Zob

Read what you're reading in Zotero, and get context-aware completions in Obsidian: citations, equations (as LaTeX), theorems/definitions, and figures sourced from the paper open in your Zotero reader.

Zob is designed in tiers — it works with just the plugin, and gets more capable as you add optional pieces. None of the optional layers are required.

Install

From the Community Plugins store

Once accepted: Settings → Community plugins → Browse → "Zob" → Install → Enable.

Via BRAT (before it's in the store)

Install the BRAT plugin, then Add beta plugin with this repository URL.

Manually

Download manifest.json, main.js, and styles.css from the latest release into <vault>/.obsidian/plugins/zob/, then enable it in Settings → Community plugins.

Usage

With a paper open in Zotero (or after running "Zob: Set current paper…"):

• Type @ for citations — inserts a bidirectional [[wikilink]] if you have a literature note for that paper, otherwise a zotero:// link (with an inline option to create & link a note).
• Type ;; for content from the current paper — figure/section references, and (with a MinerU token) equations, theorems, and figures. Insert formats are configurable, and each insert can carry a zotero:// page backlink.
• Run "Import block…" to pull a whole paragraph with its equations by meaning — type a phrase or concept (e.g. "profit maximization") and Zob ranks the paper's blocks and inserts the one you pick (see below).

Commands: Set current paper, Clear pinned paper, Create literature note, Extract paper (MinerU), Import block…, Show status & capabilities.

Semantic block import

"Import block…" finds a meaningful block (a paragraph plus its equations, tagged with its section heading) by meaning, not just keywords — so "profit maximization" matches the paragraph about maximizing expected profit even if it never uses that phrase.

• Blocks come from the same MinerU extraction as equations/figures (run "Extract paper" once).
• Ranking is hybrid: a section-heading/text keyword match, plus semantic similarity from an embedding backend. Block vectors are computed once per paper and cached.
• Two embedding backends (settings → Embedding backend):
  • Ollama — local, no key, no rate limits, offline. Run ollama serve and ollama pull nomic-embed-text. Recommended.
  • Voyage — cloud, free tier; sends block/query text to Voyage to embed.
• With neither configured it still works, ranking by keyword only.

Optional companions

These live in this repo and enable the higher tiers:

Without them, Zob falls back to Zotero's local API (recent paper / manual picker) and works fine — run "Zob: Show status & capabilities" to see which tiers are active.

Development

cd obsidian-plugin
npm install
npm run dev          # watch build → main.js
bash install.sh /path/to/vault  # copy manifest.json (root) + main.js + styles.css to your vault

Releasing

node scripts/version-bump.mjs 0.2.0     # update manifest.json + versions.json
git commit -am "0.2.0" && git tag 0.2.0 && git push --follow-tags

Pushing the tag runs .github/workflows/release.yml, which builds the plugin and attaches manifest.json, main.js, and styles.css to a GitHub release named for the version — the format the Community Plugins store expects.

Network use & privacy

Zob is desktop-only and makes network requests in these cases — all optional except the first:

• Zotero (localhost) — reads the paper you're reading, metadata, fulltext, and annotations from Zotero's local API on 127.0.0.1:23119. Local only.
• MinerU cloud (only if you run "Extract paper") — your PDF is uploaded to MinerU's servers to extract equations/theorems/figures. Requires your own MinerU token. Don't extract PDFs you can't share with a third party.
• Voyage (only if you choose the Voyage embedding backend) — block and query text is sent to Voyage to compute embeddings. The Ollama backend keeps this fully local instead.

No telemetry or analytics. Nothing is sent anywhere unless you enable the feature that needs it.

License

MIT — see LICENSE.