fix(deps): update security fixes [security] (major) #112

Merged
dawid merged 2 commits from renovate/major-security into main 2026-11-10 10:09:47 +00:00
Collaborator

This PR contains the following updates:

Package Change Age Confidence
langchain (changelog) ==0.3.30 → ==1.3.9 age confidence
langchain-core (changelog) ==0.3.86 → ==1.2.22 age confidence
langchain-text-splitters (changelog) ==0.3.11 → ==1.1.2 age confidence
langgraph (source, changelog) ==0.4.1 → ==1.0.10 age confidence
pytest (changelog) ==8.4.2 → ==9.0.3 age confidence

LangChain: 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
  • Confidentiality: disclosure of file contents outside the intended root/sandbox.
  • Authorization: path-prefix bypass can grant access to sibling resources beyond the intended subtree.
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 Score: 5.1 / 10 (Medium)
  • Vector String: CVSS:3.1/AV:L/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N

References

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 Score: 5.5 / 10 (Medium)
  • Vector String: CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N

References

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 arbitrary image_url values 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:

  • Responses are not returned to the attacker (blind SSRF)
  • Default 5-second timeout limits resource exhaustion
  • Non-image responses fail at PIL image parsing
Impact

An attacker who can control image URLs passed to get_num_tokens_from_messages() can:

  • Trigger HTTP requests from the application server to arbitrary internal or external URLs
  • Cause the server to access internal network resources (private IPs, cloud metadata endpoints)
  • Cause minor resource consumption through image downloads (bounded by timeout)

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:

  1. get_num_tokens_from_messages() processes messages containing image_url content blocks
  2. For images without detail: "low", it calls _url_to_size() to fetch the image and compute token counts
  3. _url_to_size() performs httpx.get(image_source) on any URL without validation
  4. Prior to the patch, there was no SSRF protection, size limits, or explicit timeout

File: libs/partners/openai/langchain_openai/chat_models/base.py

Patches

The vulnerability has been patched in langchain-openai==1.1.9 (requires langchain-core==1.2.11).

The patch adds:

  1. SSRF validation using langchain_core._security._ssrf_protection.validate_safe_url() to block:
    • Private IP ranges (RFC 1918, loopback, link-local)
    • Cloud metadata endpoints (169.254.169.254, etc.)
    • Invalid URL schemes
  2. Explicit size limits (50 MB maximum, matching OpenAI's payload limit)
  3. Explicit timeout (5 seconds, same as httpx.get default)
  4. Allow disabling image fetching via allow_fetching_images=False parameter
Workarounds

If you cannot upgrade immediately:

  1. Sanitize input: Validate and filter image_url values before passing messages to token counting or model invocation
  2. Use network controls: Implement egress filtering to prevent outbound requests to private IPs

Severity

  • CVSS Score: 3.7 / 10 (Low)
  • Vector String: CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:L

References

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 arbitrary image_url values 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:

  • Responses are not returned to the attacker (blind SSRF)
  • Default 5-second timeout limits resource exhaustion
  • Non-image responses fail at PIL image parsing
Impact

An attacker who can control image URLs passed to get_num_tokens_from_messages() can:

  • Trigger HTTP requests from the application server to arbitrary internal or external URLs
  • Cause the server to access internal network resources (private IPs, cloud metadata endpoints)
  • Cause minor resource consumption through image downloads (bounded by timeout)

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:

  1. get_num_tokens_from_messages() processes messages containing image_url content blocks
  2. For images without detail: "low", it calls _url_to_size() to fetch the image and compute token counts
  3. _url_to_size() performs httpx.get(image_source) on any URL without validation
  4. Prior to the patch, there was no SSRF protection, size limits, or explicit timeout

File: libs/partners/openai/langchain_openai/chat_models/base.py

Patches

The vulnerability has been patched in langchain-openai==1.1.9 (requires langchain-core==1.2.11).

The patch adds:

  1. SSRF validation using langchain_core._security._ssrf_protection.validate_safe_url() to block:
    • Private IP ranges (RFC 1918, loopback, link-local)
    • Cloud metadata endpoints (169.254.169.254, etc.)
    • Invalid URL schemes
  2. Explicit size limits (50 MB maximum, matching OpenAI's payload limit)
  3. Explicit timeout (5 seconds, same as httpx.get default)
  4. Allow disabling image fetching via allow_fetching_images=False parameter
Workarounds

If you cannot upgrade immediately:

  1. Sanitize input: Validate and filter image_url values before passing messages to token counting or model invocation
  2. Use network controls: Implement egress filtering to prevent outbound requests to private IPs

Severity

  • CVSS Score: 3.7 / 10 (Low)
  • Vector String: CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:L

References

This data is provided by OSV and the PyPI Advisory Database (CC-BY 4.0).


LangChain Core has Path Traversal vulnerabilites in legacy load_prompt functions

CVE-2026-34070 / GHSA-qh6h-p6c9-ff54 / PYSEC-2026-2193

More information

Details

Summary

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).

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 the dumpd/dumps/load/loads serialization APIs in langchain_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-core
File: langchain_core/prompts/loading.py
Affected 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
Config key Loaded by Readable extensions
template_path, suffix_path, prefix_path _load_template() .txt
examples (when string) _load_examples() .json, .yaml, .yml
example_prompt_path _load_few_shot_prompt() .json, .yaml, .yml

None 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:

  • .txt files: cloud-mounted secrets (/mnt/secrets/api_key.txt), requirements.txt, internal system prompts
  • .json/.yaml files: cloud credentials (~/.docker/config.json, ~/.azure/accessTokens.json), Kubernetes manifests, CI/CD configs, application settings

This 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
from langchain_core.prompts.loading import load_prompt_from_config

##### Reads /tmp/secret.txt via absolute path injection
config = {
    "_type": "prompt",
    "template_path": "/tmp/secret.txt",
    "input_variables": [],
}
prompt = load_prompt_from_config(config)
print(prompt.template)  # file contents disclosed

##### Reads ../../etc/secret.txt via directory traversal
config = {
    "_type": "prompt",
    "template_path": "../../etc/secret.txt",
    "input_variables": [],
}
prompt = load_prompt_from_config(config)

##### Reads arbitrary .json via few-shot examples
config = {
    "_type": "few_shot",
    "examples": "../../../../.docker/config.json",
    "example_prompt": {
        "_type": "prompt",
        "input_variables": ["input", "output"],
        "template": "{input}: {output}",
    },
    "prefix": "",
    "suffix": "{query}",
    "input_variables": ["query"],
}
prompt = load_prompt_from_config(config)
Mitigation

Update langchain-core to >= 1.2.22.

The fix adds path validation that rejects absolute paths and .. traversal sequences by default. An allow_dangerous_paths=True keyword argument is available on load_prompt() and load_prompt_from_config() for trusted inputs.

As described above, these legacy APIs have been formally deprecated. Users should migrate to dumpd/dumps/load/loads from langchain_core.load.

Credit

Severity

  • CVSS Score: 7.5 / 10 (High)
  • Vector String: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N

References

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 Score: 7.5 / 10 (High)
  • Vector String: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N

References

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 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.

Affected versions
  • langchain-text-splitters < 1.1.2
Patched versions
  • langchain-text-splitters >= 1.1.2 (requires langchain-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:

validate_safe_url(url, allow_private=False, allow_http=True)
response = requests.get(url, timeout=timeout, **kwargs)
Attack scenario
  1. A developer passes external URLs to split_text_from_url(), relying on its
    built-in validate_safe_url() check to block requests to internal networks.
  2. An attacker supplies a URL pointing to a public host they control. The URL
    passes validate_safe_url() (public hostname, public IP).
  3. The attacker's server responds with a 302 redirect 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).
  4. requests.get() follows the redirect automatically. The redirect target is
    not revalidated.
  5. The response body is parsed and returned as Document objects to the
    application.

Notes:

  • The core issue is a bypass of an explicitly provided SSRF protection.
    split_text_from_url() included validate_safe_url() specifically to be
    safe with untrusted URLs — the redirect loophole defeated that guarantee.
  • Cloud metadata endpoints that require special headers (AWS IMDSv2, GCP, Azure)
    are not reachable through this bug because the attacker does not control
    request headers. AWS IMDSv1, which requires no headers, is reachable.
  • Data exfiltration requires the application to return Document contents to the
    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 (SSRFSafeSyncTransport from langchain-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 to split_text() directly.

Severity

  • CVSS Score: 6.5 / 10 (Medium)
  • Vector String: CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N

References

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 Score: 6.5 / 10 (Medium)
  • Vector String: CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N

References

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 Score: 7.2 / 10 (High)
  • Vector String: CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H

References

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:

  • use a persistent checkpointer (database, remote store, shared filesystem, etc.),
  • load/resume from checkpoints, and
  • operate in an environment where an attacker could gain privileged write access to checkpoint data in the backing store.

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
  • Potential arbitrary code execution or other unsafe side effects during checkpoint deserialization.
  • Escalation from “write access to checkpoint store” to “code execution in the application runtime,” which may expose runtime secrets or provide access to other systems the runtime can reach.
Exploitation scenario (high level)
  1. Attacker gains privileged write access to the checkpoint store (for example, via database compromise, leaked credentials, or abuse of an administrative data path).
  2. Attacker writes a crafted checkpoint payload containing msgpack data intended to reconstruct dangerous objects.
  3. Application resumes and deserializes the checkpoint; unsafe reconstruction could execute attacker-controlled behavior.
Mitigation / remediation

LangGraph provides an allowlist-based hardening mechanism for msgpack checkpoint deserialization.

Strict mode (environment variable)
  • LANGGRAPH_STRICT_MSGPACK
    • When set truthy (1, true, yes), the default msgpack deserialization policy becomes strict.
    • Concretely: JsonPlusSerializer() will default allowed_msgpack_modules to None (strict) instead of True (warn-and-allow), unless allowed_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 datetime types, uuid.UUID, decimal.Decimal, set/frozenset/deque, ipaddress types, pathlib paths, zoneinfo.ZoneInfo, compiled regex patterns, and selected LangGraph internal types).

Automatically derived allowlist (only when compiling graphs)

When LANGGRAPH_STRICT_MSGPACK is enabled and StateGraph is compiled, LangGraph derives an allowlist from the graph’s schemas and channels and applies it to the checkpointer.

