Skip to main content

UnifyVcf

Unifies the separate structural call files a sequencing pipeline produces into the single VCF Geneyx reads as an SV VCF.

The files it unifies are:

  • SV – structural variants VCF
  • CNV – copy number variants VCF
  • Repeats – tandem repeat variants VCF
  • ROH – Regions of Homozygosity (DRAGEN BED, or an already-converted VCF)

It does not only concatenate them. Each caller writes slightly different files, and Geneyx needs specific fields to read a record properly — so records are rewritten on the way out and the result is sorted and bgzipped. Run the entry point that matches the pipeline your VCFs came from.

This package is the single implementation behind both the customer-facing scripts in geneyx.analysis.api and the Geneyx primary pipeline, so both produce the same unified file.

Installation

pip install UnifyVcf

Python 3.9–3.13 on Linux, macOS or Windows, with no external tools required and no dependencies outside the standard library. Sorting and BGZF compression are pure Python; repeat filtering uses bedtools when it is installed and computes the same intersection itself when it is not.

Wheels are published per platform and Python version because the implementation ships compiled — see How this is distributed. If pip install reports no matching distribution, your platform or Python version is outside the published set; tell us which and we will add it.

Command line

dragen-unify-vcf  -o OUTPUTPATH [-s SVPATH] [-c CNVPATH] [-r REPEATPATH] [-d ROH_BED_FILE]
ont-unify-vcf     -o OUTPUTPATH [-s SVPATH] [-c CNVPATH] [-r REPEATPATH] [-modify]
pacbio-unify-vcf  -o OUTPUTPATH [-s SVPATH] [-c CNVPATH] [-r FULLREPEATPATH] [-b REPEATLOCATIONSBEDFILEPATH]
Argument Meaning
-o Path to the unified VCF, including its name. The output is compressed, so the name should end in .vcf, not .gz.
-s Structural variants (SV) VCF. Gzipped or not.
-c Copy number variants (CNV) VCF. Gzipped or not.
-r Tandem repeats VCF. Gzipped or not.
-d DRAGEN only. The ROH BED produced by DRAGEN when run with --vc-enable-roh.
-modify ONT only. Marks repeat calls as repeats; add it when STRaglr is the repeat caller. Without it, ONT repeat records are left alone because they already carry SVTYPE=STR.
-b PacBio only. A BED restricting which repeat loci are unified. Without -b the repeats are left out of the unified file entirely, because the full TRGT catalog is far wider than a report can interpret.

A catalog of pathogenic hg38 STR loci ships with the package:

from UnifyVcf import default_repeat_catalog_path
pacbio-unify-vcf -o unified.vcf -s sv.vcf.gz \
  -r sample.GRCh38.trgt.sorted.vcf.gz \
  -b "$(python -c 'from UnifyVcf import default_repeat_catalog_path as p; print(p())')"

Python

from UnifyVcf import run

run(output_path="unified.vcf", sv_path="sv.vcf.gz", cnv_path="cnv.vcf.gz",
    repeat_path="repeats.vcf.gz", roh_bed_path="roh.bed")

run keeps the signature the standalone UnifyVcf.py script published, so code that did from UnifyVcf import run works unchanged once this package is installed.

A caller that already owns a sorting step — the Geneyx pipeline does — skips run and composes the pieces instead, so it keeps its own sorting and indexing:

from UnifyVcf import RohFormat, UnifyOptions, read_struct_lines, unify_lines

options = UnifyOptions(
    add_repeat_svtype=True,          # append ;SVTYPE=REP to repeat records
    roh_format=RohFormat.VCF,        # ROH arrives already converted
    drop_sv_filters=("InvBreakpoint",),  # exclude SV records with this FILTER
)
unify_lines(read_struct_lines(paths, options), "unified.vcf", options)   # unsorted

How the pieces fit

Function Does
read_struct_lines(paths, options) Reads each input (gzipped or not), drops excluded SV records, and skips inputs that are empty or header-only.
unify_lines(lines, output_path, options) Writes the unified, unsorted VCF. Returns None when there was nothing to write.
sort_and_compress(vcf_path) Sorts by position and bgzips; returns the .gz path.
filter_repeats_vcf(repeat_path, bed_path) Intersects a repeat VCF with a locus catalog, via bedtools or in Python.

