Fixes#36358
---
**Summary**
This PR fixes a process termination safety bug in ShellToolMiddleware
where cleanup could kill the caller process group when
create_process_group is False.
**Why this change**
When the shell child is started without a dedicated process group, it
can share the parent group. The existing cleanup path used group kill
unconditionally, which could terminate unrelated processes including the
caller. This is a high-impact availability risk.
**What changed**
Updated shell session kill logic to use group kill only when the child
is in a different process group than the caller.
Added safe fallback to child-only kill when both share the same process
group.
Added regression tests for both scenarios:
Shared process group: no group kill, child-only kill.
Dedicated process group: existing group kill behavior is preserved.
**Validation**
Targeted new tests passed.
Full shell tool unit test file passed.
AI assistance disclosure
This contribution was prepared with assistance from an AI coding agent.
I reviewed, validated, and finalized the proposed changes and test
coverage before submission.
**Areas for careful review**
Process-group detection behavior on Linux and other POSIX environments.
Any implications for existing timeout and shutdown flows in shell
middleware.
Whether additional integration-level coverage is desirable for process
cleanup behavior.
---------
Co-authored-by: Mason Daugherty <github@mdrxy.com>
## Summary
Fixes#33709.
When `AnthropicPromptCachingMiddleware` runs before
`ModelFallbackMiddleware`, Anthropic-specific `cache_control` markers
can leak into fallback attempts targeting non-Anthropic models. This
change makes fallback retries sanitize those markers only on non-primary
attempts, preserving primary-call behavior and avoiding API changes.
Related: langchain-ai/deepagentsjs#551.
## Changes
### libs/langchain_v1
- Added a private fallback sanitizer in `model_fallback.py` to remove
Anthropic `cache_control` markers from fallback requests.
- Sanitization covers `model_settings`, system/content message blocks,
and tool payloads (`BaseTool.extras` plus dict-style tools/extras).
- Applied sanitization in both sync and async fallback retry paths
(`wrap_model_call` and `awrap_model_call`) while leaving the primary
attempt unchanged.
### libs/langchain_v1 tests
- Added sync and async regression tests in `test_model_fallback.py` that
simulate primary failure and assert fallback calls only succeed when
cache markers are stripped.
- Verified non-cache settings (for example `temperature` and `top_p`)
are preserved on fallback.
- Preserved existing fallback behavior coverage (ordering, exhaustion,
and error propagation).
## Compatibility with `BedrockPromptCachingMiddleware`
The sanitizer also covers `BedrockPromptCachingMiddleware`, which
injects `cache_control` into `model_settings` only. Bedrock-specific
markers (`cachePoint` blocks, content-block `cache_control`) are applied
by the chat model classes at API-call time and never appear in the
`ModelRequest`, so no additional handling is needed.
---------
Co-authored-by: Mason Daugherty <mason@langchain.dev>
Co-authored-by: Mason Daugherty <github@mdrxy.com>
When using `ProviderStrategy`, `create_agent` unnecessarily sets
`strict=True` on tools for all providers. This is only needed for OpenAI
/ chat completions. Here we unset `strict`. For OpenAI:
1. We set it in `BaseChatOpenAI.bind_tools` (as a convenience to users
calling `model.bind_tools` directly)
2. We (redundantly) special-case OpenAI in the `create_agent` factory
logic so that things will not break for users who upgrade `langchain`
but not `langchain-openai`.
Note: payloads for OpenAI are tested here and appear unchanged:
https://github.com/langchain-ai/langchain/blob/master/libs/langchain_v1/tests/unit_tests/agents/test_response_format_integration.py
Quick test:
```python
from langchain.agents import create_agent
from langchain.agents.structured_output import ProviderStrategy
from pydantic import BaseModel
class Weather(BaseModel):
temperature: float
condition: str
def weather_tool(location: str) -> str:
"""Get the weather at a location."""
return "Sunny and 75 degrees F."
for model in [
"anthropic:claude-sonnet-4-6",
"openai:gpt-5.4",
"google_genai:gemini-3.5-flash",
]:
agent = create_agent(
model=model,
tools=[weather_tool],
response_format=ProviderStrategy(Weather),
)
result = agent.invoke({
"messages": [{"role": "user", "content": "What's the weather in SF?"}]
})
print(result["structured_response"])
```
Adds a contract note for the default summarization prompt so future
edits preserve the pieces downstream consumers depend on. The key risk
is not the prompt text itself, but the `<messages>` marker and
`{messages}` placeholder that other middleware uses to splice in
additional instructions.
Bumps [cryptography](https://github.com/pyca/cryptography) from 46.0.7
to 48.0.1.
<details>
<summary>Changelog</summary>
<p><em>Sourced from <a
href="https://github.com/pyca/cryptography/blob/main/CHANGELOG.rst">cryptography's
changelog</a>.</em></p>
<blockquote>
<p>48.0.1 - 2026-06-09</p>
<pre><code>
* Updated Windows, macOS, and Linux wheels to be compiled with OpenSSL
4.0.1.
