fix(deps): update security fixes [security] (major) #112
No reviewers
Labels
No labels
ansible
bug
ci
dependencies
digest
docker
documentation
duplicate
enhancement
fix
go
good first issue
help wanted
invalid
javascript
major
minor
patch
php
python
question
released
security
wontfix
No milestone
No project
No assignees
2 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
dawid/rag-agent!112
Loading…
Reference in a new issue
No description provided.
Delete branch "renovate/major-security"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
This PR contains the following updates:
==0.3.30→==1.3.9==0.3.86→==1.2.22==0.3.11→==1.1.2==0.4.1→==1.0.10==8.4.2→==9.0.3LangChain: Path traversal and sandbox escape in LangChain file-search middleware and loaders
CVE-2026-55443 / GHSA-gr75-jv2w-4656 / PYSEC-2026-2192 / PYSEC-2026-2556
More information
Details
Summary
Several LangChain components that resolve filesystem paths or expand search patterns do not consistently confine the resolved path to the intended root directory. Affected behaviors include: a file-search agent middleware that validates a starting directory but not the search pattern or the resolved target of matched files, so glob patterns and symlinks can reach files outside the configured root; prompt- and chain/agent-configuration loaders that accept path fields and resolve them without confining the result to a trusted base or rejecting symlink targets; and path-prefix authorization checks that compare by string prefix without a path-segment boundary, so a sibling path sharing the prefix is accepted. When these components receive path values, search patterns, or workspace contents influenced by an untrusted source — including an LLM acting on untrusted input — the result can be disclosure of files outside the intended boundary. We have no evidence of this behavior being triggered in the wild.
Affected users / systems
You may be affected if you expose an agent with filesystem-search middleware over a directory and accept prompts or retrieved content influenced by untrusted sources; load prompt or chain/agent configuration from untrusted or shared sources; or rely on path-prefix restrictions to confine tool file access. Callers that confine these components to fully trusted inputs and first-party configuration are not affected.
Impact
Patches / mitigation
The affected components will canonicalize candidate paths (resolving symlinks) and verify the resolved real path remains within the configured root before reading or returning it; search patterns will be normalized so they cannot escape the root; configuration loaders will confine resolved path fields and reject symlink escapes unless the caller explicitly opts in to dangerous loading; and path-prefix checks will enforce a path-segment boundary. Path validation will be made operating-system-portable.
Compatibility
Callers that already pass only in-root paths, validated configuration, and trusted search inputs see no behavioral change. Callers that intentionally reference external paths can opt in via the existing dangerous-loading flag.
Operational guidance
Confine filesystem-backed agent tools to a dedicated directory and prefer running them sandboxed/containerized; validate path and identifier inputs where untrusted input enters; do not enable dangerous loading for configuration whose origin you do not control.
LangSmith / hosted deployments note
This issue concerns library components executed by agents.
Severity
CVSS:3.1/AV:L/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:NReferences
This data is provided by OSV and the GitHub Advisory Database (CC-BY 4.0).
CVE-2026-55443 / GHSA-gr75-jv2w-4656 / PYSEC-2026-2192 / PYSEC-2026-2556
More information
Details
LangChain is a framework for building agents and LLM-powered applications. Prior to 1.3.9, several LangChain components that resolve filesystem paths or expand search patterns do not consistently confine the resolved path to the intended root directory. Affected behaviors include: a file-search agent middleware that validates a starting directory but not the search pattern or the resolved target of matched files, so glob patterns and symlinks can reach files outside the configured root; prompt- and chain/agent-configuration loaders that accept path fields and resolve them without confining the result to a trusted base or rejecting symlink targets; and path-prefix authorization checks that compare by string prefix without a path-segment boundary, so a sibling path sharing the prefix is accepted. When these components receive path values, search patterns, or workspace contents influenced by an untrusted source — including an LLM acting on untrusted input — the result can be disclosure of files outside the intended boundary. This vulnerability is fixed in 1.3.9.
Severity
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:NReferences
This data is provided by OSV and the PyPI Advisory Database (CC-BY 4.0).
LangChain affected by SSRF via image_url token counting in ChatOpenAI.get_num_tokens_from_messages
CVE-2026-26013 / GHSA-2g6r-c272-w58r / PYSEC-2026-2562
More information
Details
Server-Side Request Forgery (SSRF) in ChatOpenAI Image Token Counting
Summary
The
ChatOpenAI.get_num_tokens_from_messages()method fetches arbitraryimage_urlvalues without validation when computing token counts for vision-enabled models. This allows attackers to trigger Server-Side Request Forgery (SSRF) attacks by providing malicious image URLs in user input.Severity
Low - The vulnerability allows SSRF attacks but has limited impact due to:
Impact
An attacker who can control image URLs passed to
get_num_tokens_from_messages()can:Note: This vulnerability occurs during token counting, which may happen outside of model invocation (e.g., in logging, metrics, or token budgeting flows).
