Capture bibliographic metadata (journal, DOI, year, preprint-vs-published) for paper notes

[MKrbm]

Problem

When a paper is fetched, the note only stores the raw source: URL, and
_detect_content_type lumps everything into content_type: paper. So journal,
volume, etc.. are never captured but arXiv preprint
is indistinguishable from a peer-reviewed journal article. That distinction is
useful info we currently throw away.

Motivation

• The vault compounds over time; it's far more valuable if each paper carries
  real bibliographic identity, not just a URL.
• Preprint vs published, venue, etc.. are worth keeping on the note.

Proposed strategy

A standalone, opt-in hr biblio command (working name). Does not touch the
fetch path or the 16-step pipeline.

• Scans content_type: paper notes, parses the DOI/arXiv ID from the existing
  source: URL, queries OpenAlex (no key), writes bib fields into frontmatter.
• Confident-ID match only (no title-search → no mis-linking). fail-open on any
  error. Fills empty fields by default; --force to refresh (can update preprint→published).
• New optional NoteMeta fields: doi, journal, volume, issue, pages, year, authors, openalex_id, pub_type, bib_enriched_at (+ a preprint/venue field).
• Add matching SQLite columns via idempotent migration; sync rebuilds them from
  frontmatter. Surface the fields in note show / search / MCP read tools.

Out of scope (for now)

No BibTeX export, no change to report ## Sources, no fetch-time auto-enrichment.

[MKrbm]

If you are comfortable with this, I can create a PR for it.

[jordan-gibbs]

Yes, send the PR! That's pretty much the shape that I'd have asked for: a standalone opt-in command that only matches confident IDs and fails open. Two requests going in:

1. Keep the OpenAlex client on plain httpx (it's already a base dep) so hpr biblio doesn't grow the install.
2. Add a --dry-run flag that prints what it would write. It mutates frontmatter across a lot of notes and I want a preview mode before anyone points it at a big vault.

bib_enriched_at for staleness is a good call, keep it.

[jordan-gibbs]

A heads-up before you start building. The 2.0 groundwork that merged since we talked (#48) shipped hpr sources score, which already covers a chunk of this: it enriches DOI-bearing notes through OpenAlex and stores venue, citation count, retraction status, and the DOI itself. I should've flagged it here sooner.

So rather than a standalone hpr biblio, I'd ask you to build on sources score: add the preprint-vs-published distinction, the --dry-run preview, and bib_enriched_at for staleness. Those are the pieces still missing, and they're exactly the ones from your proposal that I want. Same two constraints as before: plain httpx, confident-ID matches only.

[jordan-gibbs]

Correcting my earlier comment, because I sent you after something that already exists.

I asked for preprint-vs-published as part of this. That's already shipped — core/oa.py:61-65 ranks publishedVersion / acceptedVersion / submittedVersion, and the value is persisted end to end: frontmatter (core/oa.py:534), the model (models/note.py:120), SQLite (core/db.py:59 with the migration at core/migrations.py:251), sync, and note show -j. Don't build it.

Also already in place, so the scope is smaller than the issue implies: doi, citation_count, venue and is_retracted all round-trip through frontmatter and SQLite via core/scholar.py:311-329; sources score already queries OpenAlex for DOIs and Semantic Scholar for arXiv ids over plain httpx behind a TTL cache; and sources backfill-doi already parses DOIs out of source URLs.

What's actually left:

• The missing fields — year, authors, journal, volume, issue, pages, openalex_id, pub_type. Most of the work is one line: core/scholar.py:177 narrows the OpenAlex response with ?select=cited_by_count,primary_location,is_retracted, and the payload already carries the rest. Widen that, add the columns, map them.
• --dry-run — doesn't exist on any sources subcommand.
• bib_enriched_at — staleness is currently inferred from citation_count IS NULL, so a note enriched two years ago looks identical to one enriched today.
• Making it queryable. This is the piece neither of us named, and I think it's the one that matters most for your "vault compounds over time" framing: hpr search exposes 23 filters and not one of them touches doi, venue, citation_count, is_retracted or oa_version. "Show me every preprint in this vault" needs raw SQL today. Metadata you can't filter on isn't doing much.

One schema wart worth fixing alongside: is_retracted is INTEGER NOT NULL DEFAULT 0 at core/db.py:53, so it can't represent "not checked", while models/note.py:114 types it bool | None. Frontmatter can say unknown; the database silently says not-retracted.

Still want this, and the constraints from my earlier comment stand — plain httpx, confident IDs only, no new dependency.