<p>.. _v48-0-0:</p>
<p>48.0.0 - 2026-05-04<br />
</code></pre></p>
<ul>
<li>
<p><strong>BACKWARDS INCOMPATIBLE:</strong> Support for Python 3.8 has
been removed.
<code>cryptography</code> now requires Python 3.9 or later.</p>
</li>
<li>
<p><strong>BACKWARDS INCOMPATIBLE:</strong> Loading an X.509 CRL whose
inner
<code>TBSCertList.signature</code> algorithm does not match the outer
<code>signatureAlgorithm</code> now raises <code>ValueError</code>.
Previously, such CRLs
were parsed successfully and only rejected during signature
validation.</p>
</li>
<li>
<p>Added support for
:doc:<code>/hazmat/primitives/asymmetric/mlkem</code> and
:doc:<code>/hazmat/primitives/asymmetric/mldsa</code> when using OpenSSL
3.5.0 or
later, in addition to the existing AWS-LC and BoringSSL support. This
means
post-quantum algorithms are now available to users of our wheels.</p>
<ul>
<li><strong>Note:</strong> Going forward, we do not guarantee that all
functionality
in <code>cryptography</code> will be available when building against
OpenSSL. See :doc:<code>/statements/state-of-openssl</code> for more
information.</li>
</ul>
</li>
</ul>
<p>.. _v47-0-0:</p>
<p>47.0.0 - 2026-04-24</p>
<pre><code>
* Support for Python 3.8 is deprecated and will be removed in the next
``cryptography`` release.
* **BACKWARDS INCOMPATIBLE:** Support for binary elliptic curves
(``SECT*`` classes) has been removed. These curves are rarely used and
have additional security considerations that make them undesirable.
* **BACKWARDS INCOMPATIBLE:** Support for OpenSSL 1.1.x has been
removed.
OpenSSL 3.0.0 or later is now required. LibreSSL, BoringSSL, and AWS-LC
continue to be supported.
* **BACKWARDS INCOMPATIBLE:** Dropped support for LibreSSL < 4.1.
* **BACKWARDS INCOMPATIBLE:** Loading keys with unsupported algorithms
or
keys with unsupported explicit curve encodings now raises
:class:`~cryptography.exceptions.UnsupportedAlgorithm` instead of
``ValueError``. This change affects
:func:`~cryptography.hazmat.primitives.serialization.load_pem_private_key`,
:func:`~cryptography.hazmat.primitives.serialization.load_der_private_key`,
:func:`~cryptography.hazmat.primitives.serialization.load_pem_public_key`,
:func:`~cryptography.hazmat.primitives.serialization.load_der_public_key`,
and :meth:`~cryptography.x509.Certificate.public_key` when called on
certificates with unsupported public key algorithms.
</tr></table>
</code></pre>
</blockquote>
<p>... (truncated)</p>
</details>
<details>
<summary>Commits</summary>
<ul>
<li><a
href="de987ce48c"><code>de987ce</code></a>
48.0.1 version bump and changelog (<a
href="https://redirect.github.com/pyca/cryptography/issues/14996">#14996</a>)</li>
<li><a
href="8e03e30e3a"><code>8e03e30</code></a>
bump for 48.0.0 release (<a
href="https://redirect.github.com/pyca/cryptography/issues/14796">#14796</a>)</li>
<li><a
href="295e0d254e"><code>295e0d2</code></a>
Add AGENTS.md with CLAUDE.md symlink (<a
href="https://redirect.github.com/pyca/cryptography/issues/14794">#14794</a>)</li>
<li><a
href="104a2de19e"><code>104a2de</code></a>
Bump BoringSSL, OpenSSL, AWS-LC in CI (<a
href="https://redirect.github.com/pyca/cryptography/issues/14793">#14793</a>)</li>
<li><a
href="67ec1e5198"><code>67ec1e5</code></a>
call check_length early on AesSiv::encrypt (<a
href="https://redirect.github.com/pyca/cryptography/issues/14792">#14792</a>)</li>
<li><a
href="b2da57a0d9"><code>b2da57a</code></a>
changelog for mldsa/mlkem for openssl (<a
href="https://redirect.github.com/pyca/cryptography/issues/14791">#14791</a>)</li>
<li><a
href="3cf44adee2"><code>3cf44ad</code></a>
ML-KEM OpenSSL support (<a
href="https://redirect.github.com/pyca/cryptography/issues/14781">#14781</a>)</li>
<li><a
href="2e31639666"><code>2e31639</code></a>
ML-DSA OpenSSL support (<a
href="https://redirect.github.com/pyca/cryptography/issues/14773">#14773</a>)</li>
<li><a
href="5affe5a286"><code>5affe5a</code></a>
fix rust nightly clippy (<a
href="https://redirect.github.com/pyca/cryptography/issues/14790">#14790</a>)</li>
<li><a
href="2e73ca448e"><code>2e73ca4</code></a>
bump rust-openssl dep and update EcPoint::mul_generator to
mul_generator2 (<a
href="https://redirect.github.com/pyca/cryptography/issues/1">#1</a>...</li>
<li>Additional commits viewable in <a
href="https://github.com/pyca/cryptography/compare/46.0.7...48.0.1">compare
view</a></li>
</ul>
</details>
<br />
[](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores)
Dependabot will resolve any conflicts with this PR as long as you don't
alter it yourself. You can also trigger a rebase manually by commenting
`@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits
that have been made to it
- `@dependabot show <dependency name> ignore conditions` will show all
of the ignore conditions of the specified dependency
- `@dependabot ignore this major version` will close this PR and stop
Dependabot creating any more for this major version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this minor version` will close this PR and stop
Dependabot creating any more for this minor version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this dependency` will close this PR and stop
Dependabot creating any more for this dependency (unless you reopen the
PR or upgrade to it yourself)
You can disable automated security fix PRs for this repo from the
[Security Alerts
page](https://github.com/langchain-ai/langchain/network/alerts).