  • The allowlist is built by walking the state/input/output/context schemas (plus node/branch input schemas) and channel value/update types. It includes Pydantic v1/v2 models, dataclasses, enums, TypedDict field types, and common typing constructs (containers, unions, Annotated).
  • LangGraph also includes a curated set of common LangChain message classes.

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 configure allowed_msgpack_modules themselves.

Operational guidance
  • Treat checkpoint stores as integrity-sensitive. Restrict write access and rotate credentials if compromise is suspected.
  • Enable strict mode (LANGGRAPH_STRICT_MSGPACK=true) in production if feasible, and rely on schema-driven allowlisting to reduce incompatibilities.
  • Avoid providing custom msgpack deserialization hooks that reconstruct arbitrary types unless checkpoint data is fully trusted.
Limitations / important notes
  • If a checkpointer implementation does not support allowlist application (i.e., does not implement with_allowlist), allowlist enforcement may be skipped (with a warning). In that situation, strict expectations may not hold.
  • If an application supplies a custom msgpack unpack hook (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 Score: 6.8 / 10 (Medium)
  • Vector String: CVSS:3.1/AV:A/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H

References

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 Score: 6.8 / 10 (Medium)
  • Vector String: CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:L/I:L/A:L

References

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 Score: 6.8 / 10 (Medium)
  • Vector String: CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:L/I:L/A:L

References

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==1.0.10

Compare Source

Changes since 1.0.10rc1

  • release(langgraph) 1.0.10 (#​6967)
  • release(checkpoint): 0.4.1 (#​6966)
  • chore: add serde events (#​6954)
  • chore: update defaults (#​6953)
  • release: rc2 (#​6949)
  • chore(deps-dev): bump the all-dependencies group across 1 directory with 3 updates (#​6946)

v1.0.9: langgraph==1.0.9

Compare Source

Changes since 1.0.8

  • release: langgraph + prebuilt (#​6875)
  • fix: sequential interrupt handling w/ functional API (#​6863)
  • chore: state_updated_at sort by (#​6857)
  • chore: bump orjson (#​6852)
  • chore: conformance testing (#​6842)
  • chore(deps): bump the all-dependencies group in /libs/langgraph with 6 updates (#​6815)
  • chore(deps): bump protobuf from 6.33.4 to 6.33.5 in /libs/langgraph (#​6833)
  • chore(deps): bump cryptography from 46.0.3 to 46.0.5 in /libs/langgraph (#​6837)
  • chore(deps): bump nbconvert from 7.16.6 to 7.17.0 in /libs/langgraph (#​6832)
  • chore: server runtime type (#​6774)
  • refactor: replace bare except with BaseException in AsyncQueue (#​6765)

v1.0.8: langgraph==1.0.8

Compare Source

Changes since 1.0.7

  • release(langgraph): 1.0.8 (#​6757)
  • chore: shallow copy futures (#​6755)
  • fix: pydantic messages double streaming (#​6753)
  • chore(deps-dev): bump ruff from 0.14.7 to 0.14.11 in /libs/sdk-py (#​6673)
  • chore: Omit lock when using connection pool (#​6734)
  • docs: enhance Runtime and ToolRuntime class descriptions for clarity (#​6689)
  • docs: add clarity to use of thread_id (#​6515)
  • docs: add docstrings to add_node overloads (#​6514)
  • docs: update notebook links and add archival notices for examples (#​6720)
  • release(cli): 0.4.12 (#​6716)

v1.0.7: langgraph==1.0.7

Compare Source

Changes since 1.0.6

  • release: langgraph and prebuilt 1.0.7 (#​6712)
  • fix: aiosqlite's breaking change (#​6699)
  • chore(deps): upgrade dependencies with uv lock --upgrade (#​6671)
  • chore: update twitter URLs (#​6683)

v1.0.6: langgraph==1.0.6

Compare Source

Changes since 1.0.5

  • fix: flip default on base cache (#​6677)
  • fix: change default recursion limit (#​6676)
  • fix: sanitize namespace for deeply nested graph jumps (#​6665)
  • fix: typo: saved the world "BaseMessge" to "BaseMessage" (#​6639)
  • release(cli): relax api bounds (#​6606)
  • feat(checkpoint): Validate checkpointer type at compile time (#​6586)

v1.0.5: langgraph==1.0.5

Compare Source

Changes since 1.0.4

  • release(langgraph): bump to 1.0.5 (#​6582)
  • feat(sdk-py): emit id as part of stream events (#​6581)
  • fix: update readme (#​6570)
  • release(checkpoint-postgres): 3.0.1 (#​6568)
  • release(checkpoint-sqlite): 3.0.1 (#​6566)
  • chore(cli): Pass through webhook configuration in dev server (#​6557)
  • feat: custom encryption at rest (#​6482)
  • chore: fix links for docs (#​6538)
  • chore: Bump lockfile (#​6537)
  • feat: Include pagination in assistants search response (#​6526)

v1.0.4

Compare Source

Changes since 1.0.3

  • release(langgraph): 1.0.4 (#​6502)
  • chore: pop thread ID from configurable fields in remote graph (#​6497)
  • chore: clean up some refs (#​6487)
  • chore(sdk-py): Add more type checking. (#​6479)
  • fix: interrupt stream mode values (#​6475)
  • release: langgraph-prebuilt 1.0.5 (#​6473)
  • refactor: separate prepare_push_* functions (#​6450)
  • fix: dep warnings in prebuilt (#​6443)
  • release: prebuilt 1.0.3 (#​6441)

v1.0.3

Compare Source

Changes since 1.0.2

  • chore: langgraph patch release (#​6429)
  • feat(docs): warn that StateGraph is a builder class (#​6417)
  • fix(docs): PartialState rendering in MkDocs (#​6416)
  • fix(docs): synchronize invoke and ainvoke docstrings (#​6415)
  • fix(docs): synchronize stream and astream docstrings (#​6414)
  • docs(langgraph): Fix docstring code examples of task function (#​6410)
  • fix(langgraph): Unexpected behavior for stream_mode sequences that are not lists (#​6354)
  • fix: fix previoius edge cases such as 0 (#​6379)
  • fix(langgraph): export REMOVE_ALL_MESSAGES in all to fix linting (#​6375)
  • style: update docstrings to reference StateGraph (#​6308)
  • docs(langgraph): update streaming guide links (#​6314)
  • chore: add pyproject.toml links (#​6364)
  • docs(langgraph): Fix typo in docstring of PregelLoop.tick (#​6407)
  • chore(checkpoint-postgres): bump to 3.0.1 (#​6402)
  • fix(langgraph): do not apply pending writes when updating state (#​6389)
  • chore: update ormsgpack minbound and add OPT_REPLACE_SURROGATES (#​6395)
  • release(cli): 0.4.7 expand api bounds (#​6390)
  • chore: style fixes for refs (#​6365)

v1.0.2

Compare Source

Changes since 0.6.11

  • chore: bump prebuilt dep for lg (#​6361)
  • chore: port tool node improvements back to langgraph (#​6321)
  • chore: bump core dep for prebuilt (#​6323)
  • fix(langgraph): dont persist UntrackedValue (#​6316)
  • feat(langgraph): add Overwrite to bypass reducer (#​6286)
  • fix: replace python.langchain links with new docs.langchain (#​6352)
  • release: Checkpointers 3.0 (#​6313)
  • chore: Restrict "json" type deserialization (#​6269)
  • release: langgraph + langgraph-prebuilt v1.0.0 (#​6300)
  • feat: adding cursory Python 3.14 support (#​6298)
  • style: fixes for ref docs (#​6297)
  • chore: release rcs for prebuilt + langgraph (#​6296)
  • chore: drop Python 3.9 (and syntax) (#​6289)
  • docs: style linting (#​6260)
  • release(langgraph): v1 working branch (#​6093)
  • fix: rename away from LangGraph Platform (#​6281)

v1.0.1

Compare Source

Changes since 1.0.0

  • release: Checkpointers 3.0 (#​6313)
  • chore: Restrict "json" type deserialization (#​6269)

v1.0.0

Compare Source

Changes since 1.0.0rc1

  • release: langgraph + langgraph-prebuilt v1.0.0 (#​6300)
  • feat: adding cursory Python 3.14 support (#​6298)
  • style: fixes for ref docs (#​6297)

v0.6.11

Compare Source

Changes since 0.6.10

  • chore: Allow checkpoint 3.0 in 0.6.* (#​6315)

v0.6.10

Compare Source

Changes since 0.6.9

  • chore(langgraph): bump langgraph version (#​6257)
  • fix(langgraph): revert selective interrupt task scheduling (#​6252)

v0.6.9

Compare Source

Changes since 1.0.0a4

  • chore(langgraph): bump version (#​6245)
  • chore(checkpoint): bump patch version (#​6244)
  • fix(langgraph): selective interrupt task scheduling (#​6158)
  • fix(langgraph): task result from stream mode debug / tasks should match format from get_state_history / get_state (#​6233)
  • fix(langgraph): don't use rst code blocks in docstrings (#​6231)
  • docs(langgraph): standardize version-added admonitions (#​6230)
  • fix(langgraph): fix supersteps not populating task.result field (#​6195)
  • fix(langgraph): revert -- reuse cached writes on nested resume to prevent task re-execution (#​6227)
  • chore(checkpoint-postgres): bump version (#​6222)

v0.6.8

Compare Source

Changes since 1.0.0a3

  • release(langgraph): 0.6.8 (#​6215)
  • fix(langgraph): handle multiple annotations w/ BaseChannel detection (#​6210)
  • fix(langgraph): CheckpointTask.state can be a StateSnapshot (#​6201)
  • chore(langgraph): clean up ruff format config (#​6188)
  • fix(langgraph): cleanup orphaned waiter task in AsyncPregelLoop (#​6167)
  • fix(langgraph): fix graph rendering for defer=True (#​6130)
  • fix(langgraph): reuse cached writes on nested resume to prevent task re-execution (#​6161)
  • chore(sdk-py): allow UUIDs in config (#​6151)
  • revert(langgraph): restore logic to surface interrupts for stream_mod… (#​6141)
  • chore(langgraph): Log when no values event is emitted from RemoteGraph (#​6140)
  • chore(cli): Add config schema (#​6142)
  • fix(langgraph): get_graph generates unexpected conditional edge (#​6122)
  • fix(langgraph): type checking for async w/ functional API (#​6126)
  • feat(langgraph): prevent arbitrary resumes w/ multiple pending interrupts (#​6108)
  • fix(langgraph): key error on runtime for config w/o configurable (#​6106)