Details
The vulnerable code path:
get_num_tokens_from_messages()processes messages containingimage_urlcontent blocksdetail: "low", it calls_url_to_size()to fetch the image and compute token counts_url_to_size()performshttpx.get(image_source)on any URL without validationFile:
libs/partners/openai/langchain_openai/chat_models/base.pyPatches
The vulnerability has been patched in
langchain-openai==1.1.9(requireslangchain-core==1.2.11).The patch adds:
langchain_core._security._ssrf_protection.validate_safe_url()to block:httpx.getdefault)allow_fetching_images=FalseparameterWorkarounds
If you cannot upgrade immediately:
image_urlvalues before passing messages to token counting or model invocationSeverity
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:LReferences
This data is provided by OSV and the GitHub Advisory Database (CC-BY 4.0).
LangChain affected by SSRF via image_url token counting in ChatOpenAI.get_num_tokens_from_messages
CVE-2026-26013 / GHSA-2g6r-c272-w58r / PYSEC-2026-2562
More information
Details
Server-Side Request Forgery (SSRF) in ChatOpenAI Image Token Counting
Summary
The
ChatOpenAI.get_num_tokens_from_messages()method fetches arbitraryimage_urlvalues without validation when computing token counts for vision-enabled models. This allows attackers to trigger Server-Side Request Forgery (SSRF) attacks by providing malicious image URLs in user input.Severity
Low - The vulnerability allows SSRF attacks but has limited impact due to:
Impact
An attacker who can control image URLs passed to
get_num_tokens_from_messages()can:Note: This vulnerability occurs during token counting, which may happen outside of model invocation (e.g., in logging, metrics, or token budgeting flows).
Details
The vulnerable code path:
get_num_tokens_from_messages()processes messages containingimage_urlcontent blocksdetail: "low", it calls_url_to_size()to fetch the image and compute token counts_url_to_size()performshttpx.get(image_source)on any URL without validationFile:
libs/partners/openai/langchain_openai/chat_models/base.pyPatches
The vulnerability has been patched in
langchain-openai==1.1.9(requireslangchain-core==1.2.11).The patch adds:
langchain_core._security._ssrf_protection.validate_safe_url()to block:httpx.getdefault)allow_fetching_images=FalseparameterWorkarounds
If you cannot upgrade immediately:
image_urlvalues before passing messages to token counting or model invocationSeverity
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:LReferences
This data is provided by OSV and the PyPI Advisory Database (CC-BY 4.0).
LangChain Core has Path Traversal vulnerabilites in legacy
load_promptfunctionsCVE-2026-34070 / GHSA-qh6h-p6c9-ff54 / PYSEC-2026-2193
More information
Details
Summary
Multiple functions in
langchain_core.prompts.loadingread files from paths embedded in deserialized config dicts without validating against directory traversal or absolute path injection. When an application passes user-influenced prompt configurations toload_prompt()orload_prompt_from_config(), an attacker can read arbitrary files on the host filesystem, constrained only by file-extension checks (.txtfor templates,.json/.yamlfor examples).Note: The affected functions (
load_prompt,load_prompt_from_config, and the.save()method on prompt classes) are undocumented legacy APIs. They are superseded by thedumpd/dumps/load/loadsserialization APIs inlangchain_core.load, which do not perform filesystem reads and use an allowlist-based security model. As part of this fix, the legacy APIs have been formally deprecated and will be removed in 2.0.0.Affected component
Package:
langchain-coreFile:
langchain_core/prompts/loading.pyAffected functions:
_load_template(),_load_examples(),_load_few_shot_prompt()Severity
High
The score reflects the file-extension constraints that limit which files can be read.
Vulnerable code paths
template_path,suffix_path,prefix_path_load_template().txtexamples(when string)_load_examples().json,.yaml,.ymlexample_prompt_path_load_few_shot_prompt().json,.yaml,.ymlNone of these code paths validated the supplied path against absolute path injection or
..traversal sequences before reading from disk.Impact
An attacker who controls or influences the prompt configuration dict can read files outside the intended directory:
.txtfiles: cloud-mounted secrets (/mnt/secrets/api_key.txt),requirements.txt, internal system prompts.json/.yamlfiles: cloud credentials (~/.docker/config.json,~/.azure/accessTokens.json), Kubernetes manifests, CI/CD configs, application settingsThis is exploitable in applications that accept prompt configs from untrusted sources, including low-code AI builders and API wrappers that expose
load_prompt_from_config().Proof of concept
Mitigation
Update
langchain-coreto >= 1.2.22.The fix adds path validation that rejects absolute paths and
..traversal sequences by default. Anallow_dangerous_paths=Truekeyword argument is available onload_prompt()andload_prompt_from_config()for trusted inputs.As described above, these legacy APIs have been formally deprecated. Users should migrate to
dumpd/dumps/load/loadsfromlangchain_core.load.Credit
Severity
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:NReferences
This data is provided by OSV and the GitHub Advisory Database (CC-BY 4.0).
CVE-2026-34070 / GHSA-qh6h-p6c9-ff54 / PYSEC-2026-2193
More information
Details
LangChain is a framework for building agents and LLM-powered applications. Prior to version 1.2.22, multiple functions in langchain_core.prompts.loading read files from paths embedded in deserialized config dicts without validating against directory traversal or absolute path injection. When an application passes user-influenced prompt configurations to load_prompt() or load_prompt_from_config(), an attacker can read arbitrary files on the host filesystem, constrained only by file-extension checks (.txt for templates, .json/.yaml for examples). This issue has been patched in version 1.2.22.
