Pangur docsReferences and connections
Pangur can read the works a stored PDF cites and tell you which are already in your library. This page explains how that scan works, what each status means, how to import the rest, and what the Connections graph shows.
Pangur can read the works that a stored PDF cites and tell you, for each one, whether it is already in your library, whether it can be imported, or whether it is an archival manuscript. These are the item's References. Pangur also keeps a record of which works in your library cite which others, and draws that as the item's Connections. Both are sections of the item's Study tab, and they answer different questions. References lists every work one PDF cites, most of which are usually outside your library, with buttons to bring them in. Connections shows only the citation links among works you already hold, as a graph and a list.
How Pangur finds a PDF's references
When you open References, Pangur scans the item's stored PDF. If the item has several PDFs, you choose one from the Attachment list, and an item with no stored PDF has nothing to scan.
The scan combines two sources of evidence. The first is the citation graph of OpenAlex, an open index of scholarly works, which lists the works a published article cites whenever OpenAlex knows the article's DOI. The second is the PDF's own printed bibliography and footnotes, which Pangur's server parses from the text. Pangur then looks up on OpenAlex any entries it could not match, and checks every result against your library. A work cited under several footnotes appears once, carrying all of its note numbers. Pills on each row say where the evidence came from ("text", "graph" or "text + graph"), and works found only in the OpenAlex graph are listed separately under Also cited.
The first results arrive quickly, and Pangur then refines them in the background over the next few minutes while the tab stays open. Results are kept for seven days for that version of the file, so opening References again is fast. Rescan forces a fresh scan. You can start up to six new scans a minute, after which Pangur asks you to wait a little.
Neither the parsing nor the OpenAlex lookups use a language model. An administrator can switch on an additional model pass for difficult footnoted PDFs, but it is off by default.
What each status means
Each reference has one of four statuses:
| Status | Meaning | What you can do |
|---|---|---|
| in library | The work is already in your library. The pill names how it was matched: DOI, ISBN, title, or "title (approximate)". | Open it |
| importable | Pangur identified the work on OpenAlex, or found a DOI or ISBN for it. | Import it |
| primary source | The citation is to an archival manuscript, identified by repository and shelfmark. It is never importable, because it is an object in an archive. | View facsimile if a facsimile is linked; otherwise open it at the repository |
| unresolved | Pangur could not identify the work from the citation's text. | Search for it, or Add manually |
A primary source is not the same as an unresolved reference. Pangur knows exactly what a primary source is, but an archive's object cannot be imported into a library. If the manuscript has been digitised, you can link the facsimile, after which View facsimile opens the cited leaf.
Importing the works a PDF cites
To import several references at once, select them (or Select all) and choose Import selected, or choose Import all. Before the batch starts you can choose where the works go under File into collection, or leave them in No collection (library root), and tick fetch open-access PDFs if Pangur should look for a free PDF of each work.
Pangur imports the works in one background batch and reports its progress as it goes. If an outside service limits the rate, the batch pauses and then continues on its own. When it finishes, it lists the works that were already in your library and any that failed, and Find PDFs looks for PDFs of the works imported without one. A batch takes up to 100 works. How Pangur looks for a PDF is described in Where PDFs come from.
Linking citations inside the PDF
In the PDF viewer, the Citations switch links each in-text citation to its entry in the detected reference list. Select a citation marker to see the entry it points to, and press Escape to close it. The switch is off by default.
Finding the work a citation names
Select the text of a reference in a PDF and choose Find work in the selection bar. Pangur checks the PDF's scanned references first, then scans the selection itself, then searches for it as something to import, and shows what it found. If nothing matches, it suggests selecting the whole reference and trying again.
The Connections graph
Connections draws the citation links between the item and other works in your library. Arrows run from the citing work to the cited one, and one line joins each pair, with a count when there are several citations. Choose 1 hop for the works that this item cites or that cite it, or 2 hops to go one step further. The graph shows up to the 60 closest works and says how many more lie further out. A list of the same links ("cites", "cited by") always appears beside the graph, and on a narrow screen the list is the only view.
Pangur records a link whenever it reads a stored PDF's references and finds the cited work in your library. This happens in the background after a file is read, when you open References, and when you import the works that a parent work cites. A work whose text Pangur has not read yet has no links, and Connections says so.
In a project connection limited to one collection, Connections shows only works inside that collection.
Who can scan and import
In an organisation's project, a member with the viewer role can open an item's References and Connections, which scans the PDF if it has not been scanned yet, but cannot import the works it cites (Project members and roles).
From an assistant or the command line
An assistant connected to Pangur can read a PDF's references with the MCP tool get_references, scan a document outside your library with scan_references, and import references with import_references. Over MCP Pangur runs no model, so where a bibliography needs structuring that the parser could not manage, it returns the text for the assistant's own model to structure. From a terminal, pangur refs KEY lists an item's references, pangur refs scan scans a file or web address, and pangur refs import imports them. See MCP tools and the command-line reference.
The reading room
The reading room shows one source or a sequence of them as continuous reflowed text. This page explains how to open it, move through a sequence, search it, and set the text size, line width and font you prefer.
Link a digitised facsimile
You can connect a manuscript or book in your library to a library's or archive's online digitisation, so that its leaves open in Pangur's reader and transcription desk. This page explains what you need, how the link is kept, and what happens when the images do not load.