</details>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
[](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores)
Dependabot will resolve any conflicts with this PR as long as you don't
alter it yourself. You can also trigger a rebase manually by commenting
`@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits
that have been made to it
- `@dependabot show <dependency name> ignore conditions` will show all
of the ignore conditions of the specified dependency
- `@dependabot ignore this major version` will close this PR and stop
Dependabot creating any more for this major version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this minor version` will close this PR and stop
Dependabot creating any more for this minor version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this dependency` will close this PR and stop
Dependabot creating any more for this dependency (unless you reopen the
PR or upgrade to it yourself)
You can disable automated security fix PRs for this repo from the
[Security Alerts
page](https://github.com/langchain-ai/langchain/network/alerts).
</details>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Part of https://github.com/langchain-ai/deepagents/issues/2873
---
`SummarizationMiddleware` now serializes the history passed to the
summarizer with XML formatting so URL-backed multimodal content remains
available in the prompt. The existing behavior avoided dumping raw
message metadata into the token budget, but the prefix serialization
path omitted image/audio/video URL blocks before the summary model saw
them.
## Changes
- Update `SummarizationMiddleware._create_summary` and
`SummarizationMiddleware._acreate_summary` to call
`get_buffer_string(..., format="xml")` for trimmed conversation history
- Preserve URL-backed multimodal blocks in the summary prompt while
still avoiding raw message metadata expansion
- Add sync and async unit coverage with a prompt-capturing chat model to
assert image URLs survive summarization input serialization
---------
Co-authored-by: Mason Daugherty <mason@langchain.dev>
Co-authored-by: Mason Daugherty <github@mdrxy.com>
Closes#38220
---
Users calling `create_agent(..., response_format=<schema>)` with an
OpenAI model pinned to a dated snapshot (e.g. `gpt-5.4-2026-03-05`) were
silently downgraded from native structured output (`ProviderStrategy`)
to tool-calling (`ToolStrategy`). This changes runtime behavior: extra
tool-call traces, different token usage, and no provider-side schema
enforcement.
The cause is in `_supports_provider_strategy`'s fallback patterns: the
`gpt-5.2` and `gpt-5.4` base patterns terminated with `($|[/:])`, which
— unlike their sibling families — rejected a trailing `-`, so OpenAI's
`-YYYY-MM-DD` dated-snapshot suffix matched none of the patterns. The
base patterns were deliberately strict to keep
`gpt-5.2-pro`/`gpt-5.4-pro` blocked, so rather than allowing any
trailing `-` (which would re-admit those `-pro` variants) this change
adds an optional dated-snapshot group `(-\d{4}-\d{2}-\d{2})?`. Dated
snapshots now resolve to `ProviderStrategy` while `-pro` variants stay
blocked.
Made by [Open
SWE](https://openswe.vercel.app/agents/c5ebcb29-8ce5-dda0-73f6-198e49f0c36c)
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
Bumps [pyjwt](https://github.com/jpadilla/pyjwt) from 2.12.0 to 2.13.0.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a
href="https://github.com/jpadilla/pyjwt/releases">pyjwt's
releases</a>.</em></p>
<blockquote>
<h2>2.13.0</h2>
<h1>PyJWT 2.13.0 — Security Release</h1>
<p>This release bundles five security fixes plus three additional
hardening / spec-compliance changes. We recommend all users upgrade.</p>
<h2>Security</h2>
<ul>
<li>
<p><strong><a
href="https://github.com/jpadilla/pyjwt/security/advisories/GHSA-xgmm-8j9v-c9wx"><code>GHSA-xgmm-8j9v-c9wx</code></a>
— JWK JSON accepted as HMAC secret (algorithm confusion).</strong>
<code>HMACAlgorithm.prepare_key</code> previously rejected PEM- and
SSH-formatted asymmetric keys but did not catch a JWK passed as a raw
JSON string. In a verifier configured with both symmetric and asymmetric
algorithms in <code>algorithms=[…]</code> and a raw-JSON JWK as the key,
an attacker could forge HS256 tokens using the JWK text as the HMAC
secret. The guard has been extended to reject any JWK-shaped JSON.