Severity
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:NReferences
This data is provided by OSV and the PyPI Advisory Database (CC-BY 4.0).
LangChain Text Splitters: HTMLHeaderTextSplitter.split_text_from_url SSRF Redirect Bypass
CVE-2026-41481 / GHSA-fv5p-p927-qmxr / PYSEC-2026-77
More information
Details
Summary
HTMLHeaderTextSplitter.split_text_from_url()validated the initial URL usingvalidate_safe_url()but then performed the fetch withrequests.get()with redirects enabled (the default). Because redirect targets were not revalidated, a URL pointing to an attacker-controlled server could redirect to internal, localhost, or cloud metadata endpoints, bypassing SSRF protections.The response body is parsed and returned as
Documentobjects to the calling application code. Whether this constitutes a data exfiltration path depends on the application: if it exposes Document contents (or derivatives) back to the requester who supplied the URL, sensitive data from internal endpoints could be leaked. Applications that store or process Documents internally without returning raw content to the requester are not directly exposed to data exfiltration through this issue.Affected versions
langchain-text-splitters< 1.1.2Patched versions
langchain-text-splitters>= 1.1.2 (requireslangchain-core>= 1.2.31)Affected code
File:
libs/text-splitters/langchain_text_splitters/html.py—split_text_from_url()The vulnerable pattern validated the URL once then fetched with redirects enabled:
Attack scenario
split_text_from_url(), relying on itsbuilt-in
validate_safe_url()check to block requests to internal networks.passes
validate_safe_url()(public hostname, public IP).302redirect to an internal endpoint(e.g., an unauthenticated internal admin API, or a cloud instance metadata
service that does not require request headers — such as AWS IMDSv1).
requests.get()follows the redirect automatically. The redirect target isnot revalidated.
Documentobjects to theapplication.
Notes:
split_text_from_url()includedvalidate_safe_url()specifically to besafe with untrusted URLs — the redirect loophole defeated that guarantee.
are not reachable through this bug because the attacker does not control
request headers. AWS IMDSv1, which requires no headers, is reachable.
party that supplied the URL. The SSRF itself — forcing the server to issue a
request to an internal endpoint — does not require this.
Fix
The fix replaces
requests.get()with an SSRF-safe httpx transport (SSRFSafeSyncTransportfromlangchain-core) that validates DNS results and pins connections to validated IPs on every request, including redirect targets, eliminating redirect-based bypasses.Additionally,
split_text_from_url()has been deprecated. Users should fetch HTML content themselves and pass it tosplit_text()directly.Severity
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:NReferences
This data is provided by OSV and the GitHub Advisory Database (CC-BY 4.0).
CVE-2026-41481 / GHSA-fv5p-p927-qmxr / PYSEC-2026-77
More information
Details
LangChain is a framework for building agents and LLM-powered applications. Prior to langchain-text-splitters
1.1.2, HTMLHeaderTextSplitter.split_text_from_url() validated the initial URL using validate_safe_url() but then performed the fetch with requests.get() with redirects enabled (the default). Because redirect targets were not revalidated, a URL pointing to an attacker-controlled server could redirect to internal, localhost, or cloud metadata endpoints, bypassing SSRF protections. The response body is parsed and returned as Document objects to the calling application code. Whether this constitutes a data exfiltration path depends on the application: if it exposes Document contents (or derivatives) back to the requester who supplied the URL, sensitive data from internal endpoints could be leaked. Applications that store or process Documents internally without returning raw content to the requester are not directly exposed to data exfiltration through this issue. This vulnerability is fixed in 1.1.2.
Severity
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:NReferences
This data is provided by OSV and the PyPI Advisory Database (CC-BY 4.0).
CVE-2026-28277 / GHSA-g48c-2wqr-h844 / PYSEC-2026-83
More information
Details
LangGraph SQLite Checkpoint is an implementation of LangGraph CheckpointSaver that uses SQLite DB (both sync and async, via aiosqlite). In version 1.0.9 and prior, LangGraph checkpointers can load msgpack-encoded checkpoints that reconstruct Python objects during deserialization. If an attacker can modify checkpoint data in the backing store (for example, after a database compromise or other privileged write access to the persistence layer), they can potentially supply a crafted payload that triggers unsafe object reconstruction when the checkpoint is loaded. No known patch is public.
Severity
CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:HReferences
This data is provided by OSV and the PyPI Advisory Database (CC-BY 4.0).
LangGraph checkpoint loading has unsafe msgpack deserialization
CVE-2026-28277 / GHSA-g48c-2wqr-h844 / PYSEC-2026-83
More information
Details
LangGraph checkpointers can load msgpack-encoded checkpoints that reconstruct Python objects during deserialization. If an attacker can modify checkpoint data in the backing store (for example, after a database compromise or other privileged write access to the persistence layer), they can potentially supply a crafted payload that triggers unsafe object reconstruction when the checkpoint is loaded.
This is a post-exploitation / defense-in-depth issue. Exploitation requires the ability to write attacker-controlled checkpoint bytes at rest. In most deployments that prerequisite already implies a serious incident; the additional risk is turning “checkpoint-store write access” into code execution in the application runtime, which can expand blast radius (for example by exposing environment variables or cloud credentials available to the runtime).