v0.6.7

Compare Source

Changes since 1.0.0a2

  • chore: update emphemeral local (#​6091)
  • fix: Unwrap Required/NotRequired special forms before resolving channel/reducer annotations (#​6080)
  • monorepo support in CLI (#​6028)

v0.6.6

Compare Source

Changes since 0.6.5

  • fix(langgraph): Remote Baggage (#​5964)
  • chore(langgraph): Add passthrough params/headers to invoke/stream/etc. (#​5940)
  • feat(sdk-py): client qparams (#​5918)

v0.6.5

Compare Source

Changes since 0.6.4

  • release(langgraph): 0.6.5 (#​5901)
  • fix: Persist resume_map values (#​5898)
  • feat(langgraph): implement redis node level cache (#​5834)

v0.6.4

Compare Source

Changes since 0.6.3

  • release: langgraph + prebuilt 0.6.4 (#​5854)
  • fix: mypy issue with conditional edges (#​5851)
  • fix(langgraph): support emitting messages from subgraphs when messages mode explicitly requested (#​5836)
  • perf(langgraph): Save updated_channels to checkpoint (#​5828)
  • chore(langgraph): deprecate MessageGraph (#​5843)

v0.6.3

Compare Source

Changes since 0.6.2

  • fix(langgraph): Tidy up AgentState (#​5801)
  • release: langgraph + prebuilt 0.6.3 (#​5799)
  • fix(langgraph): Add warning for incorrect node signature with mistyped config param (#​5798)
  • fix(langgraph): fix up deprecation warnings (#​5796)
  • feat(langgraph): add durability mode for invoke and ainvoke (#​5771)
  • fix(docs): Add missing imports to make examples runnable (#​5477)
  • fix(langgraph): Remove duplicate call to ensure_config (#​5768)

v0.6.2

Compare Source

Changes since 0.6.1

  • fix(prebuilt): assign context_schema to config_schema with correct condition (#​5746)
  • feat(langgraph): Add context coercion for LangGraph runtime (#​5736)

v0.6.1

Compare Source

Changes since 0.6.0

  • fix(langgraph): use parent runtime when available (#​5707)
  • fix(langgraph): inject config even when optional (#​5708)

v0.6.0

Compare 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):
from langchain_core.runnables import RunnableConfig
from langgraph.graph import StateGraph

def node(state: State, config: RunnableConfig):
    # verbose .get() access pattern for nested dicts    
    user_id = config.get("configurable", {}).get("user_id")
    db_conn = config.get("configurable", {}).get("db_connection")
    ...

builder = StateGraph(state_schema=State, config_schema=Config)

# add nodes, edges, compile the graph...

# nested runtime context passed to config's configurable key
result = graph.invoke(
    {'input': 'abc'}, 
    config={'configurable': {'user_id': '123', 'db_connection': 'conn_mock'}}
)
After (v0.6):
from dataclasses import dataclass

from langgraph.graph import StateGraph
from langgraph.runtime import Runtime

@dataclass
class Context:
    """Context schema defined by the developer."""    
    user_id: str    
    db_connection: str
    
def node(state: State, runtime: Runtime[Context]):
    # type safe access to context attributes    
    user_id = runtime.context.user_id
    db_conn = runtime.context.db_connection
    ...

builder = StateGraph(state_schema=State, context_schema=Context)

# add nodes, edges, compile the graph...

# top level context arg is typed as Context for autocomplete and type checking
result = graph.invoke(
    {'input': 'abc'},
    context=Context(user_id='123', db_conn='conn_mock')
)

The Runtime class provides a single interface for accessing information like:

  • context: static data passed at the start of a run
  • store: storage mechanism for long term memories
  • stream_writer: custom function for writing to the graph’s output stream
  • for functional API users, previous is also available: the previous return value for the given thread

Now, instead of injecting all of the above as separate parameters to node functions,
developers can access them all through a single runtime parameter.

Migration Path
  • config_schema is deprecated in favor of context_schema, and will be removed in v2.0.0
  • The new API maintains backward compatibility for existing code
  • Gradual migration is supported with deprecation warnings for config_schema
  • get_config_jsonschema is deprecated in favor of get_context_jsonschema (though this is generally only used for graph introspection and not by most langgraph users)

🔀 Dynamic model & tool selection

create_react_agent can now dynamically choose both the model and tools at runtime using a custom context object:

from langgraph.prebuilt import create_react_agent

@dataclass
class CustomContext:
    provider: Literal["anthropic", "openai"]
    tools: list[str]

def select_model(state, Runtime[Context]):
    model = {
        "openai": openai_model,
        "anthropic": anthropic_model,
    }[runtime.context.provider]

    selected_tools = [
        tool for tool in [weather, compass]
        if tool.name in runtime.context.tools
    ]

    return model.bind_tools(selected_tools)

agent = create_react_agent(
	select_model, 
         # Initialize the agent with all known tools
	tools=[weather, compass]
)

Then invoke the agent with your desired settings:

agent.invoke(
    some_input,
    context=CustomContext(provider="openai", tools=["compass"])
)

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

  • Choose between three durability modes:
    • "exit" - Save checkpoint only when the graph exits
      • equivalent to checkpoint_during=False
      • Least durable, fastest
    • "async" - Save checkpoint asynchronously while next step executes
      • equivalent to checkpoint_during=True
      • Moderately durable, mid speed
    • "sync" - Save checkpoint synchronously before next step
      • New!
      • Highly durable, slowest
Migration Path

checkpoint_during is now deprecated in favor of the new durability argument. 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 StateGraph and Pregel interfaces are now generic over a graph’s:

  • state_schema
  • context_schema
  • input_schema
  • output_schema

This means that:

  • Node signatures are type checked at graph creation time
  • Input to invoke / stream is type checked against the relevant schema
  • context available via the aforementioned Runtime class matches the context_schema

🔧 A Refined Interrupt Interface

In preparation for v1.0, we’ve made a few changes to the Interrupt interface.
Interrupts now have two attributes:

  • id - a unique identifier for the interrupt
  • value - the interrupt value

In v0.6, we’ve removed the following attributes from the Interrupt class:

  • when - this was always "during" and offered no practical value
  • resumable - functionally, this is always True
  • ns - this information is now stored in a condensed format in the id attribute
  • interrupt_id has been deprecated in favor of id, 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:

Old Import New Import Status
from langgraph.pregel.types import ... from langgraph.types import ... ⚠️ Deprecated - Will be removed in V2
from langgraph.constants import Send from langgraph.types import Send ⚠️ Deprecated - Will be removed in V2
from langgraph.constants import Interrupt from langgraph.types import Interrupt ⚠️ Deprecated - Will be removed in V2
from langgraph.channels import <ErrorClass> from langgraph.errors import <ErrorClass> ❌ Removed - All errors now centralized in langgraph.errors
from langgraph.constants import TAG_NOSTREAM_ALT from langgraph.constants import NOSTREAM ❌ Removed - Deprecated constant removed

🎯 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

  • feat(sdk-py): sdk support for context API (#​5566)
  • feat(langgraph): Implement durability mode argument (#​5432)
  • refactor(langgraph): improve Runtime interface re patch/overrides (#​5546)
  • refactor(langgraph): make constants generally private with a few select exports (#​5529)
  • refactor(langgraph): move private typing constructs in constants.py -> _internal/_typing.py (#​5518)
  • feat(langgraph): new context api (replacing config['configurable'] and config_schema) (#​5243)
  • feat(langgraph): add type checking for matching node signatures vs input_schema for add_node (#​5424)
  • change[langgraph]: clean up Interrupt interface for v1 (#​5405)
  • langgraph[change]: solidify public/private differentiations (#​5252)

v0.5.4

Compare Source

Changes since 0.5.3

  • feat(langgraph): Handle ParentCommand in RemoteGraph (#​5600)
  • fix(langgraph): ignore write to END with Command (#​5601)
  • fix(langgraph): add stacklevel=2 to the warnings to point to the caller’s codes (#​5457)

v0.5.3

Compare Source

Changes since 0.5.2

  • release(langgraph): v0.5.3 (#​5498)
  • chore[deps]: upgrade dependencies with uv lock --upgrade (#​5471)
  • docs(checkpoint-postgres): fix typo in comment (#​5486)
  • chore: add forum to readme (#​5488)
  • fix(langgraph): remove ABC spec for PregelProtocol (#​5485)
  • fix(langgraph): replace _state_schema to state_schema when accessing StateGraph (#​5436)
  • fix(checkpoint-postgres): Remove python invalid escape warning (#​5441)

v0.5.2

Compare Source

Changes since 0.5.1

  • patch[langgraph]: Fix hint for invoke/stream to allow for Command and None (#​5414)

v0.5.1

Compare Source

Changes since 0.5.0

  • langgraph[fix]: remove deprecated pydantic logic + fix schema gen behavior for typed dicts (#​5296)
  • prebuilt[patch]: import recognized tool message content block types from langchain-core (#​5275)

v0.5.0

Compare 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:

  • lock in the few API surfaces that still felt “fuzzy”
  • prune internal code paths that made distributed execution harder
  • land long-queued performance work

0.5 is that housekeeping release.


Headline changes

1. A leaner, stricter StateGraph
  1. state_schema is 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.

  2. input/output → input_schema/output_schema

    The old names still work but raise a deprecation warning.

    graph = StateGraph(
        state_schema=MyState,
        input_schema=UserQuery,
        output_schema=AssistantResponse,
    )
    
  3. New NodeBuilder utility

    A simpler, declarative way to create nodes and attach them to channels. The old Channel.subscribe_to helper 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:
graph.stream(stream_mode="tasks")        # only task-level updates
graph.stream(stream_mode="checkpoints")  # only checkpoint deltas

This makes it cheaper to subscribe only to the information you need.

3. Checkpointing overhaul
  • Smaller blobs & faster restores
    Redundant keys have been dropped and per-task writes are stored directly.
  • Seamless migrations
    Legacy “pending _sends” data is auto-migrated the first time it is loaded. Custom checkpointers continue to work unchanged.
4. Better serialization
  • JsonPlusSerializer now handles NumPy arrays stored in your state (including Fortran-ordered ones) without falling back to pickle.

Minor breaking changes you might notice

  1. state_schema required – add it if you were passing only input and output schemas instead (very rare).
  2. input / output renaming – rename to input_schema / output_schema.
  3. Subclassing internals – if you subclassed PregelNode and Runnable, drop the latter.

