bibsync 0.3.2

Synchronize BibTeX files from citation keys in LaTeX sources

bibsync

bibsync synchronizes BibTeX files from citation keys in LaTeX sources. It is inspired by adstex, with provider support for both NASA ADS and InspireHEP.

The primary workflow is to cite papers by identifier, especially arXiv ID:

\citep{2404.14498}
\citet{arXiv:2312.00752}

Then check the bibliography:

bibsync main.tex -o references.bib

To update the file, add --fix:

bibsync --fix main.tex -o references.bib

bibsync scans the TeX file, resolves missing identifier-like citekeys through NASA ADS and/or InspireHEP, rewrites provider BibTeX entries so the citekey stays the key used in TeX, and reports whether the output .bib file is current. With --fix, it writes the merged bibliography.

When a citekey cannot be resolved, the command prints the key together with a reason and the likely fix. For example, an unsupported citekey is reported as an identifier-format problem, while an arXiv ID or DOI that the provider cannot find is reported as a provider miss.

Installation

Install the Rust CLI from crates.io:

cargo install bibsync

Install the Python package from PyPI:

pip install bibsync

The PyPI package includes Python bindings and installs the bibsync command. Python 3.12 or newer is required.

You can also download a pre-built binary from the GitHub releases page. Pick the archive for your platform, extract it, and place the bibsync executable somewhere on your PATH.

To compile from source:

git clone https://github.com/isaac-cf-wong/bibsync.git
cd bibsync
cargo build --release

The compiled binary is written to target/release/bibsync on Unix-like systems or target\release\bibsync.exe on Windows.

Citing

If bibsync contributes to a scientific publication, please cite it using the Zenodo record:

https://doi.org/10.5281/zenodo.20422622

Citation formats including BibTeX can be exported directly from the Zenodo page. For convenience, you can also use:

@software{wong2026bibsync,
  author  = {Wong, Isaac C. F.},
  title   = {bibsync: A Rust package to automatically resolve, synchronize, and validate LaTeX citations across BibTeX databases},
  version = {v0.3.2},
  year    = {2026},
  month   = may,
  doi     = {10.5281/zenodo.20422622},
  url     = {https://doi.org/10.5281/zenodo.20422622}
}

Providers

By default bibsync tries NASA ADS first and InspireHEP second:

bibsync --fix main.tex -o references.bib --provider auto

NASA ADS requires an API token:

export ADS_API_TOKEN="..."

You can choose a single provider:

bibsync --fix main.tex -o references.bib --provider ads
bibsync --fix main.tex -o references.bib --provider inspire

InspireHEP supports arXiv IDs and DOIs. NASA ADS supports arXiv IDs, DOIs, and ADS bibcodes.

Python API

bibsync can also be installed from PyPI:

pip install bibsync

The PyPI package provides Python bindings backed by the Rust implementation:

import bibsync

report = bibsync.sync_files(
    ["main.tex"],
    output="references.bib",
    provider="inspire",
    check=True,
)

It also installs the bibsync command. The command delegates to the same Rust CLI implementation as the Cargo-installed binary, so command-line behavior is kept in one place.

Existing Bibliographies

If the TeX source contains \bibliography{references}, bibsync can discover references.bib automatically:

bibsync --fix main.tex

Additional read-only bibliographies can be used to avoid duplicating entries:

bibsync --fix main.tex -o references.bib -r shared.bib software.bib

Use --merge-other to copy matching entries from those read-only files into the main output file.

To update a bibliography in place, pass a single .bib file:

bibsync --fix references.bib --force-regenerate

Existing input files are validated before resolution. A missing single .bib input, --other bibliography, or --ignore-file is reported as an error with the path that could not be read. Existing bibliography files are also parsed strictly, so malformed BibTeX reports the file and the approximate failing entry instead of being treated as an empty or partial bibliography.

Update Behavior

By default bibsync leaves published entries untouched. Only entries that look like unpublished preprints — those with an archivePrefix or eprinttype field but no journal field — are re-queried to check whether they have been published. If so, the entry is updated; otherwise it is preserved.

Flag	Behavior
(default)	Re-check preprints; skip published entries
--no-update	Skip all existing entries
--update-all	Re-resolve all existing entries
--force-regenerate	Re-resolve and overwrite all existing entries

Ignoring Entries

To exclude specific entries from all resolution — for example, books or theses you have curated by hand — list their citekeys in a .bibsyncignore file:

# .bibsyncignore
knuth1997art
smith2024thesis

Pass the file with --ignore-file:

bibsync --fix main.tex -o references.bib --ignore-file .bibsyncignore

Cache

Use --cache to avoid repeated provider API calls:

bibsync --cache main.tex -o references.bib
bibsync --fix --cache main.tex -o references.bib

The cache stores provider records and mappings from arXiv IDs or DOIs to the provider's canonical record ID. Preprint entries that are re-checked for publication always bypass the cache and fetch a fresh result, then write it back. Use --refresh-cache to force a fresh fetch for all entries:

bibsync --fix --refresh-cache main.tex -o references.bib

Override the cache location with --cache-dir DIR.

If a cache file is corrupt, bibsync reports the exact cache path and asks you to refresh or remove the bad cache entry. Provider request failures include the provider and citekey or batch being resolved.

Pre-commit

The repository includes .pre-commit-hooks.yaml, so other projects can use bibsync as a pre-commit hook.

Use the pre-built binary hook for faster installs:

repos:
    - repo: https://github.com/isaac-cf-wong/bibsync
      rev: v0.3.2
      hooks:
          - id: bibsync-bin
            args: [--cache, --provider, inspire, --output, references.bib]

The binary hook downloads a platform-specific archive from the GitHub release matching rev and caches it under pre-commit's cache directory. The source hook is available for Linux x86_64/aarch64, macOS x86_64/aarch64, and Windows x86_64. The source hook is still available, but it compiles the Rust crate during hook installation:

repos:
    - repo: https://github.com/isaac-cf-wong/bibsync
      rev: v0.3.2
      hooks:
          - id: bibsync
            args: [--provider, inspire, --output, references.bib]

By default, the hook checks whether the bibliography is current and fails without writing changes. To let the hook update files, add --fix to the hook args:

repos:
    - repo: https://github.com/isaac-cf-wong/bibsync
      rev: v0.3.2
      hooks:
          - id: bibsync-bin
            args: [--fix, --cache, --provider, inspire, --output, references.bib]

To skip manually curated entries, add --ignore-file:

repos:
    - repo: https://github.com/isaac-cf-wong/bibsync
      rev: v0.3.2
      hooks:
          - id: bibsync-bin
            args:
                [
                    --fix,
                    --cache,
                    --provider,
                    inspire,
                    --output,
                    references.bib,
                    --ignore-file,
                    .bibsyncignore,
                ]

For a project-local hook while developing bibsync itself:

repos:
    - repo: local
      hooks:
          - id: bibsync
            name: bibsync
            entry: cargo run -- --fix --provider inspire --output references.bib
            language: system
            files: \.tex$