There is no evidence of exploitation in the wild, and LangGraph is not aware of a practical exploitation path in existing deployments today. This change is intended to reduce the blast radius of a checkpoint-store compromise.
Affected users / systems
Users may be affected if they:
This issue requires the attacker to be able to modify persisted checkpoint bytes (or to compromise a trusted component that writes them). It is generally not reachable by an unauthenticated remote attacker in a correctly configured deployment.
Impact
Exploitation scenario (high level)
Mitigation / remediation
LangGraph provides an allowlist-based hardening mechanism for msgpack checkpoint deserialization.
Strict mode (environment variable)
LANGGRAPH_STRICT_MSGPACK1,true,yes), the default msgpack deserialization policy becomes strict.JsonPlusSerializer()will defaultallowed_msgpack_modulestoNone(strict) instead ofTrue(warn-and-allow), unlessallowed_msgpack_modules=...is explicitly passed.allowed_msgpack_modules(serializer/checkpointer config)This setting controls what msgpack “ext” types are allowed to be reconstructed.
True(default when strict mode is not enabled): allow all ext types, but log a warning when deserializing a type that is not explicitly registered.None(strict): only a built-in safe set is reconstructed; other ext types are blocked.[(module, class_name), ...](strict allowlist): the built-in safe set plus exactly the listed symbols are reconstructed (exact-match).Built-in safe set
A small set of types is always treated as safe to reconstruct (for example
datetimetypes,uuid.UUID,decimal.Decimal,set/frozenset/deque,ipaddresstypes,pathlibpaths,zoneinfo.ZoneInfo, compiled regex patterns, and selected LangGraph internal types).Automatically derived allowlist (only when compiling graphs)
When
LANGGRAPH_STRICT_MSGPACKis enabled andStateGraphis compiled, LangGraph derives an allowlist from the graph’s schemas and channels and applies it to the checkpointer.Annotated).This derived allowlist is only applied if the selected checkpointer supports
with_allowlist(...). If a user is constructing serializers/checkpointers manually (or using a checkpointer that does not support allowlist propagation), they will need to configureallowed_msgpack_modulesthemselves.Operational guidance
LANGGRAPH_STRICT_MSGPACK=true) in production if feasible, and rely on schema-driven allowlisting to reduce incompatibilities.Limitations / important notes
with_allowlist), allowlist enforcement may be skipped (with a warning). In that situation, strict expectations may not hold.ext_hook), the custom hook controls reconstruction and can bypass the default allowlist checks (intentional escape hatch, but it weakens the protection).LangSmith / hosted deployments note
LangSmith is not aware of this issue presenting risk to existing LangSmith-hosted deployments. The described threat model requires an attacker to tamper with the checkpoint persistence layer used by the deployment; typical hosted configurations are designed to prevent such access.
First reported by: yardenporat353
Severity
CVSS:3.1/AV:A/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:HReferences
This data is provided by OSV and the GitHub Advisory Database (CC-BY 4.0).
pytest has vulnerable tmpdir handling
CVE-2025-71176 / GHSA-6w46-j5rx-g56g / PYSEC-2026-1845
More information
Details
pytest through 9.0.2 on UNIX relies on directories with the
/tmp/pytest-of-{user}name pattern, which allows local users to cause a denial of service or possibly gain privileges.Severity
CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:L/I:L/A:LReferences
This data is provided by OSV and the GitHub Advisory Database (CC-BY 4.0).
pytest has vulnerable tmpdir handling
CVE-2025-71176 / GHSA-6w46-j5rx-g56g / PYSEC-2026-1845
More information
Details
pytest through 9.0.2 on UNIX relies on directories with the
/tmp/pytest-of-{user}name pattern, which allows local users to cause a denial of service or possibly gain privileges.Severity
CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:L/I:L/A:LReferences
This data is provided by OSV and the PyPI Advisory Database (CC-BY 4.0).
Release Notes
langchain-ai/langgraph (langgraph)
v1.0.10: langgraph==1.0.10Compare Source
Changes since 1.0.10rc1
v1.0.9: langgraph==1.0.9Compare Source
Changes since 1.0.8
v1.0.8: langgraph==1.0.8Compare Source
Changes since 1.0.7
RuntimeandToolRuntimeclass descriptions for clarity (#6689)thread_id(#6515)add_nodeoverloads (#6514)v1.0.7: langgraph==1.0.7Compare Source
Changes since 1.0.6
uv lock --upgrade(#6671)v1.0.6: langgraph==1.0.6Compare Source
Changes since 1.0.5
v1.0.5: langgraph==1.0.5Compare Source
Changes since 1.0.4
v1.0.4Compare Source
Changes since 1.0.3
v1.0.3Compare Source
Changes since 1.0.2
StateGraphis a builder class (#6417)PartialStaterendering in MkDocs (#6416)invokeandainvokedocstrings (#6415)streamandastreamdocstrings (#6414)StateGraph(#6308)pyproject.tomllinks (#6364)v1.0.2Compare Source
Changes since 0.6.11
python.langchainlinks with newdocs.langchain(#6352)v1.0.1Compare Source
Changes since 1.0.0
v1.0.0Compare Source
Changes since 1.0.0rc1
v0.6.11Compare Source
Changes since 0.6.10
v0.6.10Compare Source
Changes since 0.6.9
v0.6.9Compare Source
Changes since 1.0.0a4
v0.6.8Compare Source
Changes since 1.0.0a3
BaseChanneldetection (#6210)CheckpointTask.statecan be aStateSnapshot(#6201)AsyncPregelLoop(#6167)defer=True(#6130)RemoteGraph(#6140)get_graphgenerates unexpected conditional edge (#6122)v0.6.7Compare Source
Changes since 1.0.0a2
v0.6.6Compare Source
Changes since 0.6.5
v0.6.5Compare Source
Changes since 0.6.4
v0.6.4Compare Source
Changes since 0.6.3
updated_channelsto checkpoint (#5828)MessageGraph(#5843)v0.6.3Compare Source
Changes since 0.6.2
AgentState(#5801)configparam (#5798)invokeandainvoke(#5771)v0.6.2Compare Source
Changes since 0.6.1
context_schematoconfig_schemawith correct condition (#5746)v0.6.1Compare Source
Changes since 0.6.0
v0.6.0Compare Source
LangGraph v0.6
We’re excited to announce the release of LangGraph v0.6.0, another significant step toward our v1.0 milestone. This release emphasizes providing a cleaner, more intuitive developer experience for building agentic workflows. Below we’ll cover the headline improvements and minor changes.
🚀 New Context API: Simplified Runtime Context Injection
The biggest improvement in v0.6 is the introduction of the new Context API, which makes it easier to pass run-scoped context in an intuitive and type safe way.
This pattern replaces the previously recommended pattern of injecting run-scoped context into
config['configurable'].Before (v0.5):
After (v0.6):
The
Runtimeclass provides a single interface for accessing information like:previousis also available: the previous return value for the given threadNow, instead of injecting all of the above as separate parameters to node functions,
developers can access them all through a single
runtimeparameter.Migration Path
config_schemais deprecated in favor ofcontext_schema, and will be removed in v2.0.0config_schemaget_config_jsonschemais deprecated in favor ofget_context_jsonschema(though this is generally only used for graph introspection and not by most langgraph users)🔀 Dynamic model & tool selection
create_react_agentcan now dynamically choose both the model and tools at runtime using a custom context object:Then invoke the agent with your desired settings:
Now agents can flexibly adapt their behavior based on runtime context.
🏗️ Durability Mode Support
LangGraph v0.6 introduces a new
durability****argument that gives you fine-grained control over persistence behavior. This provides finer grained control than its predecessor,checkpoint_during.This was predated
"exit"- Save checkpoint only when the graph exitscheckpoint_during=False"async"- Save checkpoint asynchronously while next step executescheckpoint_during=True"sync"- Save checkpoint synchronously before next stepMigration Path
checkpoint_duringis now deprecated in favor of the newdurabilityargument. Backwards compatibility will be maintained until v2.0.0.🛡️ Enhanced Type Safety and Validation
In an effort to make graph building easier for developers, we’ve enhanced the type safety of the
LangGraph APIs.
LangGraph’s
StateGraphandPregelinterfaces are now generic over a graph’s:state_schemacontext_schemainput_schemaoutput_schemaThis means that:
invoke/streamis type checked against the relevant schemacontextavailable via the aforementionedRuntimeclass matches thecontext_schema🔧 A Refined
InterruptInterfaceIn preparation for v1.0, we’ve made a few changes to the
Interruptinterface.Interrupts now have two attributes:
id- a unique identifier for the interruptvalue- the interrupt valueIn v0.6, we’ve removed the following attributes from the
Interruptclass:when- this was always"during"and offered no practical valueresumable- functionally, this is alwaysTruens- this information is now stored in a condensed format in theidattributeinterrupt_idhas been deprecated in favor ofid, but is still usable for backward compatibility🔒 Solidified Public API Surface
Gearing up for v1.0, we’ve solidified what’s public vs. private in the LangGraph API.
We’ve also deprecated some old import paths that have supported backports for ~1 year.
These changes make it easier to maintain a higher quality public API
and reduce the surface area for potential breaking changes.
The following table summarizes the changes:
from langgraph.pregel.types import ...from langgraph.types import ...from langgraph.constants import Sendfrom langgraph.types import Sendfrom langgraph.constants import Interruptfrom langgraph.types import Interruptfrom langgraph.channels import <ErrorClass>from langgraph.errors import <ErrorClass>langgraph.errorsfrom langgraph.constants import TAG_NOSTREAM_ALTfrom langgraph.constants import NOSTREAM🎯 Looking Toward v1.0
LangGraph v0.6 represents our final major changes before the stable v1.0 release.
We anticipate adhering strictly to SemVer post v1.0, leaning into a promise of stability and predictability.
Get Involved
LangGraph is an open source project, and we’d love to hear from you! We’ve rolled out a new LangChain forum for questions, feature requests, and discussions.
Please let us know what you think about the new Runtime API and other changes in v0.6, and if you have any difficulties with which we can help.
Full Changelog
contextAPI (#5566)Runtimeinterface re patch/overrides (#5546)constants.py->_internal/_typing.py(#5518)config['configurable']andconfig_schema) (#5243)nodesignatures vsinput_schemaforadd_node(#5424)Interruptinterface for v1 (#5405)v0.5.4Compare Source
Changes since 0.5.3
ENDwithCommand(#5601)stacklevel=2to the warnings to point to the caller’s codes (#5457)v0.5.3Compare Source
Changes since 0.5.2
uv lock --upgrade(#5471)ABCspec forPregelProtocol(#5485)v0.5.2Compare Source
Changes since 0.5.1
invoke/streamto allow forCommandandNone(#5414)v0.5.1Compare Source
Changes since 0.5.0
v0.5.0Compare Source
LangGraph 0.5 – the “Getting-Ready-for-1.0” release 🎉
TL;DR – 0.5 is not a radical rewrite, but a scrub-down and tune-up of the LangGraph core.
APIs are a little stricter, you have more control over streaming, checkpoints are lighter, etc. 99 % of users can upgrade with nothing more than a
pip install --upgrade langgraph==0.5.*.Why 0.5?
The team’s next big milestone is a 1.0 release in a few months.
To get there we needed to:
0.5 is that housekeeping release.
Headline changes
1. A leaner, stricter StateGraph
state_schemais now mandatory.“Untyped” graphs were never shown in the docs and produced surprising runtime errors. Requiring an explicit schema fixes that class of bugs and improves static analysis.
input/output→input_schema/output_schemaThe old names still work but raise a deprecation warning.
New
NodeBuilderutilityA simpler, declarative way to create nodes and attach them to channels. The old
Channel.subscribe_tohelper keeps working but will be removed in 1.0.2. Smarter streaming modes
stream_mode="debug"is now an alias for the pair["tasks", "checkpoints"]. You can now turn them on individually:This makes it cheaper to subscribe only to the information you need.
3. Checkpointing overhaul
Redundant keys have been dropped and per-task writes are stored directly.
Legacy “pending _sends” data is auto-migrated the first time it is loaded. Custom checkpointers continue to work unchanged.
4. Better serialization
JsonPlusSerializernow handles NumPy arrays stored in your state (including Fortran-ordered ones) without falling back topickle.Minor breaking changes you might notice
state_schemarequired – add it if you were passing only input and output schemas instead (very rare).input/outputrenaming – rename toinput_schema/output_schema.PregelNodeandRunnable, drop the latter.Nothing else should require code changes.
How to upgrade
If you maintain a plugin / custom checkpointer, run your test suite once; the public interfaces are untouched.
What’s next?
We’re hard at work on LangGraph 1.0, chime in here with any comments, feedback or questions, we want to hear from everyone.
Detailed Changelog
MessageGraph(#4875)"outputusage in favor ofoutput_schema(#5095)PregelNode's inheritance fromRunnable(#5093)inputandoutputin favor ofinput_schemaandoutput_schema(#4983)pep 604union syntax andpep 585generic syntax (#4963)StateGraph(dict)(#4964)retry->retry_policy(#4957)initandinvoke/stream(#4932)state_schemainStateGraph.__init__(#4897)MessageGraph(#4875)v0.4.10Compare Source
Changes since 0.4.9
MessageGraph(#4875)"outputusage in favor ofoutput_schema(#5095)PregelNode's inheritance fromRunnable(#5093)inputandoutputin favor ofinput_schemaandoutput_schema(#4983)pep 604union syntax andpep 585generic syntax (#4963)StateGraph(dict)(#4964)retry->retry_policy(#4957)initandinvoke/stream(#4932)state_schemainStateGraph.__init__(#4897)MessageGraph(#4875)v0.4.9Compare Source
Changes since 0.5.0rc1
v0.4.8Compare Source
Changes since 0.4.7
state_schemainStateGraph.__init__(#4897)MessageGraph(#4875)v0.4.7Compare Source
Changes since 0.4.6
v0.4.6Compare Source
Changes since 0.4.5
messages/message-tuplestream (#4722)messages/message-tuplestreamv0.4.5Compare Source
Changes since 0.4.4
v0.4.4Compare Source
Changes since 0.4.3
v0.4.3Compare Source
Changes since 0.4.2
v0.4.2Compare Source
Changes since 0.4.1
pytest-dev/pytest (pytest)
v9.0.3Compare Source
pytest 9.0.3 (2026-04-07)
Bug fixes
#12444: Fixed
pytest.approxwhich now correctly takes into account~collections.abc.Mappingkeys order to compare them.#13634: Blocking a
conftest.pyfile using the-p no:option is now explicitly disallowed.Previously this resulted in an internal assertion failure during plugin loading.
Pytest now raises a clear
UsageErrorexplaining that conftest files are not plugins and cannot be disabled via-p.#13734: Fixed crash when a test raises an exceptiongroup with
__tracebackhide__ = True.#14195: Fixed an issue where non-string messages passed to unittest.TestCase.subTest() were not printed.
#14343: Fixed use of insecure temporary directory (CVE-2025-71176).
Improved documentation
-pvsPYTEST_PLUGINSplugin loading and fixed an incorrect-pexample.capsysandcapfd) take precedence over the-s/--capture=nocommand-line options inAccessing captured output from a test function <accessing-captured-output>.pytest_collectionhook setssession.itemsbefore it callspytest_collection_finish, not after.Contributor-facing changes
#12689: The test reports are now published to Codecov from GitHub Actions.
The test statistics is visible on the web interface.
-- by
aleguy02v9.0.2Compare Source
pytest 9.0.2 (2025-12-06)
Bug fixes
#13896: The terminal progress feature added in pytest 9.0.0 has been disabled by default, except on Windows, due to compatibility issues with some terminal emulators.
You may enable it again by passing
-p terminalprogress. We may enable it by default again once compatibility improves in the future.Additionally, when the environment variable
TERMisdumb, the escape codes are no longer emitted, even if the plugin is enabled.#13904: Fixed the TOML type of the
tmp_path_retention_countsettings in the API reference from number to string.#13946: The private
config.inicfgattribute was changed in a breaking manner in pytest 9.0.0.Due to its usage in the ecosystem, it is now restored to working order using a compatibility shim.
It will be deprecated in pytest 9.1 and removed in pytest 10.
#13965: Fixed quadratic-time behavior when handling
unittestsubtests in Python 3.10.Improved documentation
pytest's command-line flags <command-line-flags>.v9.0.1Compare Source
pytest 9.0.1 (2025-11-12)
Bug fixes
raise unittest.SkipTest.Packaging updates and notes for downstreams
version string can be passed into its
package_envthroughthe
SETUPTOOLS_SCM_PRETEND_VERSION_FOR_PYTESTenvironmentvariable as a part of the release process -- by
webknjaz.Contributor-facing changes
creating GitHub Releases without having a Git checkout on
disk -- by
bluetechandwebknjaz.version string can be passed into its
package_envthroughthe
SETUPTOOLS_SCM_PRETEND_VERSION_FOR_PYTESTenvironmentvariable as a part of the release process -- by
webknjaz.v9.0.0Compare Source
pytest 9.0.0 (2025-11-05)
New features
#1367: Support for subtests has been added.
subtests <subtests>are an alternative to parametrization, useful in situations where the parametrization values are not all known at collection time.Example:
Each assert failure or error is caught by the context manager and reported individually, giving a clear picture of all files that are missing a docstring.
In addition,
unittest.TestCase.subTestis now also supported.This feature was originally implemented as a separate plugin in pytest-subtests, but since then has been merged into the core.
#13743: Added support for native TOML configuration files.
While pytest, since version 6, supports configuration in
pyproject.tomlfiles under[tool.pytest.ini_options],it does so in an "INI compatibility mode", where all configuration values are treated as strings or list of strings.
Now, pytest supports the native TOML data model.
In
pyproject.toml, the native TOML configuration is under the[tool.pytest]table.The
[tool.pytest.ini_options]table remains supported, but both tables cannot be used at the same time.If you prefer to use a separate configuration file, or don't use
pyproject.toml, you can usepytest.tomlor.pytest.toml:The documentation now (sometimes) shows configuration snippets in both TOML and INI formats, in a tabbed interface.
See
config file formatsfor full details.#13823: Added a "strict mode" enabled by the
strictconfiguration option.When set to
true, thestrictoption currently enablesstrict_configstrict_markersstrict_parametrization_idsstrict_xfailThe individual strictness options can be explicitly set to override the global
strictsetting.The previously-deprecated
--strictcommand-line flag now enables strict mode.If pytest adds new strictness options in the future, they will also be enabled in strict mode.
Therefore, you should only enable strict mode if you use a pinned/locked version of pytest,
or if you want to proactively adopt new strictness options as they are added.
See
strict modefor more details.#13737: Added the
strict_parametrization_idsconfiguration option.When set, pytest emits an error if it detects non-unique parameter set IDs,
rather than automatically making the IDs unique by adding 0, 1, ... to them.
This can be particularly useful for catching unintended duplicates.
#13072: Added support for displaying test session progress in the terminal tab using the OSC 9;4; ANSI sequence.
When pytest runs in a supported terminal emulator like ConEmu, Gnome Terminal, Ptyxis, Windows Terminal, Kitty or Ghostty,
you'll see the progress in the terminal tab or window,
allowing you to monitor pytest's progress at a glance.
This feature is automatically enabled when running in a TTY. It is implemented as an internal plugin. If needed, it can be disabled as follows:
-p no:terminalprogresson the command line or via an environment variablePYTEST_ADDOPTS='-p no:terminalprogress'.addopts = "-p no:terminalprogress".#478: Support PEP420 (implicit namespace packages) as --pyargs target when
consider_namespace_packagesis true in the config.Previously, this option only impacted package imports, now it also impacts tests discovery.
#13678: Added a new
faulthandler_exit_on_timeoutconfiguration option set to "false" by default to let faulthandler interrupt the pytest process after a timeout in case of deadlock.Previously, a faulthandler timeout would only dump the traceback of all threads to stderr, but would not interrupt the pytest process.
-- by
ogrisel.#13829: Added support for configuration option aliases via the
aliasesparameter inParser.addini() <pytest.Parser.addini>.Plugins can now register alternative names for configuration options,
allowing for more flexibility in configuration naming and supporting backward compatibility when renaming options.
The canonical name always takes precedence if both the canonical name and an alias are specified in the configuration file.
Improvements in existing functionality
#13330: Having pytest configuration spread over more than one file (for example having both a
pytest.inifile andpyproject.tomlwith a[tool.pytest.ini_options]table) will now print a warning to make it clearer to the user that only one of them is actually used.-- by
sgaist#13574: The single argument
--versionno longer loads the entire plugin infrastructure, making it faster and more reliable when displaying only the pytest version.Passing
--versiontwice (e.g.,pytest --version --version) retains the original behavior, showing both the pytest version and plugin information.#13823: Added
strict_xfailas an alias to thexfail_strictoption,strict_configas an alias to the--strict-configflag,and
strict_markersas an alias to the--strict-markersflag.This makes all strictness options consistently have configuration options with the prefix
strict_.#13700: --junitxml no longer prints the generated xml file summary at the end of the pytest session when --quiet is given.
#13732: Previously, when filtering warnings, pytest would fail if the filter referenced a class that could not be imported. Now, this only outputs a message indicating the problem.
#13859: Clarify the error message for pytest.raises() when a regex match fails.
#13861: Better sentence structure in a test's expected error message. Previously, the error message would be "expected exception must be <expected>, but got <actual>". Now, it is "Expected <expected>, but got <actual>".
Removals and backward incompatible breaking changes
#12083: Fixed a bug where an invocation such as pytest a/ a/b would cause only tests from a/b to run, and not other tests under a/.
The fix entails a few breaking changes to how such overlapping arguments and duplicates are handled:
If you rely on these behaviors, consider using
--keep-duplicates <duplicate-paths>, which retains its existing behavior (including the bug).#13719: Support for Python 3.9 is dropped following its end of life.
#13766: Previously, pytest would assume it was running in a CI/CD environment if either of the environment variables $CI or $BUILD_NUMBER was defined;
now, CI mode is only activated if at least one of those variables is defined and set to a non-empty value.
#13779: PytestRemovedIn9Warning deprecation warnings are now errors by default.
Following our plan to remove deprecated features with as little disruption as
possible, all warnings of type
PytestRemovedIn9Warningnow generate errorsinstead of warning messages by default.
The affected features will be effectively removed in pytest 9.1, so please consult the
deprecationssection in the docs for directions on how to update existing code.In the pytest
9.0.Xseries, it is possible to change the errors back into warnings as astopgap measure by adding this to your
pytest.inifile:But this will stop working when pytest
9.1is released.If you have concerns about the removal of a specific feature, please add a
comment to
13779.Deprecations (removal in next major release)
monkeypatch.syspath_prepend() <pytest.MonkeyPatch.syspath_prepend>now issues a deprecation warning when the prepended path contains legacy namespace packages (those usingpkg_resources.declare_namespace()).Users should migrate to native namespace packages (
420).See
monkeypatch-fixup-namespace-packagesfor details.Bug fixes
#13445: Made the type annotations of
pytest.skipand friends more spec-complaint to have them work across more type checkers.#13537: Fixed a bug in which
ExceptionGroupwith onlySkippedexceptions in teardown was not handled correctly and showed as error.#13598: Fixed possible collection confusion on Windows when short paths and symlinks are involved.
#13716: Fixed a bug where a nonsensical invocation like
pytest x.py[a](a file cannot be parametrized) was silently treated aspytest x.py. This is now a usage error.#13722: Fixed a misleading assertion failure message when using
pytest.approxon mappings with differing lengths.#13773: Fixed the static fixture closure calculation to properly consider transitive dependencies requested by overridden fixtures.
#13816: Fixed
pytest.approxwhich now returns a clearer error message when comparing mappings with different keys.#13849: Hidden
.pytest.inifiles are now picked up as the config file even if empty.This was an inconsistency with non-hidden
pytest.ini.#13865: Fixed --show-capture with --tb=line.
#13522: Fixed
pytesterin subprocess mode ignored all :attr`pytester.plugins <pytest.Pytester.plugins>` except the first.Fixed
pytesterin subprocess mode silently ignored non-strpytester.plugins <pytest.Pytester.plugins>.Now it errors instead.
If you are affected by this, specify the plugin by name, or switch the affected tests to use
pytester.runpytest_inprocess <pytest.Pytester.runpytest_inprocess>explicitly instead.Packaging updates and notes for downstreams
iniconfigandpackagingwere bumped to1.0.1and22.0.0, respectively.Contributor-facing changes
lsofcommand hanging (e.g. due to unreachable network filesystems), with the affected selftests being skipped after 10 seconds.gh pr newcommand inscripts/prepare-release-pr.py.The script now uses
gh pr createwhich is compatible with GitHub CLI v2.0+.ogrisel.tox>=4is now required when contributing to pytest.pytest_addoption(),pytest_configure(), andcacheshow()functions incacheprovider.py.Miscellaneous internal changes
-o/--override-ini) are now processed during startup rather than duringconfig.getini() <pytest.Config.getini>.Configuration
📅 Schedule: (in timezone Europe/Berlin)
🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.
♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
👻 Immortal: This PR will be recreated if closed unmerged. Get config help if that's undesired.
This PR has been generated by Mend Renovate CLI.
3cb6d7ff81ab4e74f7b3Edited/Blocked Notification
Renovate will not automatically rebase this PR, because it does not recognize the last commit author and assumes somebody else may have edited the PR.
You can manually request rebase by checking the rebase/retry box above.
⚠️ Warning: custom changes will be lost.
6f6d60a0b8539b23dd71