new-collection

Calibrating a new collection in the directory-pipeline involves a one-time setup to generate reusable OCR and NER prompts. You download the first volume, select pages, and generate ner_prompt.md and ocr_prompt.md files stored under output/{slug} for future reuse.

Setting up reusable OCR and NER prompts requires a one-time calibration on the first volume of your collection. This generates tailored ner_prompt.md and ocr_prompt.md files that you reuse for subsequent volumes with minimal changes.

Yes, you can reuse the generated extraction prompts for subsequent volumes after the initial collection setup. The calibration generates ner_prompt.md and ocr_prompt.md files requiring only minimal changes when processing later volumes in the same series.

Yes, the directory-pipeline calibration process provides end-to-end guidance through downloading the first volume and selecting pages. This ensures you correctly configure the initial extraction parameters before generating the reusable prompts.

You need to recalibrate NER and OCR prompts only once per new collection type. After the initial calibration generates your ner_prompt.md and ocr_prompt.md files, you process all subsequent volumes in that series by reusing the existing prompts with minimal adjustments.