[dependabot]

Bumps the python group with 2 updates: gliner2 and ty.

Updates gliner2 from 2.0.0 to 2.1.0

Release notes

Sourced from gliner2's releases.

v2.1.0

GLiNER2 2.1.0 is on PyPI. Paste the note below into a Hugging Face model card or discussion.

# GLiNER2 2.1.0
Released 9 October 2026. Install with:
pip install -U "gliner2==2.1.0"
</code></pre>
<p>This is the library release after <a href="https://pypi.org/project/gliner2/2.0.0/">gliner2 2.0.0</a> (24 August 2026, the GLiNER2.5 release). It is built from <a href="https://github.com/fastino-ai/GLiNER2">fastino-ai/GLiNER2</a> <code>main</code> at <code>b86d35c</code>. Checkpoint weights are unchanged. What changed is inference, training, schemas, and packaging.</p>
<p>Full history: <a href="https://github.com/fastino-ai/GLiNER2/compare/v2.0.0...b86d35c">v2.0.0...b86d35c</a></p>
<h2>Highlights</h2>
<ul>
<li><strong>Transformers 5.</strong> <code>transformers&gt;=4.43,&lt;6</code> on every install profile. The <code>disentangled-flash</code> extra requires <code>transformers&gt;=4.57,&lt;6</code>.</li>
<li><strong>DisentangledFlash attention</strong> for DeBERTa encoders, including an automatic packed (unpadded) path and a differentiable training path.</li>
<li><strong>Faster span scoring and decoder synchronization.</strong></li>
<li><strong>Schemas from JSON now match the Python builder</strong>, including thresholds, <code>top_k</code>, label descriptions, prompts, and few-shot examples.</li>
<li><strong>Precomputed encoder states</strong> can be passed into every batch inference API, and a boundary model can be loaded without the encoder.</li>
<li>Training fixes for gold-label corruption, Indic tokenization, phantom entity spans, half-precision NaNs, record choice fields, and LoRA unload.</li>
</ul>
<h2>Faster inference</h2>
<h3>DisentangledFlash</h3>
<p>New load option <code>attention_backend</code>:</p>
<table>
<thead>
<tr>
<th>Value</th>
<th>Behavior</th>
</tr>
</thead>
<tbody>
<tr>
<td><code>&quot;standard&quot;</code></td>
<td>Transformers attention</td>
</tr>
<tr>
<td><code>&quot;flashdeberta&quot;</code></td>
<td>FlashDeBERTa. <code>use_flashdeberta</code> remains a compatibility alias</td>
</tr>
<tr>
<td><code>&quot;disentangled_flash&quot;</code></td>
<td><a href="https://github.com/delyan-boychev/disentangled-flash">DisentangledFlash</a></td>
</tr>
</tbody>
</table>
<pre lang="python"><code>from gliner2 import GLiNER2
model = GLiNER2.from_pretrained(
&quot;fastino/gliner2-base-v1&quot;,
attention_backend=&quot;disentangled_flash&quot;,
)
</code></pre>
<p>Install the extra with <code>pip install &quot;gliner2[disentangled-flash]&quot;</code>. On CUDA, FP16 or BF16 is what makes the flash kernels pay off.</p>
<p>Packed layout is on by default in auto mode. A batch uses the unpadded kernel when padding is at least 15% and the batch is large enough (<code>B * L2 &gt;= 221</code>, measured on an H200). Smaller batches stay on the padded path. Tune it with <code>disentangled_flash_packed</code>, <code>disentangled_flash_min_padding</code>, <code>disentangled_flash_min_work</code>, and <code>disentangled_flash_tuning</code>.</p>
<!-- raw HTML omitted -->
</blockquote>
<p>... (truncated)</p>
</details>
<details>
<summary>Commits</summary>
<ul>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/b86d35c02f2c1a4d9840ea50197bd1b6d13b031e&quot;&gt;&lt;code&gt;b86d35c&lt;/code&gt;&lt;/a> Merge pull request <a href="https://redirect.github.com/fastino-ai/GLiNER2/issues/189&quot;&gt;#189&lt;/a> from MattThomas-fastino/transformers-v5-support</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/ec0e31271d1a627e135e91ccc2932a97cbd68494&quot;&gt;&lt;code&gt;ec0e312&lt;/code&gt;&lt;/a> Support Transformers 5 and refresh dependency caps</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/08c8c8df0803075035ac6a02e8ee8c5554d2e06b&quot;&gt;&lt;code&gt;08c8c8d&lt;/code&gt;&lt;/a> Merge pull request <a href="https://redirect.github.com/fastino-ai/GLiNER2/issues/188&quot;&gt;#188&lt;/a> from delyan-boychev/optimize-decoder-sync</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/2763f94753ef910d77e71e4bcc3018207508c640&quot;&gt;&lt;code&gt;2763f94&lt;/code&gt;&lt;/a> Merge pull request <a href="https://redirect.github.com/fastino-ai/GLiNER2/issues/187&quot;&gt;#187&lt;/a> from kfastino/fix/cls-true-label-synthetic</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/467ea4fdbbb8ee3077b545333be2a23c2331d8a5&quot;&gt;&lt;code&gt;467ea4f&lt;/code&gt;&lt;/a> Add decoder synchronization benchmark matrix</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/f59fab04eb1d3c95fe223677e289820ea677c124&quot;&gt;&lt;code&gt;f59fab0&lt;/code&gt;&lt;/a> Merge remote-tracking branch 'upstream/main' into optimize-decoder-sync</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/0b67de781678825c8cd4fda5a9a25e5563ff3137&quot;&gt;&lt;code&gt;0b67de7&lt;/code&gt;&lt;/a> Merge pull request <a href="https://redirect.github.com/fastino-ai/GLiNER2/issues/143&quot;&gt;#143&lt;/a> from delyan-boychev/optimize-span-scoring</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/879fc2d4d3e9ef481dcc1368a0eebe462e0d802d&quot;&gt;&lt;code&gt;879fc2d&lt;/code&gt;&lt;/a> Optimize boundary decoder synchronization</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/cf348748194dd878831390d4bd674bfe651b1e9b&quot;&gt;&lt;code&gt;cf34874&lt;/code&gt;&lt;/a> Optimize accelerator decoder synchronization</li>
<li><a href="https://github.com/fastino-ai/GLiNER2/commit/0fa0e8870cfc30495b5c3870e0fe87dcef7af99a&quot;&gt;&lt;code&gt;0fa0e88&lt;/code&gt;&lt;/a> Merge remote-tracking branch 'origin/main' into optimize-span-scoring</li>
<li>Additional commits viewable in <a href="https://github.com/fastino-ai/GLiNER2/compare/v2.0.0...v2.1.0&quot;&gt;compare view</a></li>
</ul>
</details>
<br />
Updates ty from 0.0.69 to 0.0.99