Nothing else should require code changes.


How to upgrade

pip install -U "langgraph>=0.5"

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

  • feat: crons sorting sdk (#​5197)
  • feat(langgraph): task masquerading with update state
  • Add print_mode= arg to invoke/stream (#​5201)
  • Fix bug where Command(update=) could be ignored if there was a 2nd interrupt after it
  • Reduce extraneous keys in checkpoint.metadata
  • If FuturesDict callback has been GCed, don't call it
  • Add migration for pending_sends (#​5106)
  • Introduce "tasks" and "checkpoints" stream modes
  • Add migration for pending_sends
  • Restore compatibility with custom checkpointer classes created in prior versions
  • Revert "Remove UntrackedValue channel"
  • Revert "Remove MessageGraph (#​4875)"
  • fix(langgraph): remove deprecated output usage in favor of output_schema (#​5095)
  • refactor(langgraph): Remove PregelNode's inheritance from Runnable (#​5093)
  • Remove support for node reading a single managed value
  • PregelLoop: Simplify tick() method (#​5080)
  • serialize/deserialize pandas with pickle fallback (#​5057)
  • Remove code paths no longer needed
  • Avoid saving checkpoints for subgraphs when checkpoint_during=False (#​5051)
  • Remove gitmcp badge (#​5055)
  • Avoid saving checkpoints for subgraphs when checkpoint_during=False
  • Clean up things for Matt!
  • Support numpy array serialization in JsonPlusSerializer (#​5035)
  • Update ormsgpack (#​5034)
  • deprecate input and output in favor of input_schema and output_schema (#​4983)
  • using StateT as default for InputT
  • lint: use pep 604 union syntax and pep 585 generic syntax (#​4963)
  • docs: remove references to StateGraph(dict) (#​4964)
  • rename retry -> retry_policy (#​4957)
  • Fix step_timeout causing ParentCommand/GraphInterrupt exception to bubble up
  • docs: fix a grammar issue in multiple files (#​4935)
  • Make it possible to run test command without docker installed (#​4948)
  • Fix async callback manager tag handling (#​4949)
  • Pass tags when configuring async callback manager
  • Allow same-name channels and nodes in StateGraph (#​4944)
  • Make it possible to run test command without docker installed
  • docs: fix ensure_config docstring
  • Allow same-name channels and nodes in StateGraph
  • Improve type checking on graph init and invoke/stream (#​4932)
  • docs: fix a grammar issue in multiple files
  • Fix makefile command file for dev server - pidfile was always empty
  • Fix Command(graph=PARENT) when used together w checkpointer=True
  • Remove unused deprecation decorator/warning (#​4917)
  • Require state_schema in StateGraph.__init__ (#​4897)
  • Remove add_conditional_edge(..., then=) (#​4893)
  • Avoid repeated runtime calls to get_type_hints
  • Remove MessageGraph (#​4875)
  • Remove non-state Graph
  • Remove UntrackedValue channel
  • Remove UntrackedValue channel (#​4859)
  • Flip default for checkpoint_during
  • Remove Channel node builder
  • Update managed value usage in local_read (#​4854)
  • Modify stream mode messages and custom to respect subgraphs=False (#​4843)
  • Remove Checkpoint.writes (#​4822)
  • Remove Checkpoint.pending_sends (#​4820)
  • Pregel: Add NodeBuilder class to replace Channel.subscribe_to (#​4819)
  • Remove dict subclasses used for values/updates stream chunks (#​4816)
  • Remove old checkpoint test fixtures (#​4814)
  • Remove postgres shallow checkpointer (#​4813)
  • Remove ChannelsManager, ManagedValues are now static classes and never instantiated (#​4812)
  • Remove Context channel / managed value, Remove SharedValue (#​4857)
  • Pregel: Add NodeBuilder class to replace Channel.subscribe_to
  • Remove old checkpoint test fixtures
  • Remove postgres shallow checkpointer
  • Remove ChannelsManager, ManagedValues are now static classes and never instantiated
  • Remove Context channel / managed value, Remove SharedValue

v0.4.10

Compare Source

Changes since 0.4.9

  • Revert change to default value of checkpoint_during arg (#​5177)
  • Revert change to default value of checkpoint_during arg
  • Fix bug where Command(update=) could be ignored if there was a 2nd interrupt after it
  • Reduce extraneous keys in checkpoint.metadata
  • If FuturesDict callback has been GCed, don't call it
  • langgraph 0.5.0rc1
  • langgraph 0.5.0rc0
  • Preparation for 0.5 release: langgraph-checkpoint (#​5124)
  • Preparation for 0.5 release
  • Add migration for pending_sends (#​5106)
  • Lint
  • Introduce "tasks" and "checkpoints" stream modes
  • Add migration for pending_sends
  • Restore compatibility with custom checkpointer classes created in prior versions
  • Revert "Remove UntrackedValue channel"
  • Revert "Remove MessageGraph (#​4875)"
  • fix(langgraph): remove deprecated output usage in favor of output_schema (#​5095)
  • refactor(langgraph): Remove PregelNode's inheritance from Runnable (#​5093)
  • Remove support for node reading a single managed value
  • PregelLoop: Simplify tick() method (#​5080)
  • serialize/deserialize pandas with pickle fallback (#​5057)
  • Remove code paths no longer needed
  • Avoid saving checkpoints for subgraphs when checkpoint_during=False (#​5051)
  • Remove gitmcp badge (#​5055)
  • Update existing
  • Add tests
  • Avoid saving checkpoints for subgraphs when checkpoint_during=False
  • Clean up things for Matt!
  • Support numpy array serialization in JsonPlusSerializer (#​5035)
  • Update ormsgpack (#​5034)
  • deprecate input and output in favor of input_schema and output_schema (#​4983)
  • using StateT as default for InputT
  • lint: use pep 604 union syntax and pep 585 generic syntax (#​4963)
  • docs: remove references to StateGraph(dict) (#​4964)
  • rename retry -> retry_policy (#​4957)
  • Fix step_timeout causing ParentCommand/GraphInterrupt exception to bubble up
  • docs: fix a grammar issue in multiple files (#​4935)
  • Make it possible to run test command without docker installed (#​4948)
  • Fix async callback manager tag handling (#​4949)
  • Pass tags when configuring async callback manager
  • Allow same-name channels and nodes in StateGraph (#​4944)
  • Make it possible to run test command without docker installed
  • docs: fix ensure_config docstring
  • Allow same-name channels and nodes in StateGraph
  • Improve type checking on graph init and invoke/stream (#​4932)
  • docs: fix a grammar issue in multiple files
  • Fix makefile command file for dev server - pidfile was always empty
  • Fix Command(graph=PARENT) when used together w checkpointer=True
  • Remove unused deprecation decorator/warning (#​4917)
  • Require state_schema in StateGraph.__init__ (#​4897)
  • Remove add_conditional_edge(..., then=) (#​4893)
  • Avoid repeated runtime calls to get_type_hints
  • Remove MessageGraph (#​4875)
  • Remove non-state Graph
  • Remove UntrackedValue channel
  • Remove UntrackedValue channel (#​4859)
  • Flip default for checkpoint_during
  • Remove Channel node builder
  • Update managed value usage in local_read (#​4854)
  • Modify stream mode messages and custom to respect subgraphs=False (#​4843)
  • Remove Checkpoint.writes (#​4822)
  • Remove Checkpoint.pending_sends (#​4820)
  • Pregel: Add NodeBuilder class to replace Channel.subscribe_to (#​4819)
  • Remove dict subclasses used for values/updates stream chunks (#​4816)
  • Remove old checkpoint test fixtures (#​4814)
  • Remove postgres shallow checkpointer (#​4813)
  • Remove ChannelsManager, ManagedValues are now static classes and never instantiated (#​4812)
  • Remove Context channel / managed value, Remove SharedValue (#​4857)
  • Update managed value usage in local_read
  • Format
  • Remove Checkpoint.writes
  • Remove Checkpoint.pending_sends
  • Pregel: Add NodeBuilder class to replace Channel.subscribe_to
  • Lint
  • Remove old checkpoint test fixtures
  • Remove postgres shallow checkpointer
  • Remove ChannelsManager, ManagedValues are now static classes and never instantiated
  • Remove Context channel / managed value, Remove SharedValue

v0.4.9

Compare Source

Changes since 0.5.0rc1

  • Revert change to default value of checkpoint_during arg (#​5177)
  • Revert change to default value of checkpoint_during arg
  • Fix bug where Command(update=) could be ignored if there was a 2nd interrupt after it
  • Reduce extraneous keys in checkpoint.metadata
  • If FuturesDict callback has been GCed, don't call it

v0.4.8

Compare Source

Changes since 0.4.7

  • Remove unused deprecation decorator/warning (#​4917)
  • Require state_schema in StateGraph.__init__ (#​4897)
  • Remove add_conditional_edge(..., then=) (#​4893)
  • Avoid repeated runtime calls to get_type_hints
  • Remove MessageGraph (#​4875)
  • Remove non-state Graph
  • Remove UntrackedValue channel
  • Remove UntrackedValue channel (#​4859)
  • Flip default for checkpoint_during
  • Remove Channel node builder
  • Update managed value usage in local_read (#​4854)
  • Modify stream mode messages and custom to respect subgraphs=False (#​4843)
  • Remove Checkpoint.writes (#​4822)
  • Remove Checkpoint.pending_sends (#​4820)
  • Pregel: Add NodeBuilder class to replace Channel.subscribe_to (#​4819)
  • Remove dict subclasses used for values/updates stream chunks (#​4816)
  • Remove old checkpoint test fixtures (#​4814)
  • Remove postgres shallow checkpointer (#​4813)
  • Remove ChannelsManager, ManagedValues are now static classes and never instantiated (#​4812)
  • Remove Context channel / managed value, Remove SharedValue (#​4857)
  • Remove SchemaCoercionMapper
  • Update managed value usage in local_read
  • prebuilt: release 0.2.2 (#​4851)
  • Format
  • Remove Checkpoint.writes
  • Remove Checkpoint.pending_sends
  • Pregel: Add NodeBuilder class to replace Channel.subscribe_to
  • Lint
  • Remove old checkpoint test fixtures
  • Remove postgres shallow checkpointer
  • Remove ChannelsManager, ManagedValues are now static classes and never instantiated
  • Remove Context channel / managed value, Remove SharedValue
  • Merge branch 'main' into fix-examples-link-readme
  • Update libs/langgraph/README.md
  • Update libs/langgraph/README.md
  • Merge branch 'main' into fix-examples-link-readme
  • Merge branch 'main' into fix-examples-link-readme
  • Merge branch 'main' into fix-examples-link-readme
  • Merge branch 'langchain-ai:main' into fix-examples-link-readme
  • fix link to examples page in readme.md

