MatchVahti · Lite

RIS is a plain-text bibliography format (.ris) that almost every reference manager reads: Zotero, EndNote, Mendeley, Papers. Each record is a short list of tagged lines: TY type · AU author · TI title · PY year · DO DOI · N1 notes · KW tags · ER end-of-record. You can open it in any text editor and read exactly what will be imported — nothing hidden.

To get it into Zotero:

1. Download .ris (or Copy) above. Each download is timestamped (e.g. matchvahti-lite-2026-06-11-1432.ris) so batches never overwrite each other.
2. In Zotero, choose File → Import… and pick the .ris file (or open the file if your system already associates .ris with Zotero). Don't drag it into the Zotero window. Depending on where it lands that can attach the file to an item instead of importing it; File → Import… always imports.
3. Zotero creates the items, with your captured sentences and note in each item's Notes/Extra and your tags attached. If a reference is already in your library (same DOI/PMID), Zotero can detect it as a duplicate (Duplicate Items view) so you can merge instead of double-adding.

Re-importing makes a new collection each time. Zotero names the import collection after the file, so importing repeatedly stacks up folders. Two ways to avoid the pile-up:

• In the import dialog, untick “Place imported collections and items into a new collection” — items land loose in your current collection instead of a new folder.
• Or keep the box ticked and let the timestamped filename name each folder as a dated batch (matchvahti-lite-2026-06-11-1432), so re-imports don't collide and you can tell them apart.

Either way, the export is cumulative (it always contains every reference you've captured this session), so import a fresh .ris once when you're done rather than after every capture, and let Zotero's duplicate detection catch any overlap.

[oo] closer look means: this abstract sentence is interesting enough to justify pulling the full text. It's a retrieval trigger, not a judgement on the study.

In the exported .ris, the flag travels as the tag cite:closer-look (alongside cite:abstract-only), carried with the paper's DOI and PMID. That tag is the contract: it is how the retrieval step recognises which references get a full-text fetch. Two ways to close the loop — both local-first and open-access only:

1. FullVahti: the Zotero plugin (no terminal). Install once from a file (two clicks in Zotero 7), then Tools → FullVahti: Find OA PDFs for tagged items. It scans your library for cite:closer-look, attaches the free, legal PDF when one exists (Unpaywall / PubMed Central, by DOI or PMID), tags each item fulltext:pdf-found / fulltext:pdf-missing / fulltext:check-needed, and leaves one run-report note listing what's still missing — ready for interlibrary-loan requests.
2. vahtian_fulltext.py: the scripted path (power users). Same scan, same fulltext:* tags, dry-run by default.

Later, MatchVahti or CiteVahti can use those retrieved full texts for stronger, full-text claim checking instead of abstract-only. MatchVahti-Lite itself never fetches PDFs and never writes to your library — it only emits the flag; retrieval is a separate, explicit step.

Because CiteVahti is an installed app and this is a single web page. CiteVahti runs as a real process on your machine with a small backend, so it can do everything an OAuth handshake needs: hold a client secret that never reaches the browser, open a localhost callback to catch the redirect, and then keep an API key to write items into your Zotero library directly — decision-gated and undoable.

MatchVahti-Lite is the opposite by design: one static HTML file, no backend, no server, and network access locked down to PubMed E-utilities. That removes every piece OAuth requires:

• No callback to return to. OAuth must send you back to a URL the app controls; a static file has none.
• Nowhere to keep a secret. Anything in the page is readable by anyone who opens it, so a client secret can't be protected.
• No path to Zotero's servers. The content-security policy only permits network calls to PubMed, so the page can't reach Zotero's API (and the browser's CORS rules would block them too).

So instead of a credentialed write, it uses the universal, no-login path: a reviewable .ris you import yourself. That is also the whole spine's no silent writes stance — nothing reaches your library until you have looked at it and clicked import. When you want the direct, authenticated Zotero write, that lives in CiteVahti / MatchVahti.