Release notes
Sourced from ty's releases.

0.0.99
Release Notes
Released on 2026-11-06.
Bug fixes

Handle output errors when generating shell completions (#29997)
Invalidate aliased conditions across loop iterations (#30074)

LSP server

Add support for workspace/willRenameFiles LSP request (#28330)

Other changes

Bound comparisons of wide builtin literal unions (#30052)
Fix call argument validation for variadic parameters (#30086)
Freshen method typevars before owner specialization (#29543)
Inform call specializations compatibility through type context (#29932)
Keep source place IDs out of the use-def map (#30077)
Preserve type structure during typevar freshening (#29803)
Preserve typevars when mapping leaves their bounds unchanged (#29547)
Preserve unions inferred during fixpoint iteration (#29927)
Record lexical place IDs with their uses (#29956)
Skip existing release artifacts when retrying mirror uploads (#4937)
Skip fixpoint iteration for simple generic calls (#30054)
Solve generic call type context through constraint sets (#26680)

Contributors

@​ibraheemdev
@​lerebear
@​ewdurbin
@​carljm
@​charliermarsh
@​dcreager

Install ty 0.0.99
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://releases.astral.sh/github/ty/releases/download/0.0.99/ty-installer.sh | sh

Install prebuilt binaries via powershell script
powershell -ExecutionPolicy Bypass -c "irm https://releases.astral.sh/github/ty/releases/download/0.0.99/ty-installer.ps1 | iex"
</tr></table> 


... (truncated)


Changelog
Sourced from ty's changelog.

