Decompress a .zst file, returning the path to the decompressed file.

Delete a manpage and its mappings (via CASCADE) from the store. Deleting from manpages cascades to parsed_manpages, which in turn cas

Download a file with curl, skipping if it already exists.

Remove options with no flags and no positional name. Returns (filtered_list, num_removed).

Full extraction pipeline: prepare → LLM calls → finalize.

Stream contents.tsv.zst and extract matching man pages as .gz files. The contents TSV has columns: id, hash, content Streams via zstd to

>>> extract_name('ab.1.gz') 'ab' >>> extract_name('ab.1.1.gz') 'ab.1' >>> extract_name('ab.1xyz.gz') 'ab' >>> extract_nam

Return source -> (extractor, meta) for all parsed manpages.

Finalize extraction from batch responses. Used by run_batch after collecting provider results. The returned stats carry ``chunks`` a

Replace invalid JSON escape sequences with their escaped form. JSON only allows: \\", \\\\, \\/, \\b, \\f, \\n, \\r, \\t, \\uXXXX. LLMs somet

Return a mapping of source_gz_sha256 → source for all stored manpages. Only includes rows that have both a sha256 and a matching parsed_manpa

Parse all metadata files. Returns parsed data structures.