v0.4.7

Compare Source

Changes since 0.4.6

  • 0.4.7
  • Fix stream mode not respected in subgraphs
  • Try to make test less flaky
  • Add tests for stream_events when using imperative api

v0.4.6

Compare Source

Changes since 0.4.5

  • 0.4.6
  • Fix
  • Lint
  • Add sync test
  • Fix
  • Fix exception handling for imperative tasks
  • prebuilt: release 0.2.1 (#​4801)
  • prebuilt: release 0.2.0 (#​4793)
  • Code review
  • feat(langgraph): push_messages should directly write to the state
  • feat(graph): add push_message method to push manually to messages / message-tuple stream (#​4722)
  • Add message state test
  • Apply same condition for stream_mode=values in suppress interrupt
  • Remove redundant cast
  • feat(graph): add push_message method to push manually to messages / message-tuple stream
  • Last uv.lock?
  • Only emit stream values chunks when the output channels have changed (#​4774)
  • docs: fix example pregel reducer (#​4740)
  • prebuilts hitl: fix branching logic + add structural snapshot tests (#​4767)
  • docs: deferred nodes (#​4759)
  • Remove local_write utility (#​4751)
  • Print output for cached @​task functions (#​4750)
  • Print output for cached @​task functions
  • sqlite: Add test for search with list filters (#​4747)
  • sqlite: update list_namespaces with max_depth (#​4746)
  • Update lockfile
  • docs: Clean up adjective use in readme (#​4734)
  • SqliteStore (#​3608)
  • ci: fix benchmark command (#​4729)
  • docs: document tuples in streaming guides (#​4690)
  • ci: migrate to uv! (#​4698)

v0.4.5

Compare Source

Changes since 0.4.4

  • langgraph: release 0.4.5 (#​4709)
  • Improve how we match cached writes for async imperative tasks (#​4691)
  • Lint
  • Improve how we match cached writes for async imperative tasks

v0.4.4

Compare Source

Changes since 0.4.3

  • langgraph: release 0.4.4 (#​4704)
  • update for consistency
  • lint again
  • use list
  • lint + update
  • update
  • update
  • update
  • langgraph: fix drawing graph with root channel
  • langgraph: fix graph drawing for self-loops
  • Merge branch 'main' into clean
  • [Proposal] add gitmcp badge for simple LLM-accessible documentation (#​4502)
  • Change LG Cloud > Platform
  • Fix invalid channels error message (#​4357)
  • docs: remove unused annotation comments in readmes (#​4656)
  • docs: General language cleanup (#​4649)
  • [docs] LangGraph / LangGraph Platform docs updates (#​4479)
  • Port checkpointer/store fixtures in langgraph-prebuilt to idiomatic pattern (#​4629)
  • Extend all tests using InMemorySaver to use all available checkpointers (#​4627)
  • Implement update_state for functional api (#​4626)
  • More idiomatic sync/async checkpointer fixtures in pytest (#​4624)
  • More idiomatic sync/async store fixtures in pytest (#​4617)
  • Overload clear method to delete all when called without args
  • Limit depth
  • Extend all tests using InMemorySaver to use all available checkpointers
  • Implement update_state for functional api
  • Fix
  • More idiomatic sync/async checkpointer fixtures in pytest
  • Move to sep file
  • More idiomatic sync/async store fixtures in pytest
  • Lint
  • Lint
  • Move FileCache to sqlite package, add InMemoryCache
  • Add clear cache methods
  • Remove refresh
  • Add namespace to cache keys
  • Lint
  • Lint
  • Lint
  • Lint
  • Lint
  • Lint
  • Lint
  • Finish implementation, add tests
  • Lint
  • Lint
  • Accept default cache_policy for graph/entrypoint/pregel
  • Fix
  • Fix
  • Output cached writes
  • Lint
  • Rename
  • Fix type annotation
  • Fixes
  • Re-do with separate cache interface
  • WIP
  • Deferred Node (#​4269)
  • Start dev server externally (#​4604)

v0.4.3

Compare Source

Changes since 0.4.2

  • release: 0.4.3 (#​4592)
  • langgraph: use tuples for streamed message events in RemoteGraph (#​4589)
  • Fix remote streaming of subgraphs (#​4590)
  • Add a limit to Pregel.draw (#​4575)

v0.4.2

Compare Source

Changes since 0.4.1

  • 0.4.2 (#​4570)
  • update
  • langgraph: decouple name from assistant ID in RemoteGraph
  • prebuilt: switch to executing parallel tool calls via Send by default (#​4438)
  • fix for langgraph
  • docstring fixes for libs/langgraph
  • docs: update add_messages API ref (#​4495)
  • update other tests
pytest-dev/pytest (pytest)

v9.0.3

Compare Source

pytest 9.0.3 (2026-04-07)

Bug fixes

  • #​12444: Fixed pytest.approx which now correctly takes into account ~collections.abc.Mapping keys order to compare them.

  • #​13634: Blocking a conftest.py file 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 UsageError explaining 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

  • #​13388: Clarified documentation for -p vs PYTEST_PLUGINS plugin loading and fixed an incorrect -p example.
  • #​13731: Clarified that capture fixtures (e.g. capsys and capfd) take precedence over the -s / --capture=no command-line options in Accessing captured output from a test function <accessing-captured-output>.
  • #​14088: Clarified that the default pytest_collection hook sets session.items before it calls pytest_collection_finish, not after.
  • #​14255: TOML integer log levels must be quoted: Updating reference documentation.

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 aleguy02

v9.0.2

Compare 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 TERM is dumb, the escape codes are no longer emitted, even if the plugin is enabled.

  • #​13904: Fixed the TOML type of the tmp_path_retention_count settings in the API reference from number to string.

  • #​13946: The private config.inicfg attribute 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 unittest subtests in Python 3.10.

Improved documentation

  • #​4492: The API Reference now contains cross-reference-able documentation of pytest's command-line flags <command-line-flags>.

v9.0.1

Compare Source

pytest 9.0.1 (2025-11-12)

Bug fixes

  • #​13895: Restore support for skipping tests via raise unittest.SkipTest.
  • #​13896: The terminal progress plugin added in pytest 9.0 is now automatically disabled when iTerm2 is detected, it generated desktop notifications instead of the desired functionality.
  • #​13904: Fixed the TOML type of the verbosity settings in the API reference from number to string.
  • #​13910: Fixed UserWarning: Do not expect file_or_dir on some earlier Python 3.12 and 3.13 point versions.

Packaging updates and notes for downstreams

  • #​13933: The tox configuration has been adjusted to make sure the desired
    version string can be passed into its package_env through
    the SETUPTOOLS_SCM_PRETEND_VERSION_FOR_PYTEST environment
    variable as a part of the release process -- by webknjaz.

Contributor-facing changes

  • #​13891, #​13942: The CI/CD part of the release automation is now capable of
    creating GitHub Releases without having a Git checkout on
    disk -- by bluetech and webknjaz.
  • #​13933: The tox configuration has been adjusted to make sure the desired
    version string can be passed into its package_env through
    the SETUPTOOLS_SCM_PRETEND_VERSION_FOR_PYTEST environment
    variable as a part of the release process -- by webknjaz.

v9.0.0

Compare 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:

    def contains_docstring(p: Path) -> bool:
        """Return True if the given Python file contains a top-level docstring."""
        ...
    
    def test_py_files_contain_docstring(subtests: pytest.Subtests) -> None:
        for path in Path.cwd().glob("*.py"):
            with subtests.test(path=str(path)):
                assert contains_docstring(path)
    

    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.subTest is now also supported.

    This feature was originally implemented as a separate plugin in pytest-subtests, but since then has been merged into the core.

    [!NOTE]
    This feature is experimental and will likely evolve in future releases. By that we mean that we might change how subtests are reported on failure, but the functionality and how to use it are stable.

  • #​13743: Added support for native TOML configuration files.

    While pytest, since version 6, supports configuration in pyproject.toml files 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.

    # pyproject.toml
    [tool.pytest]
    minversion = "9.0"
    addopts = ["-ra", "-q"]
    testpaths = [
        "tests",
        "integration",
    ]
    

    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 use pytest.toml or .pytest.toml:

    # pytest.toml or .pytest.toml
    [pytest]
    minversion = "9.0"
    addopts = ["-ra", "-q"]
    testpaths = [
        "tests",
        "integration",
    ]
    

    The documentation now (sometimes) shows configuration snippets in both TOML and INI formats, in a tabbed interface.

    See config file formats for full details.

  • #​13823: Added a "strict mode" enabled by the strict configuration option.

    When set to true, the strict option currently enables

    • strict_config
    • strict_markers
    • strict_parametrization_ids
    • strict_xfail

    The individual strictness options can be explicitly set to override the global strict setting.

    The previously-deprecated --strict command-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 mode for more details.

  • #​13737: Added the strict_parametrization_ids configuration 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:

    • On a user level, using -p no:terminalprogress on the command line or via an environment variable PYTEST_ADDOPTS='-p no:terminalprogress'.
    • On a project configuration level, using addopts = "-p no:terminalprogress".
  • #​478: Support PEP420 (implicit namespace packages) as --pyargs target when consider_namespace_packages is true in the config.

    Previously, this option only impacted package imports, now it also impacts tests discovery.

  • #​13678: Added a new faulthandler_exit_on_timeout configuration 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 aliases parameter in Parser.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.ini file and pyproject.toml with 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 --version no longer loads the entire plugin infrastructure, making it faster and more reliable when displaying only the pytest version.

    Passing --version twice (e.g., pytest --version --version) retains the original behavior, showing both the pytest version and plugin information.

    [!NOTE]
    Since --version is now processed early, it only takes effect when passed directly via the command line. It will not work if set through other mechanisms, such as PYTEST_ADDOPTS or addopts.

  • #​13823: Added strict_xfail as an alias to the xfail_strict option,
    strict_config as an alias to the --strict-config flag,
    and strict_markers as an alias to the --strict-markers flag.
    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:

    1. pytest a/b a/ or pytest a/ a/b are equivalent to pytest a; if an argument overlaps another arguments, only the prefix remains.
    2. pytest x.py x.py is equivalent to pytest x.py; previously such an invocation was taken as an explicit request to run the tests from the file twice.