0.0.99
Released on 2026-11-06.
Bug fixes

Handle output errors when generating shell completions (#29997)
Invalidate aliased conditions across loop iterations (#30074)

LSP server

Add support for workspace/willRenameFiles LSP request (#28330)

Other changes

Bound comparisons of wide builtin literal unions (#30052)
Fix call argument validation for variadic parameters (#30086)
Freshen method typevars before owner specialization (#29543)
Inform call specializations compatibility through type context (#29932)
Keep source place IDs out of the use-def map (#30077)
Preserve type structure during typevar freshening (#29803)
Preserve typevars when mapping leaves their bounds unchanged (#29547)
Preserve unions inferred during fixpoint iteration (#29927)
Record lexical place IDs with their uses (#29956)
Skip existing release artifacts when retrying mirror uploads (#4937)
Skip fixpoint iteration for simple generic calls (#30054)
Solve generic call type context through constraint sets (#26680)

Contributors

@​ibraheemdev
@​lerebear
@​ewdurbin
@​carljm
@​charliermarsh
@​dcreager

0.0.98
Released on 2026-11-05.
Bug fixes

Avoid cubic narrowing for enum OR patterns (#30019)
Avoid duplicate warnings for unknown rules and override sections (#30028)
Fix class-pattern exhaustiveness after class identity checks (#30045)
Fix conditional type and tuple annotations (#29987)
Prefer active unnamed Conda and Pixi environments (#29974)
Preserve baseline matches when a rule is renamed (#30040)
Preserve exception types after tuple truthiness checks (#30018)



... (truncated)


Commits

9b48f40 Bump version to 0.0.99 (#4956)
aabbd38 Skip existing release artifacts when retrying mirror uploads (#4937)
e8a6c93 Bump version to 0.0.98 (#4936)
1b1d9c4 Update benchmarks for ty 0.0.96 (#4930)
58e168c Bump version to 0.0.97 (#4921)
60fff56 Document baseline support (#4897)
7250934 Update PyCharm setup instructions for 2026.2 (#4914)
133bc26 Update astral-sh/setup-uv action to v10.3.3 (#4916)
23fafdc Update dependency prek to v0.5.4 (#4917)
df18177 Update uv to v0.13.7 (#4918)
Additional commits viewable in compare view



Bumps the python group with 20 updates:



Package
From
To




certifi
2026.10.19
2026.10.23


pydantic
2.12.5
2.14.3


pydantic-settings
2.12.0
2.16.2


cyclopts
4.10.1
5.3.3


pygments
2.20.0
2.21.0


tiktoken
0.12.0
0.14.0


gliner2
2.0.0
2.1.0


ipykernel
7.3.0
7.4.0


jupytext
1.19.1
1.19.6


pip
26.2.2
26.3


fastapi
0.141.1
0.162.2


pre-commit
4.5.1
4.7.0


pytest
9.0.3
9.2.0


pytest-cov
7.0.0
7.1.0


ruff
0.15.0
0.17.4


ty
0.0.69
0.0.99


uvicorn
0.40.0
0.54.0


mike
2.1.4
2.2.0


mkdocs-material
9.7.5
9.8.1


griffe-pydantic
1.3.1
1.3.2



Updates certifi from 2026.10.19 to 2026.10.23

Commits

b60466c 2026.10.23 (#445)
See full diff in compare view



Updates pydantic from 2.12.5 to 2.14.3

Release notes
Sourced from pydantic's releases.

v2.14.3 2026-11-04
v2.14.3 (2026-11-04)
What's Changed
Packaging


Build wheels for GraalPy 3.13 by @​msimacek in #14136.
The previous v2.14.2 release published GraalPy 3.11 wheels, which were not working due to a PyO3 version bump. A CI workflow issue made it so that the wheel wasn't tested before publishing. This is now fixed: starting in v2.14.3, we no longer publish GraalPy 3.11 wheels, and GraalPy 3.13 wheels are now published together with the existing 3.12 one.


