README file from
GithubPaper Finder
Find research papers from whatever scraps of information you have, without leaving Obsidian. Save each one as a source note, and optionally save its PDF too.
You heard about a paper in a talk, saw it cited in a book, or caught half a title on social media. Type what you remember:
schuller constructive gravity 2020
Penedones TASI AdS/CFT
1608.04948
hep-th/9711200
10.1103/PhysRevD.101.084046
https://journals.aps.org/prd/abstract/10.1103/PhysRevD.101.084046
Paper Finder searches several free scholarly databases at once, merges the duplicates, and shows the best matches with authors, year, venue and part of the abstract. Pick one, and it:
- creates a source note from your template, with title, authors, year, DOI, arXiv id, link and citation key. If the paper is already in your vault, it opens your existing note instead.
- opens the paper in Obsidian's web viewer, in a split next to your note. It prefers the arXiv HTML version, which renders the math.
- optionally downloads the PDF, either into your vault (embedded in the note) or into any folder on your computer.
No account and no reference manager are needed. Everything stays in Obsidian.
Status: early release. Feedback and pull requests are welcome (see CONTRIBUTING.md).
Contents
- Installation
- Quick start
- What you can type
- Commands
- Settings
- Note templates and placeholders
- PDFs
- Example setup
- Data sources
- Network use, files and privacy
- Troubleshooting
- Contributing and development
Installation
Paper Finder needs desktop Obsidian (Windows, macOS or Linux), version 1.13.2 or newer. (Version 0.1.0 of the plugin supports Obsidian 1.8 and is offered automatically on older Obsidian versions.) It isn't available on mobile because it uses desktop-only file access to save PDFs outside the vault.
From Obsidian (once the plugin is listed in the community directory):
- Open Settings → Community plugins → Browse.
- Search for Paper Finder, then install and enable it.
With BRAT (beta releases):
- Install the BRAT plugin.
- Choose Add beta plugin, enter
apratimr7/paper-finder, and enable the plugin.
Manually:
- Download
main.js,manifest.jsonandstyles.cssfrom the latest release. - Put them in
<your vault>/.obsidian/plugins/paper-finder/. - Reload Obsidian and enable Paper Finder under Settings → Community plugins.
Release files are built by GitHub Actions and carry a build provenance attestation. You can verify a download with gh attestation verify main.js --repo apratimr7/paper-finder.
Quick start
- Under Settings → Paper Finder, set Folder to where your source notes should go (default:
Papers). - Under Settings → Hotkeys, give Paper Finder: Find paper a hotkey. No hotkey is set by default. You can also click the search ribbon icon.
- Run Find paper, type something, and wait a moment or press Enter.
- Use ↑ / ↓ to move through the results and Enter to pick one. Clicking a result also picks it.
Results already in your vault are marked in vault. Picking one opens your note instead of creating a duplicate.
What you can type
| Input | Example | What happens |
|---|---|---|
| Free text | maldacena large n limit superconformal |
Fuzzy search in every enabled source. A year in the text boosts papers from that year. |
| DOI | 10.1103/PhysRevD.101.084046, doi:10.… |
Direct lookup |
| arXiv id | 1608.04948, arXiv:1608.04948v2, hep-th/9711200 |
Direct lookup (the version is ignored) |
| arXiv / ar5iv URL | https://arxiv.org/abs/1711.03220v2, …/pdf/…, …/html/… |
Direct lookup |
| doi.org URL | https://doi.org/10.1038/s41586-019-1666-5 |
Direct lookup |
| INSPIRE, OpenAlex, Semantic Scholar URL | https://inspirehep.net/literature/451647 |
Looked up in that database, then completed from the others |
| Publisher URL | https://www.nature.com/articles/s41586-019-1666-5 |
The DOI is read from the URL, or from the page's citation metadata |
Tip: select text in a note and run Find paper from selected text.