    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 PytestRemovedIn9Warning now generate errors
    instead of warning messages by default.

    The affected features will be effectively removed in pytest 9.1, so please consult the
    deprecations section in the docs for directions on how to update existing code.

    In the pytest 9.0.X series, it is possible to change the errors back into warnings as a
    stopgap measure by adding this to your pytest.ini file:

    [pytest]
    filterwarnings =
        ignore::pytest.PytestRemovedIn9Warning
    

    But this will stop working when pytest 9.1 is released.

    If you have concerns about the removal of a specific feature, please add a
    comment to 13779.

Deprecations (removal in next major release)

  • #​13807: monkeypatch.syspath_prepend() <pytest.MonkeyPatch.syspath_prepend> now issues a deprecation warning when the prepended path contains legacy namespace packages (those using pkg_resources.declare_namespace()).
    Users should migrate to native namespace packages (420).
    See monkeypatch-fixup-namespace-packages for details.

Bug fixes

  • #​13445: Made the type annotations of pytest.skip and friends more spec-complaint to have them work across more type checkers.

  • #​13537: Fixed a bug in which ExceptionGroup with only Skipped exceptions 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 as pytest x.py. This is now a usage error.

  • #​13722: Fixed a misleading assertion failure message when using pytest.approx on mappings with differing lengths.

  • #​13773: Fixed the static fixture closure calculation to properly consider transitive dependencies requested by overridden fixtures.

  • #​13816: Fixed pytest.approx which now returns a clearer error message when comparing mappings with different keys.

  • #​13849: Hidden .pytest.ini files 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 pytester in subprocess mode ignored all :attr`pytester.plugins <pytest.Pytester.plugins>` except the first.

    Fixed pytester in subprocess mode silently ignored non-str pytester.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

  • #​13791: Minimum requirements on iniconfig and packaging were bumped to 1.0.1 and 22.0.0, respectively.

Contributor-facing changes

  • #​12244: Fixed self-test failures when TERM=dumb.
  • #​12474: Added scheduled GitHub Action Workflow to run Sphinx linkchecks in repo documentation.
  • #​13621: pytest's own testsuite now handles the lsof command hanging (e.g. due to unreachable network filesystems), with the affected selftests being skipped after 10 seconds.
  • #​13638: Fixed deprecated gh pr new command in scripts/prepare-release-pr.py.
    The script now uses gh pr create which is compatible with GitHub CLI v2.0+.
  • #​13695: Flush stdout and stderr in Pytester.run to avoid truncated outputs in test_faulthandler.py::test_timeout on CI -- by ogrisel.
  • #​13771: Skip test_do_not_collect_symlink_siblings on Windows environments without symlink support to avoid false negatives.
  • #​13841: tox>=4 is now required when contributing to pytest.
  • #​13625: Added missing docstrings to pytest_addoption(), pytest_configure(), and cacheshow() functions in cacheprovider.py.

Miscellaneous internal changes

  • #​13830: Configuration overrides (-o/--override-ini) are now processed during startup rather than during config.getini() <pytest.Config.getini>.

Configuration

📅 Schedule: (in timezone Europe/Berlin)

  • Branch creation
    • At any time (no schedule defined)
  • Automerge
    • At any time (no schedule defined)

🚦 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.


  • If you want to rebase/retry this PR, check this box

This PR has been generated by Mend Renovate CLI.

This PR contains the following updates: | Package | Change | [Age](https://docs.renovatebot.com/merge-confidence/) | [Confidence](https://docs.renovatebot.com/merge-confidence/) | |---|---|---|---| | [langchain](https://github.com/langchain-ai/langchain) ([changelog](https://github.com/langchain-ai/langchain/releases?q=tag%3A%22langchain%3D%3D1%22)) | `==0.3.30` → `==1.3.9` | ![age](https://developer.mend.io/api/mc/badges/age/pypi/langchain/1.3.9?slim=true) | ![confidence](https://developer.mend.io/api/mc/badges/confidence/pypi/langchain/0.3.30/1.3.9?slim=true) | | [langchain-core](https://github.com/langchain-ai/langchain) ([changelog](https://github.com/langchain-ai/langchain/releases?q=%22langchain-core%3D%3D1%22)) | `==0.3.86` → `==1.2.22` | ![age](https://developer.mend.io/api/mc/badges/age/pypi/langchain-core/1.2.22?slim=true) | ![confidence](https://developer.mend.io/api/mc/badges/confidence/pypi/langchain-core/0.3.86/1.2.22?slim=true) | | [langchain-text-splitters](https://github.com/langchain-ai/langchain) ([changelog](https://github.com/langchain-ai/langchain/releases?q=%22langchain-text-splitters%22)) | `==0.3.11` → `==1.1.2` | ![age](https://developer.mend.io/api/mc/badges/age/pypi/langchain-text-splitters/1.1.2?slim=true) | ![confidence](https://developer.mend.io/api/mc/badges/confidence/pypi/langchain-text-splitters/0.3.11/1.1.2?slim=true) | | [langgraph](https://github.com/langchain-ai/langgraph) ([source](https://github.com/langchain-ai/langgraph/tree/HEAD/libs/langgraph), [changelog](https://github.com/langchain-ai/langgraph/releases)) | `==0.4.1` → `==1.0.10` | ![age](https://developer.mend.io/api/mc/badges/age/pypi/langgraph/1.0.10?slim=true) | ![confidence](https://developer.mend.io/api/mc/badges/confidence/pypi/langgraph/0.4.1/1.0.10?slim=true) | | [pytest](https://github.com/pytest-dev/pytest) ([changelog](https://docs.pytest.org/en/stable/changelog.html)) | `==8.4.2` → `==9.0.3` | ![age](https://developer.mend.io/api/mc/badges/age/pypi/pytest/9.0.3?slim=true) | ![confidence](https://developer.mend.io/api/mc/badges/confidence/pypi/pytest/8.4.2/9.0.3?slim=true) | --- ### LangChain: Path traversal and sandbox escape in LangChain file-search middleware and loaders [CVE-2026-55443](https://nvd.nist.gov/vuln/detail/CVE-2026-55443) / [GHSA-gr75-jv2w-4656](https://github.com/advisories/GHSA-gr75-jv2w-4656) / PYSEC-2026-2192 / PYSEC-2026-2556 <details> <summary>More information</summary> #### 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 - Confidentiality: disclosure of file contents outside the intended root/sandbox. - Authorization: path-prefix bypass can grant access to sibling resources beyond the intended subtree. ##### 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 Score: 5.1 / 10 (Medium) - Vector String: `CVSS:3.1/AV:L/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N` #### References - [https://github.com/langchain-ai/langchain/security/advisories/GHSA-gr75-jv2w-4656](https://github.com/langchain-ai/langchain/security/advisories/GHSA-gr75-jv2w-4656) - [https://nvd.nist.gov/vuln/detail/CVE-2026-55443](https://nvd.nist.gov/vuln/detail/CVE-2026-55443) - [https://github.com/langchain-ai/langchain/commit/dcaf7795a3e6590af55c3ff7bda6add6355e9ea6](https://github.com/langchain-ai/langchain/commit/dcaf7795a3e6590af55c3ff7bda6add6355e9ea6) - [https://github.com/langchain-ai/langchain](https://github.com/langchain-ai/langchain) This data is provided by [OSV](https://osv.dev/vulnerability/GHSA-gr75-jv2w-4656) and the [GitHub Advisory Database](https://github.com/github/advisory-database) ([CC-BY 4.0](https://github.com/github/advisory-database/blob/main/LICENSE.md)). </details> --- ### [CVE-2026-55443](https://nvd.nist.gov/vuln/detail/CVE-2026-55443) / [GHSA-gr75-jv2w-4656](https://github.com/advisories/GHSA-gr75-jv2w-4656) / PYSEC-2026-2192 / PYSEC-2026-2556 <details> <summary>More information</summary> #### 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 Score: 5.5 / 10 (Medium) - Vector String: `CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N` #### References - [https://github.com/langchain-ai/langchain/security/advisories/GHSA-gr75-jv2w-4656](https://github.com/langchain-ai/langchain/security/advisories/GHSA-gr75-jv2w-4656) - [https://github.com/langchain-ai/langchain/commit/dcaf7795a3e6590af55c3ff7bda6add6355e9ea6](https://github.com/langchain-ai/langchain/commit/dcaf7795a3e6590af55c3ff7bda6add6355e9ea6) This data is provided by [OSV](https://osv.dev/vulnerability/PYSEC-2026-2192) and the [PyPI Advisory Database](https://github.com/pypa/advisory-database) ([CC-BY 4.0](https://github.com/pypa/advisory-database/blob/main/LICENSE)). </details> --- ### LangChain affected by SSRF via image_url token counting in ChatOpenAI.get_num_tokens_from_messages [CVE-2026-26013](https://nvd.nist.gov/vuln/detail/CVE-2026-26013) / [GHSA-2g6r-c272-w58r](https://github.com/advisories/GHSA-2g6r-c272-w58r) / PYSEC-2026-2562 <details> <summary>More information</summary> #### Details ##### Server-Side Request Forgery (SSRF) in ChatOpenAI Image Token Counting ##### Summary The `ChatOpenAI.get_num_tokens_from_messages()` method fetches arbitrary `image_url` values 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: - Responses are not returned to the attacker (blind SSRF) - Default 5-second timeout limits resource exhaustion - Non-image responses fail at PIL image parsing ##### Impact An attacker who can control image URLs passed to `get_num_tokens_from_messages()` can: - Trigger HTTP requests from the application server to arbitrary internal or external URLs - Cause the server to access internal network resources (private IPs, cloud metadata endpoints) - Cause minor resource consumption through image downloads (bounded by timeout) **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: 1. `get_num_tokens_from_messages()` processes messages containing `image_url` content blocks 2. For images without `detail: "low"`, it calls `_url_to_size()` to fetch the image and compute token counts 3. `_url_to_size()` performs `httpx.get(image_source)` on any URL without validation 4. Prior to the patch, there was no SSRF protection, size limits, or explicit timeout **File:** `libs/partners/openai/langchain_openai/chat_models/base.py` ##### Patches The vulnerability has been patched in `langchain-openai==1.1.9` (requires `langchain-core==1.2.11`). The patch adds: 1. **SSRF validation** using `langchain_core._security._ssrf_protection.validate_safe_url()` to block: - Private IP ranges (RFC 1918, loopback, link-local) - Cloud metadata endpoints (169.254.169.254, etc.) - Invalid URL schemes 2. **Explicit size limits** (50 MB maximum, matching OpenAI's payload limit) 3. **Explicit timeout** (5 seconds, same as `httpx.get` default) 4. **Allow disabling image fetching** via `allow_fetching_images=False` parameter ##### Workarounds If you cannot upgrade immediately: 1. **Sanitize input:** Validate and filter `image_url` values before passing messages to token counting or model invocation 2. **Use network controls:** Implement egress filtering to prevent outbound requests to private IPs #### Severity - CVSS Score: 3.7 / 10 (Low) - Vector String: `CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:L` #### References - [https://github.com/langchain-ai/langchain/security/advisories/GHSA-2g6r-c272-w58r](https://github.com/langchain-ai/langchain/security/advisories/GHSA-2g6r-c272-w58r) - [https://nvd.nist.gov/vuln/detail/CVE-2026-26013](https://nvd.nist.gov/vuln/detail/CVE-2026-26013) - [https://github.com/langchain-ai/langchain/commit/2b4b1dc29a833d4053deba4c2b77a3848c834565](https://github.com/langchain-ai/langchain/commit/2b4b1dc29a833d4053deba4c2b77a3848c834565) - [https://github.com/langchain-ai/langchain](https://github.com/langchain-ai/langchain) - [https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.2.11](https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.2.11) This data is provided by [OSV](https://osv.dev/vulnerability/GHSA-2g6r-c272-w58r) and the [GitHub Advisory Database](https://github.com/github/advisory-database) ([CC-BY 4.0](https://github.com/github/advisory-database/blob/main/LICENSE.md)). </details> --- ### LangChain affected by SSRF via image_url token counting in ChatOpenAI.get_num_tokens_from_messages [CVE-2026-26013](https://nvd.nist.gov/vuln/detail/CVE-2026-26013) / [GHSA-2g6r-c272-w58r](https://github.com/advisories/GHSA-2g6r-c272-w58r) / PYSEC-2026-2562 <details> <summary>More information</summary> #### Details ##### Server-Side Request Forgery (SSRF) in ChatOpenAI Image Token Counting ##### Summary The `ChatOpenAI.get_num_tokens_from_messages()` method fetches arbitrary `image_url` values 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: - Responses are not returned to the attacker (blind SSRF) - Default 5-second timeout limits resource exhaustion - Non-image responses fail at PIL image parsing ##### Impact An attacker who can control image URLs passed to `get_num_tokens_from_messages()` can: - Trigger HTTP requests from the application server to arbitrary internal or external URLs - Cause the server to access internal network resources (private IPs, cloud metadata endpoints) - Cause minor resource consumption through image downloads (bounded by timeout) **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: 1. `get_num_tokens_from_messages()` processes messages containing `image_url` content blocks 2. For images without `detail: "low"`, it calls `_url_to_size()` to fetch the image and compute token counts 3. `_url_to_size()` performs `httpx.get(image_source)` on any URL without validation 4. Prior to the patch, there was no SSRF protection, size limits, or explicit timeout **File:** `libs/partners/openai/langchain_openai/chat_models/base.py` ##### Patches The vulnerability has been patched in `langchain-openai==1.1.9` (requires `langchain-core==1.2.11`). The patch adds: 1. **SSRF validation** using `langchain_core._security._ssrf_protection.validate_safe_url()` to block: - Private IP ranges (RFC 1918, loopback, link-local) - Cloud metadata endpoints (169.254.169.254, etc.) - Invalid URL schemes 2. **Explicit size limits** (50 MB maximum, matching OpenAI's payload limit) 3. **Explicit timeout** (5 seconds, same as `httpx.get` default) 4. **Allow disabling image fetching** via `allow_fetching_images=False` parameter ##### Workarounds If you cannot upgrade immediately: 1. **Sanitize input:** Validate and filter `image_url` values before passing messages to token counting or model invocation 2. **Use network controls:** Implement egress filtering to prevent outbound requests to private IPs #### Severity - CVSS Score: 3.7 / 10 (Low) - Vector String: `CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:L` #### References - [https://github.com/langchain-ai/langchain/security/advisories/GHSA-2g6r-c272-w58r](https://github.com/langchain-ai/langchain/security/advisories/GHSA-2g6r-c272-w58r) - [https://nvd.nist.gov/vuln/detail/CVE-2026-26013](https://nvd.nist.gov/vuln/detail/CVE-2026-26013) - [https://github.com/langchain-ai/langchain/commit/2b4b1dc29a833d4053deba4c2b77a3848c834565](https://github.com/langchain-ai/langchain/commit/2b4b1dc29a833d4053deba4c2b77a3848c834565) - [https://github.com/langchain-ai/langchain](https://github.com/langchain-ai/langchain) - [https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.2.11](https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.2.11) - [https://pypi.org/project/langchain-core](https://pypi.org/project/langchain-core) - [https://github.com/advisories/GHSA-2g6r-c272-w58r](https://github.com/advisories/GHSA-2g6r-c272-w58r) This data is provided by [OSV](https://osv.dev/vulnerability/PYSEC-2026-2562) and the [PyPI Advisory Database](https://github.com/pypa/advisory-database) ([CC-BY 4.0](https://github.com/pypa/advisory-database/blob/main/LICENSE)). </details> --- ### LangChain Core has Path Traversal vulnerabilites in legacy `load_prompt` functions [CVE-2026-34070](https://nvd.nist.gov/vuln/detail/CVE-2026-34070) / [GHSA-qh6h-p6c9-ff54](https://github.com/advisories/GHSA-qh6h-p6c9-ff54) / PYSEC-2026-2193 <details> <summary>More information</summary> #### Details ##### Summary 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). **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 the `dumpd`/`dumps`/`load`/`loads` serialization APIs in `langchain_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-core` **File:** `langchain_core/prompts/loading.py` **Affected 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 | Config key | Loaded by | Readable extensions | |---|---|---| | `template_path`, `suffix_path`, `prefix_path` | `_load_template()` | `.txt` | | `examples` (when string) | `_load_examples()` | `.json`, `.yaml`, `.yml` | | `example_prompt_path` | `_load_few_shot_prompt()` | `.json`, `.yaml`, `.yml` | None 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: - **`.txt` files:** cloud-mounted secrets (`/mnt/secrets/api_key.txt`), `requirements.txt`, internal system prompts - **`.json`/`.yaml` files:** cloud credentials (`~/.docker/config.json`, `~/.azure/accessTokens.json`), Kubernetes manifests, CI/CD configs, application settings This 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 ```python from langchain_core.prompts.loading import load_prompt_from_config ##### Reads /tmp/secret.txt via absolute path injection config = { "_type": "prompt", "template_path": "/tmp/secret.txt", "input_variables": [], } prompt = load_prompt_from_config(config) print(prompt.template) # file contents disclosed ##### Reads ../../etc/secret.txt via directory traversal config = { "_type": "prompt", "template_path": "../../etc/secret.txt", "input_variables": [], } prompt = load_prompt_from_config(config) ##### Reads arbitrary .json via few-shot examples config = { "_type": "few_shot", "examples": "../../../../.docker/config.json", "example_prompt": { "_type": "prompt", "input_variables": ["input", "output"], "template": "{input}: {output}", }, "prefix": "", "suffix": "{query}", "input_variables": ["query"], } prompt = load_prompt_from_config(config) ``` ##### Mitigation **Update `langchain-core` to >= 1.2.22.