refkit 0.0.2

Fast Python citation parsing, rendering, and BibTeX editing backed by Rust

refkit

refkit reads BibTeX, BibLaTeX, and Hayagriva YAML, renders CSL citations, and edits raw BibTeX documents from Python.

Install

pip install refkit

refkit is pure Python and depends on the exact matching refkit-core release. refkit-core contains the Rust/PyO3 extension as refkit_core._refkit_core, including PyEmscripten wheels for the Python 3.14 Pyodide runtime.

refkit supports CPython 3.11 through 3.14. Native wheels from refkit-core use the Python 3.11 stable ABI.

Render A Citation

import refkit as rk

library = rk.Library.parse_bibtex(
    """
@article{doe2024,
  author = {Doe, Jane},
  title = {Fast Citations},
  journal = {Journal of Citation Tests},
  year = {2024}
}
@book{roe2022,
  author = {Roe, Richard},
  title = {Batch References},
  publisher = {Example Press},
  year = {2022}
}
"""
)
style = rk.Style.load("apa")
doc = rk.Document(library, style, locale="en-US")

rendered = doc.render(
    [
        rk.Citation("intro", "doe2024"),
        rk.Citation(
            "detail",
            rk.CitationGroup([rk.Cite("doe2024", locator="12", label="page"), "roe2022"]),
        ),
    ]
)

print(rendered["intro"].text)
print(rendered["detail"].text)
print(rendered.bibliography.text)

Expected output:

(Doe, 2024)
(Doe, 2024, p. 12; Roe, 2022)
Doe, J. (2024). Fast Citations. Journal of Citation Tests.
Roe, R. (2022). Batch References. Example Press.

Document.render renders the whole citation document at once. It returns RenderedDocument, where rendered["intro"] and rendered["detail"] are named citation outputs and rendered.bibliography is the cited bibliography for those citations. Cite names one citation item. CitationGroup renders several items as one citation. Citation(id, group) gives that rendered citation a stable lookup name. Citation ids must be unique inside one Document.render call.

For one-off scripts, pass a bibliography path directly:

rk.cite("refs.bib", "doe2024", style="ieee").text
rk.full_bibliography("refs.bib", style="chicago-author-date").html

Use Library.parse_bibtex, Library.parse_yaml, and Document when the bibliography source is already in memory or when several citations share the same library and style.

Capabilities

Capability	Python surface
Read normalized bibliography data	Library.read, Library.parse_bibtex, Library.parse_yaml
Render citations	Document.render, Citation, Cite, CitationGroup, cite
Render bibliographies	Document.cited_bibliography, Document.full_bibliography, full_bibliography
Load styles and locales	Style.load, Style.from_path, Style.from_xml, Locale.load
Inspect entries	mapping access, keys, get, get_many, select, project, to_dicts
Edit raw BibTeX	BibDocument.read, BibDocument.parse, field assignment, write
Inspect rendered output	Rendered.text, Rendered.html, Rendered.tree

Input Formats

API	Input	Result
Library.read(path)	.bib	Normalized citation library from BibTeX or BibLaTeX.
Library.read(path)	.yaml, .yml	Normalized citation library from Hayagriva YAML.
Library.parse_bibtex(source)	BibTeX or BibLaTeX string	Normalized citation library.
Library.parse_yaml(source)	Hayagriva YAML string	Normalized citation library.
BibDocument.read(path)	.bib	Raw document model with comments, preambles, strings, failed blocks, order, spans, and editable fields.
Style.load(name)	Bundled style name such as apa	CSL style for rendering.
Style.from_path(path)	Independent CSL XML file	CSL style for rendering.
Style.from_xml(xml)	Independent CSL XML string	CSL style for rendering.
Locale.load(code)	Bundled locale code such as en-US	Locale object for rendering.

Hayagriva YAML is a mapping from citation keys to entry mappings:

doe2024:
  type: Article
  author: Doe, Jane
  title: Refkit for Bibliographies
  date: 2024
  parent:
    type: Periodical
    title: Journal of Citation Systems
    volume: 12

library = rk.Library.parse_yaml(
    """
doe2024:
  type: Article
  title: Refkit for Bibliographies
  date: 2024
"""
)

Inspect A Library

Library is the normalized citation database. Use it for rendering, selectors, mapping access, and bulk export.

library = rk.Library.read("refs.bib")

print(library.keys())
print(library.project(["key", "title", "doi", "volume"]))
print(library.to_dicts())

Library.select accepts Hayagriva selector strings:

for entry in library.select("article > periodical[volume]"):
    print(entry.key, entry.title, entry.parents[0].title)

Edit Raw BibTeX

BibDocument preserves the raw .bib structure that normalized rendering does not need: comments, preambles, string definitions, failed blocks, order, and source spans.

raw = rk.BibDocument.read("refs.bib")
raw.entries["doe2024"].fields["title"].value = "Corrected title"
raw.write("refs.bib")

Direct map lookup requires one matching entry key and one matching field name. unique_keys() returns one key per name. occurrence_keys() returns keys in source order, including duplicates. When a file contains duplicates, choose the source-order occurrence explicitly:

raw = rk.BibDocument.read("refs.bib")

second_entry = raw.entries.get_all("doe2024")[1]
second_entry.fields.get_all("title")[0].value = "Corrected title"
raw.write("refs.bib")

Inspect Rendered Output

Document.render, Document.cited_bibliography, Document.full_bibliography, cite, and full_bibliography return rendered values.

rendered = doc.render([rk.Citation("intro", "doe2024")])
citation = rendered["intro"]

print(citation.text)
print(citation.html)
print(citation.tree)

Rendered.tree returns structured nodes for text, links, element metadata, transparent citation fragments, and bibliography entries.

Use With Polars

Install polars-refkit when BibTeX source lives in a dataframe and the result should stay in a Polars query plan.

import polars as pl
import polars_refkit as prk

df = pl.DataFrame(
    {
        "bibtex": ["@article{doe2024, title={Fast Citations}, year={2024}}"],
        "key": ["doe2024"],
        "keys": [["doe2024"]],
    }
)

out = df.select(
    citation=pl.col("bibtex").refkit.cite(pl.col("key")),
    each_citation=pl.col("bibtex").refkit.cite_each(pl.col("keys")),
    grouped_citation=pl.col("bibtex").refkit.cite_group(pl.col("keys")),
    count=pl.col("bibtex").refkit.entry_count(),
    keys=pl.col("bibtex").refkit.keys(),
    entries=pl.col("bibtex").refkit.entries(),
)

Development

uv sync --all-packages --group dev
(cd packages/refkit-core && uv run maturin develop)
uv run pytest packages/refkit/tests --no-cov

The workspace also provides:

make lint
make typecheck
make test
make rust
make build

License

refkit is licensed under the Apache License, Version 2.0, available in LICENSE. See NOTICE for upstream citation and bibliography component acknowledgements.