<em>Reported by <a
href="https://github.com/aradona91"><code>@aradona91</code></a>.</em></p>
</li>
<li>
<p><strong><a
href="https://github.com/jpadilla/pyjwt/security/advisories/GHSA-jq35-7prp-9v3f"><code>GHSA-jq35-7prp-9v3f</code></a>
— Algorithm allow-list bypass with <code>PyJWK</code> /
<code>PyJWKClient</code>.</strong> When verifying with a
<code>PyJWK</code>, the caller's <code>algorithms=[…]</code> allow-list
was checked against the token header <code>alg</code> as a string only;
actual verification used the algorithm bound to the <code>PyJWK</code>.
An attacker who controlled a registered JWKS key could sign with one
algorithm and advertise another on the header. PyJWT now requires the
token header <code>alg</code> to match the <code>PyJWK</code>'s
algorithm before verification. <em>Reported by <a
href="https://github.com/sushi-gif"><code>@sushi-gif</code></a>.</em></p>
</li>
<li>
<p><strong><a
href="https://github.com/jpadilla/pyjwt/security/advisories/GHSA-w7vc-732c-9m39"><code>GHSA-w7vc-732c-9m39</code></a>
— DoS via base64 decode of unused payload segment when
<code>b64=false</code>.</strong> For detached-payload JWS
(<code>b64=false</code>), the compact-form payload segment was
base64-decoded before being discarded in favor of the caller-supplied
<code>detached_payload</code>. An attacker could inflate the unused
segment to force CPU + memory cost without holding a valid signature.
The segment is now required to be empty per RFC 7515 Appendix F, and is
no longer decoded. <em>Reported by <a
href="https://github.com/thesmartshadow"><code>@thesmartshadow</code></a>.</em></p>
</li>
<li>
<p><strong><a
href="https://github.com/jpadilla/pyjwt/security/advisories/GHSA-993g-76c3-p5m4"><code>GHSA-993g-76c3-p5m4</code></a>
— <code>PyJWKClient</code> accepts non-HTTP(S) URIs.</strong>
<code>PyJWKClient.fetch_data</code> passed its URI to
<code>urllib.request.urlopen</code>, which by default also handles
<code>file://</code>, <code>ftp://</code>, and <code>data:</code>
schemes. An application that fed an attacker-influenced URI into
<code>PyJWKClient</code> could be coerced into reading local files or
reaching other unintended schemes. <code>PyJWKClient</code> now rejects
any URI whose scheme isn't <code>http</code> or <code>https</code>.
<em>Reported by <a
href="https://github.com/KEIJOT"><code>@KEIJOT</code></a>.</em></p>
</li>
<li>
<p><strong><a
href="https://github.com/jpadilla/pyjwt/security/advisories/GHSA-fhv5-28vv-h8m8"><code>GHSA-fhv5-28vv-h8m8</code></a>
— <code>PyJWKClient</code> cache wiped on fetch error.</strong> A
<code>finally</code>-block <code>put(jwk_set=None)</code> cleared the
JWK Set cache whenever a fetch raised, turning a transient JWKS-endpoint
outage into application-wide auth failure. The cache write was moved
into the success path; transient errors no longer evict valid cached
keys. <em>Reported by <a
href="https://github.com/eddieran"><code>@eddieran</code></a>.</em></p>
</li>
</ul>
<h2>Fixed</h2>
<ul>
<li>Reject empty HMAC keys outright in
<code>HMACAlgorithm.prepare_key</code> with <code>InvalidKeyError</code>
instead of accepting them with only a warning. Defends against the
<code>os.getenv("JWT_SECRET", "")</code> footgun.
<em>Thanks to <a
href="https://github.com/SnailSploit"><code>@SnailSploit</code></a> and
<a href="https://github.com/spartan8806"><code>@spartan8806</code></a>
for the reports.</em></li>
<li>Forward per-call <code>options</code> (including
<code>enforce_minimum_key_length</code>) from <code>PyJWT.decode</code>
through to <code>PyJWS._verify_signature</code>. The option was
previously silently dropped between the two layers, so it only took
effect when set on the <code>PyJWT</code> instance. <em>Thanks to <a
href="https://github.com/WLUB"><code>@WLUB</code></a> for the
report.</em></li>
<li><strong>RFC 7797 §3 compliance for <code>b64=false</code>:</strong>
the encoder now auto-adds <code>"b64"</code> to
<code>crit</code>, and the decoder rejects tokens that set
<code>b64=false</code> without listing it in <code>crit</code>.