The first present input in SV → CNV → repeats order donates the header; the rest contribute records only.

Platform support

Everything runs on Linux, macOS and Windows. Two things used to be Linux-only and are no longer:

Sorting and compression. The output has to be sorted by position and compressed as BGZF — a gzip variant written in <64KB blocks — because Geneyx indexes it with tabix, and tabix cannot index plain gzip. Both are done in pure Python (UnifyVcf.compression), so no sort or bgzip binary is needed. The ordering matches what sort -k1,1 -k2,2n produced under the previous shell implementation, including GNU sort's whole-line tie-break, so files keep the order Geneyx has always received.

Repeat filtering. bedtools intersect is used where bedtools is on PATH. Where it is not, the same intersection is computed in Python: a record spans [POS-1, END), taking END from INFO where present — repeat callers emit symbolic <STR> alleles whose length is not the length of REF — and falling back to POS-1 + len(REF). The engine that ran is logged. The CI runs the suite on both Linux and Windows, and the Linux leg installs bedtools so the two engines are compared against each other rather than one being skipped.

Force one or the other when you need to:

from UnifyVcf import BEDTOOLS, BUILTIN, filter_repeats_vcf

filter_repeats_vcf(repeats, catalog, engine=BUILTIN)

How this is distributed

The wheel ships compiled modules (.pyd on Windows, .so elsewhere), not .py files. What it publishes instead is the interface:

  • UnifyVcf/*.pyi type stubs carrying every public signature and its description, generated from the source by tools/generate_stubs.py
  • the same docstrings inside the compiled modules, so help(UnifyVcf.run) works
  • py.typed, so mypy and IDEs read the stubs
  • this README

So autocomplete, type checking and help() behave normally; the function bodies are just not there to read. __init__.py stays interpreted deliberately — it is only re-exports, and it is the surface this README documents.

Consequences worth knowing:

  • No sdist is published. An sdist would carry the sources, undoing the point. pip download --no-binary :all: therefore has nothing to fetch.
  • One wheel per platform and Python version, built by cibuildwheel in CI, rather than the single pure-Python wheel a normal package gets.
  • This is a deterrent, not a control. A compiled extension can still be analysed, and string literals — log messages, ;SVTYPE=REP and the like — remain visible in the binary because the code has to use them at run time. It stops casual reading and copy-paste reuse; it is not a security boundary.

Working on the package

The sources live in this repository and run interpreted, so development and the test suite are unaffected:

pip install -e .            # needs a C compiler; builds the extensions in place
python -m pytest tests
python tools/generate_stubs.py          # after changing any public signature
python tools/generate_stubs.py --check  # what CI enforces

tests/test_published_surface.py fails if a stub drifts from the source, so the published description cannot silently stop matching the code.

Per-caller differences

These are options rather than separate implementations:

DRAGEN ONT PacBio Geneyx pipeline
;SVTYPE=REP on repeats yes only with -modify yes yes
ROH input BED – – VCF
Repeat filtering – – bedtools against -b done as its own pipeline step
SV FILTER exclusions none none none InvBreakpoint (PacBio)

The PacBio row and the pipeline row differ on InvBreakpoint deliberately: the Geneyx pipeline drops those records from the unified file it analyses, and the customer-facing script does not. Aligning them would change what customers upload, so it is a product decision rather than a packaging one.

Known defect carried over

Pinned by a characterization test rather than fixed, so that packaging stayed a refactor. Worth its own change:

is_valid_line misses single-sample no-calls. The GT is read from the last tab-separated cell, which on a single-sample record still carries the trailing newline, so "./.\n" != "./." and the no-call is kept. Multi-sample records take the intended path. The neighbouring gt_index == '0' compares an int to a str and is never true.

