.

Shipped 2026-03-13 · AI Voice transcription and OCR usually work well. The risk is when they almost work — a misread digit, a mangled unit, a transcription that looks plausible but isn’t. Bird now reads every capture after the fact, scores its confidence, and flags any that look uncertain.

​
What’s new

  • Background quality check after every voice transcription and every photo OCR.
  • Confidence scoring — Bird scores the result against the source (audio recording or original image) and assigns high, medium, or low confidence.
  • Proactive toast — if confidence is medium or low, Bird surfaces a short notification on your screen pointing you at the specific ambiguous values, with a one-click jump to the note.
  • No interruption to capture — the check runs after the capture is saved. It never blocks you from finishing what you were doing.

​
Why this matters

Silent errors are the worst kind. A typo you can see is fixable. A misread “240 µmol/L” that looks the same as “240 µmol/I” only surfaces months later when someone tries to reproduce the result. Proactive confidence flags catch the ambiguous cases at the moment of capture, when the source is fresh and the fix is cheap.