<em>Thanks to <a
href="https://github.com/MachineLearning-Nerd"><code>@MachineLearning-Nerd</code></a>
for the report.</em></li>
</ul>
<h2>Changed</h2>
<ul>
<li>Migrate the <code>dev</code>, <code>docs</code>, and
<code>tests</code> package extras to dependency groups, by <a
href="https://github.com/kurtmckee"><code>@kurtmckee</code></a> in <a
href="https://redirect.github.com/jpadilla/pyjwt/pull/1152">#1152</a>.</li>
</ul>
<h2>Upgrade notes</h2>
<p>Most fixes are invisible to correctly-configured callers. A few
behavioral changes you may encounter:</p>
<ul>
<li><strong>Empty HMAC keys now raise.</strong> If your app passed
<code>""</code> or <code>b""</code> as a secret
(often via a missing env var, e.g.
<code>os.getenv("JWT_SECRET", "")</code>),
<code>encode</code>/<code>decode</code> will now raise
<code>InvalidKeyError</code>. This is the intended behavior — fix the
configuration.</li>
<li><strong><code>PyJWK</code> decoding now requires the token's
<code>alg</code> to match the JWK's algorithm.</strong> Previously a
mismatch was silently honored if the header <code>alg</code> appeared in
the allow-list. Tokens that relied on this mismatch will now fail with
<code>InvalidAlgorithmError</code>.</li>
<li><strong><code>PyJWKClient</code> now rejects non-HTTP(S) URIs at
construction time.</strong> Tests or dev environments that fetched JWKS
from <code>file://</code> URIs need to switch to a local HTTP server or
load the JWKS by other means (e.g. construct
<code>PyJWKSet.from_dict(...)</code> directly).</li>
<li><strong><code>b64=false</code> tokens are now strictly RFC 7515 /
7797 compliant.</strong> Tokens with a non-empty compact-form payload
segment, or that omit <code>"b64"</code> from
<code>crit</code>, will be rejected. PyJWT-produced tokens always
satisfy both invariants, so round-trips through PyJWT are
unaffected.</li>
<li><strong><code>enforce_minimum_key_length</code> set per-call now
takes effect.</strong> Callers who passed
<code>options={"enforce_minimum_key_length": True}</code> to
<code>jwt.decode()</code> previously got no enforcement; they will now
get <code>InvalidKeyError</code> on undersized keys, as documented.</li>
</ul>
<p><strong>Full changelog:</strong> <a
href="https://github.com/jpadilla/pyjwt/compare/2.12.1...2.13.0">https://github.com/jpadilla/pyjwt/compare/2.12.1...2.13.0</a></p>
<h2>2.12.1</h2>
<h2>What's Changed</h2>
<ul>
<li>Add typing_extensions dependency for Python < 3.11 by <a
href="https://github.com/jpadilla"><code>@jpadilla</code></a> in <a
href="https://redirect.github.com/jpadilla/pyjwt/pull/1151">jpadilla/pyjwt#1151</a></li>
</ul>
<p><strong>Full Changelog</strong>: <a
href="https://github.com/jpadilla/pyjwt/compare/2.12.0...2.12.1">https://github.com/jpadilla/pyjwt/compare/2.12.0...2.12.1</a></p>
</blockquote>
</details>
<details>
<summary>Changelog</summary>
<p><em>Sourced from <a
href="https://github.com/jpadilla/pyjwt/blob/master/CHANGELOG.rst">pyjwt's
changelog</a>.</em></p>
<blockquote>
<h2><code>v2.13.0
<https://github.com/jpadilla/pyjwt/compare/2.12.1...2.13.0></code>__</h2>
<p>Security</p>
<pre><code>
- Reject JWK JSON documents passed as raw HMAC secrets in
``HMACAlgorithm.prepare_key`` to close an algorithm-confusion gap that
the existing PEM/SSH guard did not cover. Reported by @aradona91 in
`GHSA-xgmm-8j9v-c9wx
<https://github.com/jpadilla/pyjwt/security/advisories/GHSA-xgmm-8j9v-c9wx>`__.
- Bind the JWT header ``alg`` to ``PyJWK.algorithm_name`` during
verification so the caller's ``algorithms=[...]`` allow-list cannot be
bypassed when decoding with a ``PyJWK`` / ``PyJWKClient`` key. Reported
by @sushi-gif in `GHSA-jq35-7prp-9v3f
<https://github.com/jpadilla/pyjwt/security/advisories/GHSA-jq35-7prp-9v3f>`__.