Two other defects in the original shell-based sorter are gone as a consequence of rewriting it for Windows, rather than by choice:

  • It swallowed its own failures — a failed sort or bgzip was logged and execution continued, leaving an unsorted, uncompressed .vcf where the caller expected .vcf.gz. Failures now raise.
  • Its temp directory was vcf-temp in the current working directory, so two runs in one directory deleted each other's temp files. Temporary files are now written beside the output and removed in a finally.

Tests

pytest tests

No bedtools, bgzip or network needed — the suite is file manipulation only. One test compares the two repeat-filtering engines and skips where bedtools is not installed.

Metadata

Release files for UnifyVcf 0.1.0

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

Built distributions (wheels)

Table of built distributions (wheels) for UnifyVcf 0.1.0
File Actions
unifyvcf-0.1.0-cp313-cp313-win_amd64.whl CPython 3.13 CPython 3.13 Windows x86-64 Details for unifyvcf-0.1.0-cp313-cp313-win_amd64.whl
unifyvcf-0.1.0-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl CPython 3.13 CPython 3.13 Linux glibc 2.17+ x86-64, Linux glibc 2.28+ x86-64 Details for unifyvcf-0.1.0-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
unifyvcf-0.1.0-cp313-cp313-macosx_10_13_x86_64.whl CPython 3.13 CPython 3.13 macOS 10.13+ x86-64 Details for unifyvcf-0.1.0-cp313-cp313-macosx_10_13_x86_64.whl
unifyvcf-0.1.0-cp312-cp312-win_amd64.whl CPython 3.12 CPython 3.12 Windows x86-64 Details for unifyvcf-0.1.0-cp312-cp312-win_amd64.whl
unifyvcf-0.1.0-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl CPython 3.12 CPython 3.12 Linux glibc 2.17+ x86-64, Linux glibc 2.28+ x86-64 Details for unifyvcf-0.1.0-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
unifyvcf-0.1.0-cp312-cp312-macosx_10_13_x86_64.whl CPython 3.12 CPython 3.12 macOS 10.13+ x86-64 Details for unifyvcf-0.1.0-cp312-cp312-macosx_10_13_x86_64.whl
unifyvcf-0.1.0-cp311-cp311-win_amd64.whl CPython 3.11 CPython 3.11 Windows x86-64 Details for unifyvcf-0.1.0-cp311-cp311-win_amd64.whl
unifyvcf-0.1.0-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl CPython 3.11 CPython 3.11 Linux glibc 2.17+ x86-64, Linux glibc 2.28+ x86-64 Details for unifyvcf-0.1.0-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
unifyvcf-0.1.0-cp311-cp311-macosx_10_9_x86_64.whl CPython 3.11 CPython 3.11 macOS 10.9+ x86-64 Details for unifyvcf-0.1.0-cp311-cp311-macosx_10_9_x86_64.whl
unifyvcf-0.1.0-cp310-cp310-win_amd64.whl CPython 3.10 CPython 3.10 Windows x86-64 Details for unifyvcf-0.1.0-cp310-cp310-win_amd64.whl
unifyvcf-0.1.0-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl CPython 3.10 CPython 3.10 Linux glibc 2.17+ x86-64, Linux glibc 2.28+ x86-64 Details for unifyvcf-0.1.0-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
unifyvcf-0.1.0-cp310-cp310-macosx_10_9_x86_64.whl CPython 3.10 CPython 3.10 macOS 10.9+ x86-64 Details for unifyvcf-0.1.0-cp310-cp310-macosx_10_9_x86_64.whl
unifyvcf-0.1.0-cp39-cp39-win_amd64.whl CPython 3.9 CPython 3.9 Windows x86-64 Details for unifyvcf-0.1.0-cp39-cp39-win_amd64.whl
unifyvcf-0.1.0-cp39-cp39-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl CPython 3.9 CPython 3.9 Linux glibc 2.17+ x86-64, Linux glibc 2.28+ x86-64 Details for unifyvcf-0.1.0-cp39-cp39-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
unifyvcf-0.1.0-cp39-cp39-macosx_10_9_x86_64.whl CPython 3.9 CPython 3.9 macOS 10.9+ x86-64 Details for unifyvcf-0.1.0-cp39-cp39-macosx_10_9_x86_64.whl

