Iris
Published:
Iris turns scanned lab notes (PDF, PNG, JPG) into corrected, searchable markdown. Scans are read by an OCR model (Chandra OCR 2), then a local LLM served by Ollama corrects what the OCR misread and translates any non-English text into English. Everything runs locally, so lab notes never leave the machine.
When a context document for the same work exists, such as a markdown protocol or a Hermetica protocol lock file, the LLM uses it to fill gaps and fix values in the note.
- The raw OCR is kept next to the corrected note, as a record of what the scan said before the LLM touched it
- Long OCR runs resume page by page after an interruption
- Scans are paired with their context documents by file name