Skip to main content

citation-integrity-check

PyPI Python License: MIT

Verify answer citations refer to supplied source ids and that cited sources actually support the claims. Zero runtime dependencies.

Python port of @mukundakatta/citation-integrity-check. The JS sibling has the original API; this README sticks to the Python surface.

Install

pip install citation-integrity-check

Usage

from citation_integrity_check import verify

sources = [
    {"id": "1", "text": "Photosynthesis converts light into chemical energy in plants."},
    {"id": "abc123", "text": "Chlorophyll absorbs red and blue wavelengths of light."},
]
answer = (
    "Plants use photosynthesis to convert light into energy [1]. "
    "Chlorophyll absorbs red and blue light [id:abc123]."
)

result = verify(answer, sources)

result.ok            # True if no missing ids and no unsupported claims
result.missing       # list[str]    -- cited ids that don't exist in sources
result.unsupported   # list[Claim]  -- sentences with no valid supporting citation
result.coverage      # float in [0, 1] -- fraction of sentences with a valid citation

Citation forms

Two markers are recognized inside the answer:

Form Resolves to
[1] sources[0] (1-based index) and source.id == "1"
[id:abc] source.id == "abc"

Anything else inside brackets (like [Note]) is ignored, so stylistic prose doesn't count as a citation.

How "unsupported" is decided

A sentence is unsupported when any of these is true:

  • It has no citation marker at all (reason="no_citation").
  • All cited ids are missing from sources (reason="missing_source").
  • The cited source's text doesn't share enough non-stopword tokens with the sentence (reason="insufficient_overlap").

Token-overlap is |claim_tokens & source_tokens| / |claim_tokens|, with a small built-in stopword list. The threshold is tunable:

verify(answer, sources, support_threshold=0.5)  # stricter

API differences from the JS sibling

  • Returns a CitationResult dataclass with unsupported claims (per-sentence) instead of the JS unused ids list.
  • Adds the [id:foo] named-citation form alongside numeric [N].
  • Adds the token-overlap support_threshold to verify the cited source actually mentions the claim.

See the JS sibling's README for the full design notes.

Metadata

Release files for citation-integrity-check 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for citation-integrity-check 0.1.0
File Size Uploaded Provenance Actions
citation_integrity_check-0.1.0.tar.gz 7.7 kB Without provenance Details for citation_integrity_check-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for citation-integrity-check 0.1.0
File Interpreter ABI Platform Actions
citation_integrity_check-0.1.0-py3-none-any.whl Python 3 none any Details for citation_integrity_check-0.1.0-py3-none-any.whl

Total release size: 15.4 kB

Release files / citation_integrity_check-0.1.0.tar.gz

Download URL citation_integrity_check-0.1.0.tar.gz
Size 7.7 kB
Tags Source
SHA-256 checksum
How to use checksums
1ad29342327a1db1ad5596bfa8164e11eccedacef876e428693a17c51b4a186d
BLAKE2b-256 checksum
How to use checksums
d32b01bd3925f79165b4076ca03a429838bf8d0c3b211d0019110d8e95b5ef11
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.14.4

Release files / citation_integrity_check-0.1.0-py3-none-any.whl

Download URL citation_integrity_check-0.1.0-py3-none-any.whl
Size 7.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
ee8184e312640fa644487de43d6c78e6b8b244b9e81fbccb2318271c6d37dec2
BLAKE2b-256 checksum
How to use checksums
4ae876cb2186ff8c3400470be02da7c528e11e58c57744168bb1c44748ff288e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.14.4

Security for citation-integrity-check 0.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page