- Reject non-``http(s)`` URI schemes in ``PyJWKClient`` so attacker-
influenced URIs cannot read local files or reach unintended schemes via
urllib's default ``file://`` / ``ftp://`` / ``data:`` handlers. Reported
by @KEIJOT in `GHSA-993g-76c3-p5m4
<https://github.com/jpadilla/pyjwt/security/advisories/GHSA-993g-76c3-p5m4>`__.
- Preserve the cached JWK Set on fetch errors in
``PyJWKClient.fetch_data``.
The previous ``finally``-block ``put(None)`` pattern cleared the cache
on any transient outage, turning one bad JWKS request into application-
wide auth failure. Reported by @eddieran in `GHSA-fhv5-28vv-h8m8
<https://github.com/jpadilla/pyjwt/security/advisories/GHSA-fhv5-28vv-h8m8>`__.
- Skip the unconditional base64 decode of the compact-form payload
segment
when ``b64=false`` is set in the protected header, and require that
segment to be empty (RFC 7515 Appendix F detached form). Closes an
unauthenticated DoS amplifier. Reported by @thesmartshadow in
`GHSA-w7vc-732c-9m39
<https://github.com/jpadilla/pyjwt/security/advisories/GHSA-w7vc-732c-9m39>`__.
<p>Fixed</p>
<pre><code>
- Reject empty HMAC keys outright in ``HMACAlgorithm.prepare_key`` with
``InvalidKeyError`` instead of accepting them with only a warning.
Thanks to @SnailSploit and @spartan8806 for independently flagging the
footgun.
- Forward per-call ``options`` (including
``enforce_minimum_key_length``)
from ``PyJWT.decode`` through to ``PyJWS._verify_signature`` so the
option actually takes effect when set at the call site rather than only
on the ``PyJWT`` instance. Thanks to @WLUB for the report.
- RFC 7797 §3 compliance for ``b64=false``: the encoder now auto-adds
``&quot;b64&quot;`` to the ``crit`` header parameter, and the
decoder rejects
tokens that set ``b64=false`` without listing it in ``crit``. Thanks to
@MachineLearning-Nerd for the report.
Changed
</code></pre>
<ul>
<li>Migrate the <code>dev</code>, <code>docs</code>, and
<code>tests</code> package extras to dependency groups by <a
href="https://github.com/kurtmckee"><code>@kurtmckee</code></a> in
<code>[#1152](https://github.com/jpadilla/pyjwt/issues/1152)
&lt;https://github.com/jpadilla/pyjwt/pull/1152&gt;</code>__</li>
</ul>
<p><code>v2.12.1
&lt;https://github.com/jpadilla/pyjwt/compare/2.12.0...2.12.1&gt;</code>__
</tr></table>
</code></pre></p>
</blockquote>
<p>... (truncated)</p>
</details>
<details>
<summary>Commits</summary>
<ul>
<li><a
href="7144e4534c"><code>7144e45</code></a>
Apply ruff format</li>
<li><a
href="d2f4bec496"><code>d2f4bec</code></a>
Restore <code>cast()</code> calls with cross-version <code>type:
ignore</code> for <code>prepare_key</code></li>
<li><a
href="22f478cebd"><code>22f478c</code></a>
Remove redundant casts in <code>RSAAlgorithm.prepare_key</code> and
`ECAlgorithm.prepare...</li>
<li><a
href="95791b1759"><code>95791b1</code></a>
Bundle security fixes and hardening into 2.13.0</li>
<li><a
href="dcc27a9d31"><code>dcc27a9</code></a>
[pre-commit.ci] pre-commit autoupdate (<a
href="https://redirect.github.com/jpadilla/pyjwt/issues/1155">#1155</a>)</li>
<li><a
href="9d08a9a189"><code>9d08a9a</code></a>
[pre-commit.ci] pre-commit autoupdate (<a
href="https://redirect.github.com/jpadilla/pyjwt/issues/1146">#1146</a>)</li>
<li><a
href="b87c10014d"><code>b87c100</code></a>
Bump codecov/codecov-action from 5 to 6 (<a
href="https://redirect.github.com/jpadilla/pyjwt/issues/1154">#1154</a>)</li>
<li><a
href="40e3147eb5"><code>40e3147</code></a>
Migrate development extras to dependency groups (<a
href="https://redirect.github.com/jpadilla/pyjwt/issues/1152">#1152</a>)</li>
<li><a
href="a4e1a3d121"><code>a4e1a3d</code></a>
Add typing_extensions dependency for Python < 3.11 (<a
href="https://redirect.github.com/jpadilla/pyjwt/issues/1151">#1151</a>)</li>
<li>See full diff in <a
href="https://github.com/jpadilla/pyjwt/compare/2.12.0...2.13.0">compare
view</a></li>
</ul>
</details>
<br />
[](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores)
Dependabot will resolve any conflicts with this PR as long as you don't
alter it yourself. You can also trigger a rebase manually by commenting
`@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits
that have been made to it
- `@dependabot show <dependency name> ignore conditions` will show all
of the ignore conditions of the specified dependency
- `@dependabot ignore this major version` will close this PR and stop
Dependabot creating any more for this major version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this minor version` will close this PR and stop
Dependabot creating any more for this minor version (unless you reopen
the PR or upgrade to it yourself)
- `@dependabot ignore this dependency` will close this PR and stop
Dependabot creating any more for this dependency (unless you reopen the
PR or upgrade to it yourself)
You can disable automated security fix PRs for this repo from the
[Security Alerts
page](https://github.com/langchain-ai/langchain/network/alerts).
