pyautomemory.bib is the canonical metadata layer shared by every sub-wiki under ../wiki/ in this
repository. Wiki source entries explain which claims papers support; the BibTeX file records
citation metadata and canonical keys. Keep PDFs, local paths, abstracts, and long paper
summaries out of both layers.
All canonical metadata lives in this folder; loose .bib files must not be
added elsewhere in the repo. (The file was consolidated from a legacy
personal library in 2026-07 — the merge audit is in git history.)
- Verify the paper from an authoritative public record or the paper itself.
- Search
pyautomemory.bibby DOI, arXiv ID, and title. Reuse the existing canonical key; otherwise add verified metadata under a unique, stable author-year key. - Add a compact section to the relevant
../wiki/<domain>/sources/*.mdusing the inherited schema in../wiki/CLAUDE.md. - Add concept/entity links only where the paper materially supports existing text.
- Run
make validate-literature-citations.
Never record a local PDF path or infer claims from a filename. If metadata or support is uncertain, add a TODO rather than guessing.
Before patching another project's LaTeX, inspect that project's .bib. Match papers by DOI,
then arXiv ID, then title/authors, and use the project's existing key when present. Do not
assume a PyAutoMemory canonical key exists downstream. (A key-alias subsystem existed but was
never used and was retired in #34; renaming a canonical key is a sed across the wiki, which
validation guards.)
make validate-literature-citations
# or: python scripts/validate_literature_citations.pyDuplicate canonical keys, missing source keys, and new claim entries without canonical keys
fail validation. Canonical entries not yet referenced by a wiki source entry are reported but
do not fail; use --show-all for the complete list.