Performance

Improve schema generation for fixed type handlers by @​alexmojaki in #14130
Skip general schema dispatch for builtin scalars by @​alexmojaki in #14141
Skip annotation handler setup when annotations are empty by @​alexmojaki in #14149
Fix quadratic serialization time with per-index include/exclude on sequences by @​Viicos in #14160
Avoid type lookups when checking for None during serialization by @​Viicos in #14167
Improve performance of model serialization by @​Viicos in #14171
Improve performance of dict serialization by @​Viicos in #14172
Improve performance of type inference during serialization by @​Viicos in #14173
Avoid cloning the field name when using the default value during validation by @​Viicos in #14175
Do not evaluate model_config annotations by @​Viicos in #14195
Avoid slow runtime Protocol check when expanding GroupedMetadata annotations by @​Viicos in #14196
Use a lightweight @contextmanager implementation internally by @​Viicos in #14199

Fixes

Fix various issues about GC tracking in pydantic-core by @​Viicos in #13819
Include null in serialization JSON Schema for unless-none field serializers by @​Kropiunig in #13848
Fix RecursionError when copying or unpickling models with defer_build by @​Viicos in #14115
Remove dead default_factory check in unsupported Field() attribute by @​owtaylor in #14129
Disallow non string extra keys during validation by @​Viicos in #14135
Fix deeply nested data being reported as a cyclic reference by @​Viicos in #14164

New Contributors

@​owtaylor made their first contribution in #14129

v2.14.2 2026-10-29
v2.14.2 (2026-10-29)
This release addresses 2 security issues, https://github.com/pydantic/pydantic/security/advisories/GHSA-qpmf-fgh3-ff25 and https://github.com/pydantic/pydantic/security/advisories/GHSA-3r8g-cqwr-jr29.
Packaging

Bump PyO3 to 0.30.1, jiter to 0.18.0, use PyFrozenDict by @​Viicos in #14003
Bump speedate to v0.18.0 by @​Viicos in #14007
Publish pydantic-core wheels for PyPy 8.0 (pp80 ABI) by @​Viicos in #14083



... (truncated)


Changelog
Sourced from pydantic's changelog.




v2.14.0 (2026-10-08)
GitHub release
What's Changed
The highlights of the v2.14 release are available in the blog post.
Several minor changes (considered non-breaking changes according to our versioning policy)
are also included in this release. Make sure to look into them before upgrading.
This release drops support for Python 3.9 and adds support for Python 3.15.
New Features

Add support for multiple_of constraint in fraction and timedelta core schemas by @​Viicos in #13873
Add __namespace__ parameter to create_model() by @​Viicos in #13895

Changes

Add an ordered-dict core schema by @​Viicos in #13796
Add a counter core schema by @​Viicos in #13824
Do not mutate model_config attribute by @​Viicos in #13825
Reject newlines in NameEmail validation by @​Viicos in #13897

Fixes

Call enum constructors in both validation modes by @​Viicos in #13792
Fix documentation examples not raising and remove unused string_sub_type unused error code by @​Viicos in #13821
Catch OverflowError during ByteSize validation by @​Viicos in #13850
Require multiple_of constraints to be positive by @​Viicos in #13862
Always pop field name and model type stacks during schema generation by @​saquibjawedbit in #13859
Handle TypeError gracefully during docstring extraction by @​Viicos in #13874
Fix config propagation of stdlib dataclasses and TypedDicts in JSON Schema by @​Viicos in #13891
Don't apply ser_json_timedelta to all datetime types in serialization inference by @​Viicos in #13892
Encode JSON Schema defaults with a consistent configuration by @​Viicos in #13893
Propagate init and kw_only=False from Field() to stdlib dataclass fields by @​Viicos in #13898
Respect propagated extra config in the JSON Schema of stdlib dataclasses by @​Viicos in #13899
Handle OS errors during ZoneInfo validation by @​Viicos in #13912
Handle exceptions during Decimal validation by @​Viicos in #13940
Rebuild incomplete types during serialization inference by @​Viicos in #13922
Reject large exponents in Fraction validation by @​Viicos in #13944
Avoid validating URL prefix multiple times in multi host URLs by @​Viicos in #13945

New Contributors

@​saquibjawedbit made their first contribution in #13859



... (truncated)


Commits

ea492e0 Bump pydantic-core to v2.50.4
4f4a33f Prepare release v2.14.3
dca067b Benchmark TypedDict schema generation across builds
a2d1f6c Skip annotation handler setup when annotations are empty
3ae0a40 Construct builtin scalar union schemas directly
358d36f Skip general schema dispatch for builtin scalars
3cc701a Improve schema generation for fixed type handlers
762db7e Remove redundant None check in union schema generation
c214986 Bump pytest-codspeed to v5.0.4
038ee97 Bump typing-inspection to v0.4.5
Additional commits viewable in compare view