Commands
| Command | What it does |
|---|---|
| Find paper | Opens the search box. |
| Find paper from selected text | Searches for the current selection. |
| Open paper link of current note | Opens the current source note's paper, following your Link opened for reading order. |
| Download PDF for current note | Downloads the PDF for the paper in the current note. Works on notes the plugin didn't create, as long as they have a DOI, arXiv id or link. |
| Cite paper | Pick a source note and insert a citation at the cursor, by default [[Note|Maldacena 1998]]. |
Settings
New notes
| Setting | Default | Description |
|---|---|---|
| Folder | Papers |
Where new notes are created. Can contain placeholders, e.g. Papers/{{year}}. Missing folders are created. |
| File name | {{author}} {{year}} - {{title}} |
Gives e.g. Maldacena 1998 - The large N limit of superconformal field theories and supergravity. Characters that aren't allowed in file names are removed, and : becomes -. |
| Note template | (empty) | Path to a template note in your vault, e.g. Templates/Paper.md. |
| Template text | (empty) | A template written directly in the settings. Used when no template note is set. If both are empty, the built-in template is used. |
| Year | Journal publication year | Which year {{year}} means when they differ, e.g. a 1997 preprint published in 1998. The other choice is Year it first appeared (preprint). |
Sources
You can turn each database on or off, and set how many results each one returns (default 8). See Data sources.
Links
| Setting | Default | Description |
|---|---|---|
| Link saved in the note | DOI → arXiv abstract → publisher → INSPIRE → OpenAlex | The first link that exists for a paper becomes {{url}}. |
| Link opened for reading | arXiv HTML → ar5iv → DOI → publisher | Tried in order. arXiv HTML and ar5iv are checked to exist before they're used. |
| Open links in | Web viewer, in a split | Or Web viewer, in a new tab, or System browser. If the core Web viewer plugin is off, the system browser is used. |
| Open the paper after creating its note | on | |
| Open the new note | on |
Both link lists can be reordered by dragging, shortened and extended (the + button). Available link types: arXiv HTML, ar5iv, arXiv abstract, arXiv PDF, DOI, publisher page, open-access PDF, INSPIRE-HEP, OpenAlex, Semantic Scholar.
PDFs
See PDFs below.
Search
| Setting | Default | Description |
|---|---|---|
| Search as you type | on | Off means search only when you press Enter. |
| Typing delay | 800 ms | How long to wait after the last keystroke. Pasted identifiers are searched almost immediately. |
Citations
| Setting | Default | Description |
|---|---|---|
| Citation format | [[{{link}}|{{authorsShort}} {{year}}]] |
Inserted by Cite paper. {{link}} is the link to the note; the other placeholders are read from the note's properties. |
| Cite from folder | (the notes folder) | Which notes Cite paper offers. Use / for the whole vault. |
Property names
These are the frontmatter keys your template uses for title, authors, year, DOI, arXiv id and link. Defaults are title, authors, year, doi, arxiv and url, and matching ignores case. Paper Finder uses them to:
- recognise papers already in your vault;
- build citations;
- open links and download PDFs for existing notes.
If your template uses other keys (e.g. Year of Publication), enter them here.
Optional API access
| Setting | Description |
|---|---|
| Contact e-mail | Sent to OpenAlex and Crossref only if you fill it in. They give identified clients faster, more reliable service (their "polite pool"). |
| OpenAlex API key | Optional. |
| Semantic Scholar API key | Recommended if you enable Semantic Scholar, which strictly rate-limits requests without a key. |
API keys are kept in Obsidian's secret storage, not in your vault, so they don't end up in synced or shared vault folders. (Keys entered in version 0.1.0 are moved there automatically.) The contact e-mail is stored in the plugin's data.json in your vault.
All settings can be found through the search box in Obsidian's settings.
