Spaces:
Running
Running
Commit History
28 days ago
-
Improve PMC auto-fetch fallback
diogo.rodrigues.silvaea1d3e2
29 days ago
-
Add PubMed Central fallback for auto-fetch PDF collection
diogo.rodrigues.silva +1: Claude Sonnet 5Some publishers (e.g. JMIR) gate their PDF endpoint behind a bot challenge (AWS WAF/Cloudflare) that a plain HTTP client can't pass, so neither the direct URL nor Unpaywall (which often just points back at that same blocked URL) can recover the PDF. PMC's and Europe PMC's own article/PDF websites are now similarly gated behind a proof-of-work check, so this doesn't attempt those either — instead it resolves a PMCID (from a PMCID/PMID column, or from the DOI via NCBI's ID Converter API) and fetches full-text JATS XML through Europe PMC's REST API, a sanctioned machine endpoint separate from the gated website. That XML is flattened to plain text and re-rendered as a PDF with PyMuPDF so it flows through the existing extraction pipeline unchanged (no original figures/layout, text only). Verified end-to-end against the JMIR article that surfaced this: the direct URL and Unpaywall both fail (202/blocked), and PMC recovers the full 10-page text. Co-Authored-By: Claude Sonnet 5 <noreply@anth…defc2b4 -
Fall back to a DOI column when the article URL has no DOI
diogo.rodrigues.silva +1: Claude Sonnet 5When a row's URL doesn't contain a parseable DOI (e.g. a publisher landing page URL without one in the path), auto-fetch now looks for a dedicated DOI column (DOI / Digital Object Identifier / Article DOI) and uses that value for the Unpaywall lookup instead of giving up. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>7b0e506 -
Fix Unpaywall auto-fetch always failing due to placeholder email
diogo.rodrigues.silva +1: Claude Sonnet 5Unpaywall's API rejects the example.com default address with a 422 ("Please use your own email address"), so the Unpaywall fallback silently failed on every call unless UNPAYWALL_EMAIL was explicitly set. Now it's skipped cleanly (with a clear status message) when no real contact email is configured, and failures are logged with a reason under APP_DEBUG=1 instead of being swallowed silently. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>ae2da38 -
Add automatic PDF collection with legal open-access fallback
diogo.rodrigues.silva +1: Claude Sonnet 5Rows now try to auto-fetch their PDF on load (from the article URL, then via Unpaywall by DOI) before falling back to the existing manual upload flow. Wired into initial load, accept/skip, row navigation, and session restore, plus a manual "Try auto-fetch PDF" retry button. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>e5ddb25 -
Prepare screener for Hugging Face Spaces
diogo.rodrigues.silva96017f7
-
improve table preview 2
diogo.rodrigues.silvadead85b -
improve table preview
diogo.rodrigues.silvaa6ea449
-
Debug
diogo.rodrigues.silva079ecc0 -
Revert "debug extracted fields"
diogo.rodrigues.silvaThis reverts commit 891f9dab3d60bbf2ff1a41060bd5f0f25ee3750a.a554729 -
debug extracted fields
diogo.rodrigues.silva891f9da -
Optimize table preview
diogo.rodrigues.silva619ce12 -
Add table preview
diogo.rodrigues.silvab025cdf -
“Debug”
diogo.rodrigues.silva20d4104 -
“Debug”
diogo.rodrigues.silva94b35bf -
“Debug”
diogo.rodrigues.silva42b9c28 -
Add row navigation
diogo.rodrigues.silva395d057 -
Add row navigation
diogo.rodrigues.silva8822049 -
Remove labels and expande editability
diogo.rodrigues.silva6d86944