Documentation
¶
Overview ¶
Command validation-m3 is the M3 validation run: ARCHITECTURE_PHASE3.md's step 11, the 20 seeds of M1's run (spikes/8-validation) again, now for the papers on disk — PDFs, their text, search, citation context and import.
Each seed gets a library folder of its own. For each: add it, expand to 200 works (M1 used 500; the PDFs are what is under test here), fetch at most 25 PDFs — open access only, a refusal is recorded and never worked around (hard rule 1) — which reads them too. Then, for every paper with text, search for its own title's longest words across the library and see whether it comes back in the first five papers.
After every seed, an import check per field: up to 8 of the field's downloaded PDFs, copied under meaningless names into a fresh library's Inbox/, imported, and the paper each was filed as compared with the paper it was downloaded for.
Last, a sample of 10 citation sentences per field is written to mentions-sample.json for checking by hand.
No contact email is sent. Everything is written under -root, which should not be a temporary folder: the hand check reads it afterwards.
Usage:
go run ./spikes/12-validation-m3 -root <folder> > report.md go run ./spikes/12-validation-m3 -root <folder> -recheck-imports go run ./spikes/12-validation-m3 -root <folder> -recheck-search
Source Files