Note templates and placeholders
The built-in template creates a metadata-only note:
---
title: "{{title}}"
authors:
- "{{authors}}"
year: "{{year}}"
venue: "{{venue}}"
doi: "{{doi}}"
arxiv: "{{arxiv}}"
url: "{{url}}"
citekey: "{{citekey}}"
type: article
created: "{{date}} {{time}}"
---
Your own template can contain anything. These placeholders work in templates, file names, folders and the citation format:
| Placeholder | Value |
|---|---|
{{title}} |
Title |
{{author}} |
First author's family name. Particles are kept: 't Hooft, van der Waals. |
{{authorsShort}} |
Smith, Smith and Doe, or Smith et al. |
{{firstAuthor}} |
First author's full name |
{{authors}} |
All authors. Alone on a YAML list line ( - "{{authors}}"), it becomes one list item per author. Elsewhere it's A; B; C. |
{{year}} |
Year (see the Year setting) |
{{venue}} |
Journal or conference |
{{doi}} |
DOI, e.g. 10.1103/physrevd.101.084046 |
{{arxiv}} |
arXiv id, e.g. 1608.04948 |
{{url}} |
The link chosen by Link saved in the note |
{{citekey}} |
First author's family name + year, e.g. Maldacena1998 |
{{abstract}} |
Abstract, as plain text |
{{date}}, {{time}} |
When the note was created: YYYY-MM-DD, HH:mm |
Your frontmatter can't break. Values are YAML-escaped automatically, whether you put quotes around the placeholder or not, so a title such as AdS/CFT: a "review" #2 is safe. The result is parsed before anything is written; if your template produces invalid YAML, the note isn't created and you get an error message instead.
An example template with custom keys is in examples/academic-paper-template.md.
PDFs
PDF download is off by default. Turn it on under Settings → Paper Finder → PDFs.
| Setting | Default | Description |
|---|---|---|
| Download PDF when creating a note | Never | Never, Ask each time or Always. Download PDF for current note works whatever this is set to. |
| Save PDFs outside the vault | off | See below. |
| PDF folder | (next to the note) | See below. |
| PDF file name | (same as the note) | Placeholders as above. |
| PDF sources | arXiv PDF → open-access PDF | Order in which download sources are tried. |
| Embed the PDF in the note | on | Adds ![[file.pdf]] before the note's References heading, or at the end if it has none. |
| PDF property | pdf |
Frontmatter key that gets the link to the PDF. Leave empty for none. |
| Maximum PDF size | 50 MB | Larger files are skipped. |
Where PDFs go:
- Inside the vault (Save PDFs outside the vault off): PDF folder is a vault folder such as
Attachments/PapersorPDFs/{{year}}. Leave it empty to save next to the note. The PDF can be embedded and opened in Obsidian's PDF viewer. - Outside the vault (Save PDFs outside the vault on): PDF folder must be an absolute path such as
~/Papers,/data/papersorC:\Users\me\Papers. This is useful to keep large files out of a synced vault. The note gets afile://link in the PDF property. Obsidian can't embed files outside the vault, so there's no embed.
What is downloaded:
- Only free, legal copies: the arXiv PDF, or an open-access PDF that OpenAlex knows about.
- A file is only saved if it really is a PDF; it must start with
%PDF-. Paywall, login and error pages are rejected. - Existing files are never overwritten. A second copy is saved as
name (2).pdf.
Example setup
A physics-oriented setup with custom property names, PDFs outside a cloud-synced vault, and one-click citations:
| Setting | Value |
|---|---|
| Folder | 2 - Source Material/Academic Papers |
| File name | {{author}} {{year}} - {{title}} |
| Template text | the contents of examples/academic-paper-template.md |
| Property names | Title, Authors, Year of Publication, DOI, arXiv, URL/Source/DOI |
| Cite from folder | 2 - Source Material |
| Download PDF | Ask each time |
| Save PDFs outside the vault | on |
| PDF folder | ~/Papers |
| Link opened for reading | arXiv HTML → ar5iv → DOI |
Data sources
| Source | Used for | Default |
|---|---|---|
| OpenAlex | Broad search across all fields (about 250 million works); open-access PDF links | on |
| arXiv | Preprints, including the latest versions. Requests are limited to one every three seconds, as arXiv asks. | on |
| INSPIRE-HEP | High-energy physics, gravitation and cosmology, with curated journal references | on |
| Crossref | Publisher metadata and DOI lookup | on |
| Semantic Scholar | Extra coverage for vague queries. Strictly rate-limited without an API key. | off |
How results are ranked: results are combined with reciprocal-rank fusion. A paper that several sources rank highly comes first.