Total release size: 9.0 MB

Release files / unifyvcf-0.1.0-cp313-cp313-win_amd64.whl

Download URL unifyvcf-0.1.0-cp313-cp313-win_amd64.whl
Size 189.4 kB
Tags CPython 3.13 Windows x86-64
SHA-256 checksum
How to use checksums
d333cb62d762f7d0faccfb83cd6ebe82f4d9176a8780dac704a5c432384183f1
BLAKE2b-256 checksum
How to use checksums
793d417ef3b0015a57a1218531ccc57c624670a31df9ba30679887940954951f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl

Download URL unifyvcf-0.1.0-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
Size 1.4 MB
Tags CPython 3.13 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum
How to use checksums
d5da3b53fd1b80883d3dcd10936cc68157b50005a3bc2efca18d66d6b2e45a0b
BLAKE2b-256 checksum
How to use checksums
97067aefd1b634ec0a0909fb4ad6faf33c099a9aea2efe2ccbc628190c152449
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp313-cp313-macosx_10_13_x86_64.whl

Download URL unifyvcf-0.1.0-cp313-cp313-macosx_10_13_x86_64.whl
Size 221.0 kB
Tags CPython 3.13 macOS 10.13+ x86-64
SHA-256 checksum
How to use checksums
3ecc437576f9e182d8b6855ace8a9bb131c27b3e94f2d89f75c43663d58d864e
BLAKE2b-256 checksum
How to use checksums
c0259b271fe80c4d78f3b6b34620dcebfa72537d81930b8ab5dd1a73ffbe2d47
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp312-cp312-win_amd64.whl

Download URL unifyvcf-0.1.0-cp312-cp312-win_amd64.whl
Size 191.1 kB
Tags CPython 3.12 Windows x86-64
SHA-256 checksum
How to use checksums
bc62a3342aac8313a18f2eeabdfcf0f002cce4d7788271856811ce878ccd86ff
BLAKE2b-256 checksum
How to use checksums
560bfe9d67f41bd62fb014c9da428a75294f845be7cf4f4613a31517907b963f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl

Download URL unifyvcf-0.1.0-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
Size 1.5 MB
Tags CPython 3.12 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum
How to use checksums
88a75967ae5a93fd20c7853971548a6ab263041fb10066263a8e2e471ca62071
BLAKE2b-256 checksum
How to use checksums
1fbbcf05bb6726857443cc3c14c9055cae2f36802d643bcc930e078d6b4f223a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp312-cp312-macosx_10_13_x86_64.whl

Download URL unifyvcf-0.1.0-cp312-cp312-macosx_10_13_x86_64.whl
Size 222.1 kB
Tags CPython 3.12 macOS 10.13+ x86-64
SHA-256 checksum
How to use checksums
49f14f412b42aad745448b5699acd363c15a858a8f5ab3353742099367260b36
BLAKE2b-256 checksum
How to use checksums
7098a4793e265c8602806918220aad466d0f60d5db1cba0c4ce112b15a80e72b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp311-cp311-win_amd64.whl

Download URL unifyvcf-0.1.0-cp311-cp311-win_amd64.whl
Size 196.3 kB
Tags CPython 3.11 Windows x86-64
SHA-256 checksum
How to use checksums
3e641a7b6528a00af1de78fd7ff1da8df36acc5216b08343a9d0c4c791028b6f
BLAKE2b-256 checksum
How to use checksums
e20600f8ba7cea90a8c3cadd21b95e856421fb35437e4c7d10c0959fc0865422
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl

Download URL unifyvcf-0.1.0-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
Size 1.4 MB
Tags CPython 3.11 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum
How to use checksums
616cf1dc7aac123ca2bcbcdfbf1f82c63b0c386de21653181558b054af794595
BLAKE2b-256 checksum
How to use checksums
a8b7b73544ca990717a8bac6274f8865941c72922d5a21b3a4dfb0eece9fd631
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp311-cp311-macosx_10_9_x86_64.whl

Download URL unifyvcf-0.1.0-cp311-cp311-macosx_10_9_x86_64.whl
Size 217.8 kB
Tags CPython 3.11 macOS 10.9+ x86-64
SHA-256 checksum
How to use checksums
55fb9ed42aa91409e14bd3c4c5013744a894653fecfbdb56841633d2a2289e3f
BLAKE2b-256 checksum
How to use checksums
f67e7f7914a9f134294c07f9cd4099349d7a7220c331a2875e41e0efb3c10162
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp310-cp310-win_amd64.whl

Download URL unifyvcf-0.1.0-cp310-cp310-win_amd64.whl
Size 197.8 kB
Tags CPython 3.10 Windows x86-64
SHA-256 checksum
How to use checksums
db9dc626dda34ac4d1cf3bb52b730fbcbbde2a0e8a18ca91bbd140ffcffc2893
BLAKE2b-256 checksum
How to use checksums
d557c94f636a4f5dc48d5fe087c935609e527fe7d662eafb58f2b37f779c1982
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl

Download URL unifyvcf-0.1.0-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
Size 1.4 MB
Tags CPython 3.10 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum
How to use checksums
3c6c68688f1e51627804efe61f3a4ffb45ebadb08b8936a023dcb59926344fe5
BLAKE2b-256 checksum
How to use checksums
c7cf0c5c838d8bca28bf2b089d9a05bf2798c6dd8e9d3ac9cb1d2a7d620b1c3b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp310-cp310-macosx_10_9_x86_64.whl

Download URL unifyvcf-0.1.0-cp310-cp310-macosx_10_9_x86_64.whl
Size 221.4 kB
Tags CPython 3.10 macOS 10.9+ x86-64
SHA-256 checksum
How to use checksums
d6f87297dd57a72c92bc35641003fa02ecf7305b041ecc26cd24928accacdc84
BLAKE2b-256 checksum
How to use checksums
4b6cba031b21936990fd822700b8ba45e5d4c73a05f3a5a123f9eb4a5cdf2d28
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp39-cp39-win_amd64.whl

Download URL unifyvcf-0.1.0-cp39-cp39-win_amd64.whl
Size 198.3 kB
Tags CPython 3.9 Windows x86-64
SHA-256 checksum
How to use checksums
691c4bc6d840d609d0c86ce973fc1fa101b6f0ce04478af6e475c022b5f119eb
BLAKE2b-256 checksum
How to use checksums
37b7f5a9cb3cb83ff25504f10da115e4bf15c88e140647f40d4045be2899b3fc
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp39-cp39-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl

Download URL unifyvcf-0.1.0-cp39-cp39-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
Size 1.3 MB
Tags CPython 3.9 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum
How to use checksums
fad60a31649ef911577d81ffc8905f5da1176b808ac4eb3f406f8b1369c446e6
BLAKE2b-256 checksum
How to use checksums
65f31bf5b335e1ad0e212c1de835861efbcb038b30dadb0a8c6ce5fe1137933c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Release files / unifyvcf-0.1.0-cp39-cp39-macosx_10_9_x86_64.whl

Download URL unifyvcf-0.1.0-cp39-cp39-macosx_10_9_x86_64.whl
Size 222.9 kB
Tags CPython 3.9 macOS 10.9+ x86-64
SHA-256 checksum
How to use checksums
29913cd5fb9f42d04ff440e747fe0fd091509b73221351c77c7506be2f216f91
BLAKE2b-256 checksum
How to use checksums
df3da80a0d928f9aaea8cec49161151187b2f38fdf28c17007fa53e2f9c0824e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.11.16

Security for UnifyVcf 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.

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