Updates pydantic-settings from 2.12.0 to 2.16.2

Release notes
Sourced from pydantic-settings's releases.

v2.16.2
Highlights
This patch release fixes env decoding of nested dictionary values, CliApp handling of AliasPath, nested secrets directory precedence and discriminated union defaults with nested_model_default_partial_update.
Behavior changes

An explicit secrets_dir on NestedSecretsSettingsSource now takes precedence (#1093). Previously it was ignored when the wrapped SecretsSettingsSource inherited a directory from model config, and an explicit empty list was masked too. The order is now: the wrapper's secrets_dir argument, then the wrapped source's directory, then model config. If you pass both Settings(_secrets_dir=...) and NestedSecretsSettingsSource(..., secrets_dir=...), the wrapper's argument now wins.

Bug fixes
Env and dotenv sources

Dictionary values typed as a union, such as dict[str, list[str] | str] or dict[str, list[str] | None], now JSON-decode values set through env_nested_delimiter (#1089, #1097). Only JSON arrays and objects are decoded, so plain values like 123, true or "quoted" are kept as strings. Decoding follows enable_decoding and the NoDecode/ForceDecode of the field that holds the dict, and type aliases and Annotated value types are resolved.
Nested dictionary values such as dict[str, list[dict[str, str]]] are decoded at the right level again (#1109), so APP_VALUES__X='[{"k":"v"}]' no longer reaches validation as a string.

CLI

CliApp.run now rebuilds AliasPath inputs for BaseModel and pydantic dataclass commands (#1096). A supplied path used to be reduced to the field name, which failed validation. AliasChoices priority is respected and unsupplied path defaults are left to the model.

Discriminated union defaults

With nested_model_default_partial_update=True, inner discriminated unions written as Annotated[..., Field(discriminator='...')] are now detected (#1076). When the incoming tag selects another member, the default's fields are no longer merged into it.
A stdlib dataclass default on a discriminated union field no longer raises NameError when another field of the dataclass has a string annotation naming a local class (#1077). Only the discriminator's own annotation is resolved now.


What's Changed

build(deps): bump the python-packages group with 3 updates by @​dependabot[bot] in pydantic/pydantic-settings#1056
Detect nested Field discriminators in default partial updates by @​digit50 in pydantic/pydantic-settings#1076
Resolve only the discriminator annotation of stdlib dataclass defaults by @​digit50 in pydantic/pydantic-settings#1077
Decode nested dictionary values with complex union types by @​chering14 in pydantic/pydantic-settings#1089
Fix regressions from nested dict union value decoding (#1089) by @​hramezani in pydantic/pydantic-settings#1097
Fix explicit nested secrets directory precedence by @​DivyamTalwar in pydantic/pydantic-settings#1093
Fix CliApp reconstruction of AliasPath inputs by @​DivyamTalwar in pydantic/pydantic-settings#1096
Fix JSON decoding of nested dictionary environment values by @​Hugohong258 in pydantic/pydantic-settings#1109
Prepare release 2.16.2 by @​hramezani in pydantic/pydantic-settings#1111

New Contributors

@​digit50 made their first contribution in pydantic/pydantic-settings#1076
@​chering14 made their first contribution in pydantic/pydantic-settings#1089
@​DivyamTalwar made their first contribution in pydantic/pydantic-settings#1093
@​Hugohong258 made their first contribution in pydantic/pydantic-settings#1109

Full Changelog: https://github.com/pydantic/pydantic-settings/compare/v2.16.1...v2.16.2

v2.16.1
Highlights


... (truncated)


Commits

94150a0 Prepare release 2.16.2 (#1111)
9e02de0 Fix JSON decoding of nested dictionary environment values (#1109)
7e2dacb Fix CliApp reconstruction of AliasPath inputs (#1096)
2d1f924 Fix explicit nested secrets directory precedence (#1093)
e9588ae Fix regressions from nested dict union value decoding (#1089) (#1097)
e71a53e Decode nested dictionary values with complex union types (#1089)
8dc9e05 Resolve only the discriminator annotation of stdlib dataclass defaults (#1077)
c50c853 Detect nested Field discriminators in default partial updates (#1076)
9aabb8a build(deps): bump the python-packages group with 3 updates (#1056)
4e286e2 Prepare release 2.16.1 (#1087)
Additional commits viewable in compare view



