Create a literature database

Define your modules → upload sources → extract → review and publish

PDFs stay private. Only reviewed structured results are published.

Template editing is available. Cloud storage, the task queue and server secrets must be configured before extraction can start.

Authorized workspace

Database nameModel

Saved module definitions are frozen for reproducibility. Start a new project to use a different configuration. Trial guardrail: 500 article × module tasks per project; up to 5 attempts per task.

2. Source PDFs

Main articlesAutomatic P001_Title.pdf naming; originals are retained in the mapping.Supporting Information · optionalExplicit pairing prevents mismatches. No SI is required.

Up to 10 parallel uploads. PDFs are parsed one at a time to limit memory use. Keep this tab open until uploads finish.

Text-based PDFs only in this version. Figures are not read; scanned documents require OCR. Text is never silently truncated.

3. Extraction & review

Each article × module is checkpointed separately. Progress refreshes every 5 seconds; you can close the browser after submission.

Cloud extraction: up to 10 article × module tasks at once, shared across projects. Completed tasks are not rerun.

0 / 0 schema-validated modules · 0 tokens

Schema validation checks structure, not scientific correctness. Check source evidence before publishing. Missing SI is never inferred.

4. Publish structured results

I have reviewed the results and approve public access to this dataset (excluding PDFs).