</details>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
README installation examples now use `uv add` consistently, matching the
repo's `uv`-based Python workflow. The top-level README also gets a
cleaner quickstart and resource section with current links for docs,
community, learning, and contribution guidance.
## Changes
- Replaced `pip install` snippets with `uv add` across package quick
install docs, including the Hugging Face extras and
`sentence-transformers` upgrade examples.
- Updated the top-level quickstart to show only `uv add langchain` and
refreshed the example model to `openai:gpt-5.5`.
- Pointed the LangGraph orchestration link at the LangGraph GitHub
repository.
- Consolidated top-level documentation and additional-resource links
under a single `Resources` section covering docs, ecosystem overview,
API reference, discussions, Academy, contributing, and the Code of
Conduct.
- Added LangChain Academy and Code of Conduct links to package README
resource sections.
Core serialization tests now opt into the object allowlists they rely on
instead of assuming default deserialization permits core objects.
Compatibility tests that intentionally exercise deprecated runnable
streaming and history APIs also suppress the expected deprecation
warnings so they can keep covering those legacy paths cleanly.
## Changes
- Updated serialization and prompt round-trip tests to pass
`allowed_objects="core"` or targeted allowlists when loading
`AIMessage`, prompt templates, structured prompts, runnable maps, and
related core objects.
- Adjusted secret-injection regression coverage to keep testing
`secrets_from_env=True` behavior while explicitly allowing core
deserialization paths.
- Tightened prompt deserialization rejection tests so attribute-access
payloads are loaded only through the specific prompt-template allowlist
needed to reach validation.
- Added module-level warning filters around legacy runnable
compatibility coverage for `astream_log`,
`astream_events(version="v1")`, and `RunnableWithMessageHistory`.
- Bumped the `langchain` package's minimum `langgraph` dependency from
`1.2.4` to `1.2.5`.
## Testing
- Updated unit tests across core serialization, prompt, fake chat model,
runnable history, and runnable event coverage.
Bumps `langchain-core` to `1.4.7` for the next patch release and updates
downstream minimum `langchain-core` requirements so package locks
resolve against the new core version.
This also refreshes the runnable snapshots that embed `lc_versions`
metadata so the version consistency check continues to validate
checked-in artifacts.
Validated with `python libs/core/scripts/check_version.py`, `uv lock
--check` across package lockfiles, and the core runnable tests that own
the updated snapshots with local LangSmith tracing env disabled.
Standardizes inline code markup in Python docstrings and comments by
replacing Sphinx-style double backticks with single-backtick Markdown.
The cleanup keeps existing code fences intact while aligning inline
references with the repo's docstring convention.
## Changes
- Converted inline code references in core prompt-loading docs and
LangSmith tracer comments, including `..`, `allow_dangerous_paths`, and
inheritable metadata keys.
- Normalized agent-related docstrings and comments around
`wrap_model_call`, `ExtendedModelResponse`, `Command`,
`create_structured_chat_agent`, and `DockerExecutionPolicy`.
- Updated partner package docstrings for inline references such as
`json_schema`, `ToolCall`, `apply_patch_call_output`, OpenRouter content
block keys, and Perplexity tool-call serialization.
- Cleaned test and helper docstrings that referenced command separators,
fake `resource` modules, stream event names, and xdist rate-limit
environment variables.
Originally a narrow bump of mypy to `1.20` in four packages. Expanded to
get the whole monorepo onto a single, current mypy and a consistent
type-check configuration, so contributors no longer hit different mypy
versions and divergent behavior depending on which package they touch.
### What changed
- **Unified the mypy pin to `>=2.1.0,<2.2.0`** in every mypy-using
package (6 libs + 14 partners), replacing the previously scattered pins
(`1.10`/`1.17`/`1.18`/`1.19`/`1.20`, with assorted upper bounds).
- **Unified the `[tool.mypy]` base per tier:**
- libs: `plugins = ["pydantic.mypy"]`, `strict = true`,
`enable_error_code = "deprecated"`, `warn_unreachable = true`
- partners: `disallow_untyped_defs = true`
- Normalized style (`disallow_untyped_defs = "True"` string → bool,
quote/key consistency).
