Manual downloads¶
Some papers can only be downloaded by the user in a normal authorized browser: ScienceDirect always, and any paper that acquisition wrote to manual-download.csv. manual-fetch takes over the files you download, identifies and verifies them, and places them in the reserved delivery paths.
ScienceDirect manual handoff¶
ScienceDirect remains manual by design. Configure an Elsevier API key if the task needs official expected-identity resolution:
Start the watcher before downloading:
python "$skill/scripts/arp.py" manual-fetch `
--input "10.1016/j.asieco.2024.101746" `
--watch "$HOME\Downloads" --output C:\Research\papers
The CLI opens only the canonical article page. The user completes organization login and downloads the official PDF plus raw BibTeX in a normal browser. The CLI observes only stable new or changed files, verifies the pair, and delivers it. It never controls ScienceDirect, reads the user's Chrome profile, exports cookies, or turns an Article Retrieval API 403 into scripted website login.
Files you already have¶
If both files already exist:
python "$skill/scripts/arp.py" manual-fetch --input <DOI> `
--pdf .\paper.pdf --bibtex .\citation.bib --output C:\Research\papers
Items of a frozen corpus¶
For a frozen corpus item, use its immutable selection ID:
python "$skill/scripts/arp.py" manual-fetch `
--selection C:\Research\corpus-run\selection-manifest.json `
--key <selection-id> --output C:\Research\corpus-delivery `
--watch "$HOME\Downloads"
The whole manual queue¶
To work through the whole manual-download.csv of an acquisition in one session, use --queue. It opens each paper's official page in turn, imports pairs in whatever order they are downloaded, and updates the queue. Add --from <folder> instead of --watch when the files are already downloaded: