Arguably the most important feature of this template is its ability to automatically generate citations from simple identifiers. This makes it easy to cite and maintain large lists of publications.
The template is able to do this thanks to Manubot, a suite of tools that (among many other things) lets you automatically generate a citation with full details from just a short identifier, like doi, url, isbn, pmc, pmcid, pubmed, arxiv, and many many more.
{% embed url="https://manubot.org/" %}
First let's define some consistent terminology to make things easier to explain:
- source - A paper, book, article, web page, film, or any other published item you want to cite.
- metasource - A single item that lists multiple sources, like how an author's ORCID number can be used to get a list of their published works.
- citation - Full, detailed information about a source, like title, author(s), publisher, publish date, URL, etc.
For most content on your site, you just need change the contents of the appropriate file. Citations have a special additional step. When you add new sources or metasources to be cited, the template has to run a special "cite process" to generate your full citations.
At a high level, here's how it all works:
%%{
init: {
"fontFamily": "Arial",
"stroke": "none",
"flowchart": {
"curve": "monotoneX"
}
}
}%%
graph LR
subgraph input
sources([sources*.yaml])
subgraph metasources
orcid([orcid*.yaml])
etc([etc...])
end
end
subgraph cite ["cite process"]
expand([expand to sources])
merge([merge sources])
manubot([Manubot])
preserve([preserve input fields])
end
subgraph output
citations([citations.yaml])
title[titles, authors, etc...\ninput fields]
end
subgraph site
display[citation component\nlist component]
end
sources --> merge
orcid --> expand
etc --> expand
expand --> merge
merge --> manubot
manubot --> preserve
preserve --> citations
citations --> display
classDef node fill:#38bdf8,stroke:none,color:white
classDef cluster fill:#e0f2fe,stroke:white,stroke-width:3px
classDef input fill:#e0f2fe
classDef cite fill:#ffedd5
classDef output fill:#dcfce7
classDef title fill:none,color:black
class input, input
class cite, cite
class output, output
class title, title
class display, title
- Input your sources and metasources in
/_datafiles, e.g.sources-2020.yaml,orcid-students.yaml, etc. - The cite process runs – either on GitHub or locally – to convert your sources and metasources to full citations.
- The cite process outputs a single
citations.yamlfile in/_data. - Display and filter citations on your site with the list and citation components.
{% hint style="info" %}
Do not edit citations.yaml! It will get overwritten each time the cite process runs. If you need to manually input or correct things, see below.
{% endhint %}
The cite process in detail...
- Each source/metasource file gets processed by the appropriate cite plugin (see
/_cite/plugins) based on filename prefix. - In metasource files, each list entry gets expanded into a list of regular sources. Any fields you put in the original entry get copied to each source in the expanded list.
- In source files, each list entry stays as-is.
- Metadata about the cite process is attached to each source, like which input file it originated from and which cite plugin it ran with.
- A full list of regular sources is compiled, with duplicates merged together by
id. - When merging two entries, if the same field is present in both, the one that comes later in the full list overwrites the one that comes earlier (because there cannot be two keys with the same name on an dict/object).
- Manubot generates full citation details for each source that has an
id. - Any field originally on each source is preserved.
You can mix and match as many sources and metasources as you want, and display them however and wherever you'd like! For example, you may want to have a "CV" page that lists all of the papers under your PI's ORCID, then reserve your "Research" page for just a few special papers by various members in your lab that you want to highlight.
Filename must start with sources.
{% code title="/_data/sources.yaml" %}
- id: doi:10.1098/rsif.2017.0387
- id: pubmed:29424689
- id: pmc:PMC5640425
- id: arxiv:1806.05726
# ...more sources{% endcode %}
| Parameter | Description |
|---|---|
id | Identifier for the source that Manubot can understand and cite. If Manubot is unable to generate a citation for this ID, the template will log an error message and exit with an error code. |
Optionally, you can manually pass extra "rich" details that the template can display nicely. Manubot can't automatically determine these.
{% code title="/_data/sources.yaml" %}
- id: arxiv:1806.05726
type: paper
description: Lorem ipsum _dolor sit amet_.
image: https://publisher.com/striking-image-for-your-paper.jpg
buttons:
- type: source
link: https://github.com/your-lab/some-repo
- type: website
text: My Personal Website
link: http://some-website.com/
tags:
- biology
- big data
- medicine
repo: your-lab/some-repo
# ...another source{% endcode %}
| Parameter | Description |
|---|---|
type | The type of the source. Determines the icon to show. See /_data/types.yaml for what types are built-in or to add your own. |
description | Brief description of the source. Can contain Markdown. |
image | URL to a striking image for the source. Highly recommended. Displays as a thumbnail next to the citation details. |
buttons | List of buttons to show underneath the citation details. |
tags | List of tags to show underneath the citation details. |
repo | GitHub repository to automatically fetch additional tags from. |
{% hint style="info" %} Always provide a good thumbnail for your publications. Use a figure from the source, or if there are none, a journal logo or issue cover. Here are some best practices for making a good thumbnail image. {% endhint %}
All fields you attach to a source (or metasource, see below) get passed through to the generated citation untouched. This allows you to manually input or correct details of a citation.
{% code title="/_data/sources.yaml" %}
# manually correct specific citation detail
- id: pmc:PMC5640425
publisher: Cold Spring Harbor
# manually provide all citation details
- title: Some Publication Title
authors:
- "**Steve McQueen**"
- Lightning McQueen
publisher: bioRxiv
date: 2021-01-01
link: biorxiv.org/1234
# attach an arbitrary field
- id: pmc:PMC5640425
some-field: 123{% endcode %}
| Parameter | Description |
|---|---|
title / authors / publisher / date / link | Basic citation information normally returned from Manubot and displayed by the citation component. Date should be in YYYY-MM-DD format. Authors can contain Markdown. |
If you don't provide an id, Manubot has nothing to cite and so doesn't run. You'd only want to do this if manually providing all the citation details manually. This defeats the main benefit of the template, but is sometimes necessary.
You can also attach arbitrary fields. The template won't explicitly use them, but you could use them to filter citations with the list component.
ORCID is the recommended way to retrieve a complete set of published works for an author.
Filename must start with orcid.
{% code title="/_data/orcid.yaml" %}
- orcid: 0000-0002-4655-3773
jane-doe: true
# ...another author{% endcode %}
Each ORCID gets expanded into a full list of regular sources with ids. Any fields you put in the original entry get copied to each source in the expanded list. This applies to the other types of metasources below as well.
To conveniently and explicitly associate an ORCID with a person for filtering with the list component, you can add a unique field for that person like in the example above and use a filter like this:
{% raw %}
{% include list.html data="citations" component="citation" filters="jane-doe: true" %}
{% endraw %}Why do the filter like that?
Why have the unique author be the key (e.g. jane-doe: true) instead of the value (e.g. member: jane-doe)? If you do the latter, and two people are authors on the same source, only one of the authors will be kept (whichever is listed last). This is a result of behavior described elsewhere on this page (duplicate sources merged into one, cannot have duplicate keys on dict/object).
Because the cite process merges duplicate sources by id, you can also use the manual override method above to manually correct sources returned from metasources. You can also use the special remove field to ignore a source returned from a metasource, discarding it from the citations output. Example:
{% code title="/_data/sources.yaml" %}
# some paper returned by an ORCID with a wrong title
- id: doi:123/456
title: Correct Title
# some paper returned by an ORCID that you don't want to keep at all
- id: pubmed:123456
remove: true{% endcode %}
{% hint style="info" %} If a metasource returns a source with a type of identifier that Manubot doesn't know how to cite, that source will be ignored. When this happens, the cite process prints a warning, not a critical error, so you'll have to check your CLI output or GitHub Actions logs. {% endhint %}
See ORCID above for general metasource functionality.
Uses NCBI eutils to search PubMed for terms. This is a brittle way to select an author's papers, very vulnerable to false positives. ORCID is recommended.
Filename must start with pubmed.
{% code title="/_data/pubmed.yaml" %}
- term: "Greene, C[Author] NOT Greene CE[Author]"{% endcode %}
See ORCID above for general metasource functionality.
Unfortunately, Google does not provide APIs for many of its services, and that includes Scholar. Luckily there is a 3rd-party API to access it, SerpAPI. First you'll have to sign up and get an API key. Then, if running the cite process on GitHub, make a new repository secret named GOOGLE_SCHOLAR_API_KEY with your API key as its value, or if running locally, put the same name/value in a .env file.
Filename must start with google-scholar.
{% code title="/_data/google-scholar.yaml" %}
- gsid: ETJoidYAAAAJ{% endcode %}
Metasources like ORCID update over time as you publish new sources. To accommodate this, the template tries to automatically re-run the cite process periodically. In your repo's "
{% hint style="info" %} GitHub imposes certain limitations on scheduled workflows. You may have to manually enable the workflow, or re-enable it after a while. {% endhint %}
When the workflow runs, it will get the latest sources associated with your metasources, generate citations like normal, and open a pull request to let you review the changes before publishing them.
{% hint style="info" %} Due to another GitHub limitation, if you want to see a live preview link for this pull request, you'll have to close (not merge) and reopen it. That should trigger the "build preview" workflow. {% endhint %}
If you have many sources, generating the citations for all of them can take a while. To save time, the cite process keeps a time-to-live cache for Manubot/ORCID/etc. network requests. If a cached response hasn't expired, it is used instead of making a new request. To clear this, simply delete the cache file in _cite. You can also customize the default expire times in the Python code in _cite.