- **Fixed the 20 real errors** mypy 2.1 surfaces: `redundant-cast` from
improved narrowing (`core`, `langchain-classic`), a `var-annotated` for
`_LOGGED`, a return-type widening in `langchain-groq`'s
`_convert_from_v1_to_groq` (it can legitimately return a bare `str`),
and stale `type-arg`/`unused-ignore` in `langchain-model-profiles`
tests.
### Deliberate non-uniformity (documented inline in the relevant
`pyproject.toml`s)
Going fully byte-identical would surface ~196 additional errors that are
*not* real bugs, so two settings are kept package-appropriate:
- **`warn_unreachable`** is enabled on every strict lib **except
`core`**, where it false-flags intentional defensive code — including
the SSRF / IP-policy guards in `_security/` — as unreachable.
- **`pydantic.mypy` plugin** is used only on `anthropic` and
`perplexity` (their code is authored against it and reports ~99/~132
errors without it). It is *not* added to the other partners, where it
only flags the public alias constructor API (e.g. `ChatGroq(model=...)`)
in tests rather than finding bugs.
- **`ollama`** is left on its `ty` type checker; it does not use mypy.
---------
Co-authored-by: Mason Daugherty <github@mdrxy.com>
Fixes#35244
Users can write async agent middleware with `@wrap_model_call`, and
LangChain already supports that behavior at runtime by detecting
coroutine functions and wiring them to `awrap_model_call`.
However, the decorator's public typing currently describes only the sync
callable shape. As a result, valid async middleware is rejected by type
checkers such as mypy and ty, even though the same code runs correctly.
This updates the middleware decorator types so async `wrap_model_call`
and `wrap_tool_call` functions type-check consistently with their
runtime behavior. It also simplifies related callable aliases and uses
casts where `iscoroutinefunction` narrows the callable at runtime but
static type checkers cannot follow that narrowing.
---------
Co-authored-by: Mason Daugherty <github@mdrxy.com>
Provider-native structured output fallback detection now uses bounded
model-name patterns instead of broad substring checks, reducing false
positives for unrelated model IDs. The model examples and test fixtures
across OpenAI/OpenRouter-facing code were refreshed around current
OpenAI model families while preserving shipped defaults.
## Changes
- Tightened `FALLBACK_MODELS_WITH_STRUCTURED_OUTPUT` from loose string
fragments to regex patterns, with `_supports_provider_strategy` matching
full model-name segments instead of arbitrary substrings.
- Expanded structured-output fallback coverage for newer OpenAI,
Anthropic, and xAI/Grok model families, including `gpt-5.x`, newer
Claude 4/5-style names, and `grok-build`.
- Reused `_attempt_infer_model_provider` in provider tool search routing
so `_provider_from_model_name` follows the same provider inference
behavior as `init_chat_model`.
- Suppressed irrelevant provider-inference deprecation warnings during
provider tool search registry lookup.
- Refreshed OpenAI, Azure OpenAI, OpenRouter, core metadata, and example
model references from older fixtures like `gpt-4`, `gpt-4o`, `o1`, and
`o4-mini` to current test/profile models such as `gpt-5.5`,
`gpt-5-nano`, and `gpt-4.1-mini`.
- Removed outdated OpenAI test assumptions around legacy `o1` behavior
and narrowed legacy structured-output checks to explicitly legacy model
names.
Simplify test for `create_agent` errors.
* Remove duplicate tests
* Test sync and async with common logic
---------
Co-authored-by: Mason Daugherty <github@mdrxy.com>
In this order:
* used `@override` when overriding a parent method.
* prefixed param with `_` when the param could be renamed.
* used `*_args, **_kwargs` when it was not possible to rename (eg:
protocol)
* used `_ = some_variable` when the variable name is inspected (in
tools)
---------
Co-authored-by: Mason Daugherty <github@mdrxy.com>
Co-authored-by: Mason Daugherty <mason@langchain.dev>
`SummarizationMiddleware._trigger_conditions` is now explicitly marked
as a temporary compatibility view for private consumers. The regression
test is tied to the package major version so the 2.0 release path fails
loudly until the legacy attr and test are removed.
`SummarizationMiddleware` now uses `_trigger_clauses` as the canonical
internal representation for AND/OR trigger evaluation while keeping
`_trigger_conditions` as a tuple-shaped compatibility view. This keeps
the new dict-style `TriggerClause` behavior intact without breaking
private consumers that still inspect the old tuple-normalized trigger
state.
## Changes
- Added `_trigger_clauses` as the source of truth for summarization
trigger evaluation, profile requirement checks, and compound AND clause
handling.
- Restored `_trigger_conditions` as a legacy compatibility projection
for tuple-expressible triggers, so tuple and single-key dict triggers
remain visible in the previous private shape.
- Avoided misrepresenting compound `TriggerClause` inputs like
`{"tokens": 1000, "messages": 5}` as independent OR-style tuple
conditions.
