Extraction pipeline

Image upload

Choose one reaction figure.

Image URL

Fetch a public HTTPS image.

PDF document

Select a document for page screening.

Render pages

Render page overviews and detailed tiles on demand.

Index PDF text

Link captions, footnotes and referenced sections.

Request gateway

Validate and forward BYOK requests.

Select reaction pages

Include detected reactions and explicit procedures.

Extract candidate

Read chemistry from images and matched text evidence.

Validate against source evidence

Recheck images, text and evidence citations.

Normalize response

Parse and standardize model JSON.

Combine page results

Keep PDF findings tied to pages.

Review and export

Inspect, copy, or download output.