How duplicates are merged: results are merged by DOI or arXiv id. Without either, they're merged by title, if the title is long enough and the years agree.
Where merged details come from: each field comes from the source that is best at it, so a paper gets the same file name however you found it:
- titles from Crossref;
- the original journal DOI from INSPIRE;
- journal names from Crossref or INSPIRE.
All five services are free and need no account. Thank you to arXiv for use of its open access interoperability.
Network use, files and privacy
As required by Obsidian's developer policies, here is what the plugin accesses.
Network use. Paper Finder only connects to the internet when you search, create a note, open a link or download a PDF. It contacts:
- the sources you have enabled:
api.openalex.org,export.arxiv.org,inspirehep.net,api.crossref.org,api.semanticscholar.org. Your search text is sent to them to find papers. arxiv.organdar5iv.labs.arxiv.org, to check whether an HTML version exists and to download arXiv PDFs.- a publisher's page, only when you paste its URL, to read the DOI from it.
- the host of an open-access PDF, only when downloading that PDF.
Requests go through Obsidian's requestUrl. Nothing identifying is sent unless you add a contact e-mail or API key yourself.
Files outside the vault. These are written only if you turn on Save PDFs outside the vault, and only PDF files inside the folder you chose. The plugin checks that every path stays inside the folder, and never overwrites existing files.
No telemetry, no ads, no account, no payment. The plugin collects no usage data.
Security
- No runtime dependencies.
main.jsis built only from this repository's source code. - Three exact-pinned build tools (Obsidian's type definitions, TypeScript and esbuild), installed with install scripts disabled. Package signatures are verified in CI.
- Text from APIs and web pages is never treated as HTML. It's converted to plain text, and the plugin never uses
innerHTML. - Only
http(s)links are stored or opened. - API responses are validated at runtime. Every field is checked to have the expected type before it is used.
- API keys are kept in Obsidian's secret storage, never in the vault.
- The code passes Obsidian's recommended lint rules (
eslint-plugin-obsidianmd, including type-checkedtypescript-eslintrules) with no warnings.
Please report vulnerabilities privately. See SECURITY.md.
Troubleshooting
| Problem | What to do |
|---|---|
| "Crossref failed (HTTP 429)" or another source failed | That database is rate-limiting you. The other sources still return results. Wait a moment, or add a contact e-mail under Optional API access. |
| Semantic Scholar always fails | It needs an API key for reliable use; request one or leave it off. |
| The right paper isn't in the results | Add an author's surname or the year, paste the DOI or arXiv id, or enable more sources. Lecture notes and talks often aren't indexed as papers. |
| Links open in the browser instead of a split | Enable the core Web viewer plugin under Settings → Core plugins. |
| "No PDF saved: not a PDF" | The paper has no free PDF that the plugin knows of. Publisher pages behind a paywall are rejected on purpose. |
| A paper isn't recognised as already in the vault | Check that Property names match the keys in your notes, and that the note has a DOI, arXiv id or link. |
Contributing and development
Pull requests are welcome. See CONTRIBUTING.md for the setup, code layout and rules. In short:
npm ci # installs exactly the lockfile; install scripts are disabled
npm run dev # watch mode
npm run build # type-check + production build
npm run test:live # search core against the real APIs
npm run test:note # template, YAML and file-name checks (offline)
npm run test:pdf # PDF and file-safety checks
License
MIT © Apratim Rastogi