Updates cyclopts from 4.10.1 to 5.3.3

Release notes
Sourced from cyclopts's releases.

v5.3.3
Bug Fixes

convert NewType nested inside containers and unions to its underlying type by @​Enesuygurs in BrianPugh/cyclopts#1036
show choices for an enum NewType nested in a container or union by @​BrianPugh in BrianPugh/cyclopts#1038
complete default-command positionals alongside subcommands in bash/zsh/fish by @​BrianPugh in BrianPugh/cyclopts#1037

New Contributors

@​Enesuygurs made their first contribution in BrianPugh/cyclopts#1036

Full Changelog: https://github.com/BrianPugh/cyclopts/compare/v5.3.2...v5.3.3
v5.3.2
Bug Fixes

Fix enum members in a Literal (including Parameter(choices=Literal[...])) being shown as Color.RED in help, completion and errors instead of the name that parses (red) by @​youdie006 in BrianPugh/cyclopts#1033

New Contributors

@​youdie006 made their first contribution in BrianPugh/cyclopts#1033

Full Changelog: https://github.com/BrianPugh/cyclopts/compare/v5.3.1...v5.3.2
v5.3.1
Bug Fixes

respect explicit delimiters for path environment values by @​Don-Yin in BrianPugh/cyclopts#1021
validate Parameter(name="*") behind a type alias at setup time by @​Rahul-cs50 in BrianPugh/cyclopts#1030
resolve subscripted generic type aliases (#1028) by @​yash-bitla in BrianPugh/cyclopts#1031
bind config values to a meta app's own parameters (regressed in v4.0.0) by @​BrianPugh in BrianPugh/cyclopts#1027

New Contributors

@​Don-Yin made their first contribution in BrianPugh/cyclopts#1021
@​Rahul-cs50 made their first contribution in BrianPugh/cyclopts#1030
@​yash-bitla made their first contribution in BrianPugh/cyclopts#1031

Full Changelog: https://github.com/BrianPugh/cyclopts/compare/v5.3.0...v5.3.1
v5.3.0
Features

Number validator supports Decimal and Fraction. #988

Bug Fixes
Config & Environment Variables

Empty config containers now bind for dict, list, tuple, set and dataclass parameters (Any parameters keep the container as-is). #993, #1006
A config null is rejected for types that don't allow None. #1004
A config array is rejected for single-value types, including a pydantic Model | bool (previously bound True). #1005, #1014
A JSON config.Env value containing spaces stays one token. #997
Symlinked configuration filenames are preserved. @​ac1982 #995

JSON Parsing

dict[str, X] parameters accept a JSON object from the CLI or an env var. #1001
A JSON object for a nested pydantic model field no longer crashes with IndexError. #999



... (truncated)


Commits

e512a49 fix: complete default-command positionals alongside subcommands in bash/zsh/fish
7919e99 fix: list choices for an enum NewType nested in a container or union
11004d5 fix: convert NewType nested inside containers and unions to its underlying type
146828c Merge pull request #1033 from youdie006/literal-enum-member-choices
3f57429 fix: don't crash on a nameless Flag value in a Literal
243513d List enum members in a Literal by their transformed name
5c217f9 Merge pull request #1027 from BrianPugh/fix/meta-app-config-params
eea85d9 Merge pull request #1031 from yash-bitla/fix-generic-type-alias-1028
3357a6b fix: resolve generic type alias whose value ignores its type parameter
f98cbb2 fix: validate Parameter(name="*") behind a type alias at setup time
Additional commits viewable in compare view



Updates pygments from 2.20.0 to 2.21.0

Release notes
Sourced from pygments's releases.

2.21.0


New lexers:

BitBake (#3103)
Caddyfile (#3225)
CEL (#3048)
PureScript (#1077, #3054)



Updated lexers:

Bash: Fix coloured keyword at the beginning of a name (#2926)
Boogie: Add missing Boogie and Civl Verifier keywords (#3156)
C#:

Recognize interpolated verbatim strings with either $@ or @$
prefixes (#2685)
Support dollar-prefixed and multi-quote raw strings (#3129, #2897)
Recognize union (#3182)


C/C++:

Add C23/C++26 attributes (#3084)
Add more C2Y keywords (#3092)
Highlight a function following a namespace body (#2928)
Fix C/C++ lexer support for multiline pre-processor comments (#3051)
Add .ipp as a file extension (#3141, #1008)


Clojure: Recognize named, octal and unicode character literals such as
\space and \o377 as a single token (#979)
Csound: Add missing opcode parameter type letter (#3161)
CUDA: Derive from the C++ lexer instead of C to highlight C++
constructs such as template, class and namespace (#3127)
D: Allow non-ASCII (Unicode) identifiers (#1088)
Fish: Fix single quote backslash escape (#3138, #2821)
Go: Various lexer improvements (#3199)
GoogleSQL: Require a word break after SET (#3167)
Hexdump: Only match valid digits (#3200, #2847)
JavaScript: Highlight the arguments object (#3146)
Jsonnet: Recognize colons in array slice expressions (#2828)
JSX: Allow apostrophes in element text (#2816)
Julia: Fix rstrings backslash (#3140, #2537)
Kotlin: Support companion objects without an explicit name (#2525)
Kotlin: Don't let a nullable type marker (?) consume the following
character, so Foo?, and a?:b tokenize correctly (#2964)
Kusto: Recognize member-access dots in dynamic objects (#2779)
Lua: Various improvements (#3143)
Macaulay2: Update symbols to 1.26.05 (#3120)
Markdown:

Highlight bold-italics (***...*** and ___...___) (#3067)
Fix mention regex to support hyphens in usernames (#3139, #3135)


Markdown, reStructuredText, TiddlyWiki5: Fix wrong token offsets for
embedded code blocks (#3133)
Mathematica: Recognize \[Name] named-character escapes such as
\[Nu] instead of emitting an Error token (#3097)





... (truncated)


Changelog
Sourced from pygments's changelog.

Version 2.21.0
(released August 17th, 2026)


New lexers:

BitBake (#3103)
Caddyfile (#3225)
CEL (#3048)
PureScript (#1077, #3054)



Updated lexers:

Bash: Fix coloured keyword at the beginning of a name (#2926)
Boogie: Add missing Boogie and Civl Verifier keywords (#3156)
C#:

Recognize interpolated verbatim strings with either $@ or @$
prefixes (#2685)
Support dollar-prefixed and multi-quote raw strings (#3129, #2897)
Recognize union (#3182)


C/C++:

Add C23/C++26 attributes (#3084)
Add more C2Y keywords (#3092)
Highlight a function following a namespace body (#2928)
Fix C/C++ lexer support for multiline pre-processor comments (#3051)
Add .ipp as a file extension (#3141, #1008)


Clojure: Recognize named, octal and unicode character literals such as
\space and \o377 as a single token (#979)
Csound: Add missing opcode parameter type letter (#3161)
CUDA: Derive from the C++ lexer instead of C to highlight C++
constructs such as template, class and namespace (#3127)
D: Allow non-ASCII (Unicode) identifiers (#1088)
Fish: Fix single quote backslash escape (#3138, #2821)
Go: Various lexer improvements (#3199)
GoogleSQL: Require a word break after SET (#3167)
Hexdump: Only match valid digits (#3200, #2847)
JavaScript: Highlight the arguments object (#3146)
Jsonnet: Recognize colons in array slice expressions (#2828)
JSX: Allow apostrophes in element text (#2816)
Julia: Fix rstrings backslash (#3140, #2537)
Kotlin: Support companion objects without an explicit name (#2525)
Kotlin: Don't let a nullable type marker (?) consume the following
character, so Foo?, and a?:b tokenize correctly (#2964)
Kusto: Recognize member-access dots in dynamic objects (#2779)
Lua: Various improvements (#3143)
Macaulay2: Update symbols to 1.26.05 (#3120)
Markdown:

Highlight bold-italics (***...*** and ___...___) (#3067)
Fix mention regex to support hyphens in usernames (#3139, #3135)


Markdown, reStructuredText, TiddlyWiki5: Fix wrong token offsets for





... (truncated)


Commits

a43b45d Get ready for the 2.21.0 release.
d8f14cb Fix version_added for Purescript.
19c5817 Remove superfluous parentheses from PostgresExplainLexer (#3232)
9992e09Description has been truncated