extract_assessment() over a set of files. A file that fails to extract is
reported and skipped rather than taking the batch down with it – one
unreadable PDF in a folder of twelve should not cost the other eleven.
Arguments
- files
Character: Paths to the documents.
.docx,.html,.odt,.rtf,.epuband.tex(each requiring thepandocpackage),.pdf(requiringpdftools),.csv, or.R,.qmd,.Rmd,.mdand.txtread as they are. Seeread_document(). They need not be of one format: a folder of.docxhomework sets and a.pdffinal exam is one batch.- llmconfig
rtemis.llm::LLMConfig: Model configuration.- names
Character: Assessment names, one per file. Defaults to the titles found, else the file names.
- verbosity
Integer
[0, Inf): Verbosity level.- ...
Additional arguments passed to
extract_assessment().
Value
List of Assessment objects, named by assessment. Where a document
was skipped, the "failed" attribute carries a data frame of its path and
the error that stopped it.
Details
The run opens with the number of documents and closes with how many of them
produced items, because that is the number a skipped file changes. The
per-file warning has scrolled past by the time a batch of twelve finishes,
and the returned list is named by assessment rather than by file, so nothing
else in the result says that three documents are missing from it. What was
skipped, and why, comes back in the "failed" attribute.
Assessment names come from the document titles where those identify a
document, and from the file names otherwise – a course whose every question
set is titled "Homework Questions" is named from its files. Two names can
still coincide, and the second is suffixed rather than allowed to overwrite
the assessment already holding it, since the name becomes the YAML key in
write_assessments() and the id prefix after that. Pass names to settle
all of this yourself.