** The fix adds path validation that rejects absolute paths and `..` traversal sequences by default. An `allow_dangerous_paths=True` keyword argument is available on `load_prompt()` and `load_prompt_from_config()` for trusted inputs. As described above, these legacy APIs have been formally deprecated. Users should migrate to `dumpd`/`dumps`/`load`/`loads` from `langchain_core.load`. ##### Credit - [jiayuqi7813](https://github.com/jiayuqi7813) reporter - [VladimirEliTokarev](https://github.com/VladimirEliTokarev) reporter - [Rickidevs](https://github.com/Rickidevs) reporter - Kenneth Cox (cczine@gmail.com) reporter #### Severity - CVSS Score: 7.5 / 10 (High) - Vector String: `CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N` #### References - [https://github.com/langchain-ai/langchain/security/advisories/GHSA-qh6h-p6c9-ff54](https://github.com/langchain-ai/langchain/security/advisories/GHSA-qh6h-p6c9-ff54) - [https://nvd.nist.gov/vuln/detail/CVE-2026-34070](https://nvd.nist.gov/vuln/detail/CVE-2026-34070) - [https://github.com/langchain-ai/langchain/commit/27add913474e01e33bededf4096151130ba0d47c](https://github.com/langchain-ai/langchain/commit/27add913474e01e33bededf4096151130ba0d47c) - [https://github.com/langchain-ai/langchain](https://github.com/langchain-ai/langchain) - [https://github.com/langchain-ai/langchain/releases/tag/langchain-core==1.2.22](https://github.com/langchain-ai/langchain/releases/tag/langchain-core==1.2.22) This data is provided by [OSV](https://osv.dev/vulnerability/GHSA-qh6h-p6c9-ff54) and the [GitHub Advisory Database](https://github.com/github/advisory-database) ([CC-BY 4.0](https://github.com/github/advisory-database/blob/main/LICENSE.md)). </details> --- ### [CVE-2026-34070](https://nvd.nist.gov/vuln/detail/CVE-2026-34070) / [GHSA-qh6h-p6c9-ff54](https://github.com/advisories/GHSA-qh6h-p6c9-ff54) / PYSEC-2026-2193 <details> <summary>More information</summary> #### 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 Score: 7.5 / 10 (High) - Vector String: `CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N` #### References - [https://access.redhat.com/security/cve/CVE-2026-34070](https://access.redhat.com/security/cve/CVE-2026-34070) - [https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-34070.json](https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-34070.json) - [https://access.redhat.com/errata/RHSA-2026:24766](https://access.redhat.com/errata/RHSA-2026:24766) - [https://access.redhat.com/errata/RHSA-2026:37275](https://access.redhat.com/errata/RHSA-2026:37275) - [https://github.com/langchain-ai/langchain/releases/tag/langchain-core==1.2.22](https://github.com/langchain-ai/langchain/releases/tag/langchain-core==1.2.22) - [https://bugzilla.redhat.com/show_bug.cgi?id=2453287](https://bugzilla.redhat.com/show_bug.cgi?id=2453287) - [https://github.com/langchain-ai/langchain/commit/27add913474e01e33bededf4096151130ba0d47c](https://github.com/langchain-ai/langchain/commit/27add913474e01e33bededf4096151130ba0d47c) - [https://github.com/langchain-ai/langchain/security/advisories/GHSA-qh6h-p6c9-ff54](https://github.com/langchain-ai/langchain/security/advisories/GHSA-qh6h-p6c9-ff54) This data is provided by [OSV](https://osv.dev/vulnerability/PYSEC-2026-2193) and the [PyPI Advisory Database](https://github.com/pypa/advisory-database) ([CC-BY 4.0](https://github.com/pypa/advisory-database/blob/main/LICENSE)). </details> --- ### LangChain Text Splitters: HTMLHeaderTextSplitter.split_text_from_url SSRF Redirect Bypass [CVE-2026-41481](https://nvd.nist.gov/vuln/detail/CVE-2026-41481) / [GHSA-fv5p-p927-qmxr](https://github.com/advisories/GHSA-fv5p-p927-qmxr) / PYSEC-2026-77 <details> <summary>More information</summary> #### Details ##### Summary `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. ##### Affected versions - `langchain-text-splitters` < 1.1.2 ##### Patched versions - `langchain-text-splitters` >= 1.1.2 (requires `langchain-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: ```python validate_safe_url(url, allow_private=False, allow_http=True) response = requests.get(url, timeout=timeout, **kwargs) ``` ##### Attack scenario 1. A developer passes external URLs to `split_text_from_url()`, relying on its built-in `validate_safe_url()` check to block requests to internal networks. 2. An attacker supplies a URL pointing to a public host they control. The URL passes `validate_safe_url()` (public hostname, public IP). 3. The attacker's server responds with a `302` redirect 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). 4. `requests.get()` follows the redirect automatically. The redirect target is **not** revalidated. 5. The response body is parsed and returned as `Document` objects to the application. **Notes:** - The core issue is a bypass of an explicitly provided SSRF protection. `split_text_from_url()` included `validate_safe_url()` specifically to be safe with untrusted URLs — the redirect loophole defeated that guarantee. - Cloud metadata endpoints that require special headers (AWS IMDSv2, GCP, Azure) are not reachable through this bug because the attacker does not control request headers. AWS IMDSv1, which requires no headers, is reachable. - Data exfiltration requires the application to return Document contents to the 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 (`SSRFSafeSyncTransport` from `langchain-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 to `split_text()` directly. #### Severity - CVSS Score: 6.5 / 10 (Medium) - Vector String: `CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N` #### References - [https://github.com/langchain-ai/langchain/security/advisories/GHSA-fv5p-p927-qmxr](https://github.com/langchain-ai/langchain/security/advisories/GHSA-fv5p-p927-qmxr) - [https://nvd.nist.gov/vuln/detail/CVE-2026-41481](https://nvd.nist.gov/vuln/detail/CVE-2026-41481) - [https://github.com/langchain-ai/langchain](https://github.com/langchain-ai/langchain) - [https://github.com/pypa/advisory-database/tree/main/vulns/langchain-text-splitters/PYSEC-2026-77.yaml](https://github.com/pypa/advisory-database/tree/main/vulns/langchain-text-splitters/PYSEC-2026-77.yaml) This data is provided by [OSV](https://osv.dev/vulnerability/GHSA-fv5p-p927-qmxr) and the [GitHub Advisory Database](https://github.com/github/advisory-database) ([CC-BY 4.0](https://github.com/github/advisory-database/blob/main/LICENSE.md)). </details> --- ### [CVE-2026-41481](https://nvd.nist.gov/vuln/detail/CVE-2026-41481) / [GHSA-fv5p-p927-qmxr](https://github.com/advisories/GHSA-fv5p-p927-qmxr) / PYSEC-2026-77 <details> <summary>More information</summary> #### 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 Score: 6.5 / 10 (Medium) - Vector String: `CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N` #### References - [https://github.com/langchain-ai/langchain/security/advisories/GHSA-fv5p-p927-qmxr](https://github.com/langchain-ai/langchain/security/advisories/GHSA-fv5p-p927-qmxr) This data is provided by [OSV](https://osv.dev/vulnerability/PYSEC-2026-77) and the [PyPI Advisory Database](https://github.com/pypa/advisory-database) ([CC-BY 4.0](https://github.com/pypa/advisory-database/blob/main/LICENSE)). </details> --- ### [CVE-2026-28277](https://nvd.nist.gov/vuln/detail/CVE-2026-28277) / [GHSA-g48c-2wqr-h844](https://github.com/advisories/GHSA-g48c-2wqr-h844) / PYSEC-2026-83 <details> <summary>More information</summary> #### 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 Score: 7.2 / 10 (High) - Vector String: `CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H` #### References - [https://github.com/langchain-ai/langgraph/security/advisories/GHSA-g48c-2wqr-h844](https://github.com/langchain-ai/langgraph/security/advisories/GHSA-g48c-2wqr-h844) This data is provided by [OSV](https://osv.dev/vulnerability/PYSEC-2026-83) and the [PyPI Advisory Database](https://github.com/pypa/advisory-database) ([CC-BY 4.0](https://github.com/pypa/advisory-database/blob/main/LICENSE)). </details> --- ### LangGraph checkpoint loading has unsafe msgpack deserialization [CVE-2026-28277](https://nvd.nist.gov/vuln/detail/CVE-2026-28277) / [GHSA-g48c-2wqr-h844](https://github.com/advisories/GHSA-g48c-2wqr-h844) / PYSEC-2026-83 <details> <summary>More information</summary> #### 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: - use a persistent checkpointer (database, remote store, shared filesystem, etc.), - load/resume from checkpoints, and - operate in an environment where an attacker could gain privileged write access to checkpoint data in the backing store. 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 - Potential **arbitrary code execution** or other unsafe side effects during checkpoint deserialization. - Escalation from “write access to checkpoint store” to “code execution in the application runtime,” which may expose runtime secrets or provide access to other systems the runtime can reach. ##### Exploitation scenario (high level) 1. Attacker gains privileged write access to the checkpoint store (for example, via database compromise, leaked credentials, or abuse of an administrative data path). 2. Attacker writes a crafted checkpoint payload containing msgpack data intended to reconstruct dangerous objects. 3. Application resumes and deserializes the checkpoint; unsafe reconstruction could execute attacker-controlled behavior. ##### Mitigation / remediation LangGraph provides an allowlist-based hardening mechanism for msgpack checkpoint deserialization. ##### Strict mode (environment variable) - **`LANGGRAPH_STRICT_MSGPACK`** - When set truthy (`1`, `true`, `yes`), the default msgpack deserialization policy becomes strict. - Concretely: `JsonPlusSerializer()` will default `allowed_msgpack_modules` to `None` (strict) instead of `True` (warn-and-allow), unless `allowed_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 `datetime` types, `uuid.UUID`, `decimal.Decimal`, `set`/`frozenset`/`deque`, `ipaddress` types, `pathlib` paths, `zoneinfo.ZoneInfo`, compiled regex patterns, and selected LangGraph internal types). ##### Automatically derived allowlist (only when compiling graphs) When `LANGGRAPH_STRICT_MSGPACK` is enabled and `StateGraph` is compiled, LangGraph derives an allowlist from the graph’s schemas and channels and applies it to the checkpointer. - The allowlist is built by walking the state/input/output/context schemas (plus node/branch input schemas) and channel value/update types. It includes Pydantic v1/v2 models, dataclasses, enums, TypedDict field types, and common typing constructs (containers, unions, `Annotated`). - LangGraph also includes a curated set of common LangChain message classes. 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 configure `allowed_msgpack_modules` themselves. ##### Operational guidance - Treat checkpoint stores as integrity-sensitive. Restrict write access and rotate credentials if compromise is suspected. - Enable strict mode (`LANGGRAPH_STRICT_MSGPACK=true`) in production if feasible, and rely on schema-driven allowlisting to reduce incompatibilities. - Avoid providing custom msgpack deserialization hooks that reconstruct arbitrary types unless checkpoint data is fully trusted. ##### Limitations / important notes - If a checkpointer implementation does **not** support allowlist application (i.e., does not implement `with_allowlist`), allowlist enforcement may be skipped (with a warning). In that situation, strict expectations may not hold. - If an application supplies a custom msgpack unpack hook (`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 Score: 6.8 / 10 (Medium) - Vector String: `CVSS:3.1/AV:A/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H` #### References - [https://github.com/langchain-ai/langgraph/security/advisories/GHSA-g48c-2wqr-h844](https://github.com/langchain-ai/langgraph/security/advisories/GHSA-g48c-2wqr-h844) - [https://nvd.nist.gov/vuln/detail/CVE-2026-28277](https://nvd.nist.gov/vuln/detail/CVE-2026-28277) - [https://github.com/langchain-ai/langgraph](https://github.com/langchain-ai/langgraph) - [https://github.com/pypa/advisory-database/tree/main/vulns/langgraph/PYSEC-2026-83.yaml](https://github.com/pypa/advisory-database/tree/main/vulns/langgraph/PYSEC-2026-83.yaml) This data is provided by [OSV](https://osv.dev/vulnerability/GHSA-g48c-2wqr-h844) and the [GitHub Advisory Database](https://github.com/github/advisory-database) ([CC-BY 4.0](https://github.com/github/advisory-database/blob/main/LICENSE.md)). </details> --- ### pytest has vulnerable tmpdir handling [CVE-2025-71176](https://nvd.nist.gov/vuln/detail/CVE-2025-71176) / [GHSA-6w46-j5rx-g56g](https://github.com/advisories/GHSA-6w46-j5rx-g56g) / PYSEC-2026-1845 <details> <summary>More information</summary> #### 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 Score: 6.8 / 10 (Medium) - Vector String: `CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:L/I:L/A:L` #### References - [https://nvd.nist.gov/vuln/detail/CVE-2025-71176](https://nvd.nist.gov/vuln/detail/CVE-2025-71176) - [https://github.com/pytest-dev/pytest/issues/13669](https://github.com/pytest-dev/pytest/issues/13669) - [https://github.com/pytest-dev/pytest/pull/14343](https://github.com/pytest-dev/pytest/pull/14343) - [https://github.com/pytest-dev/pytest/commit/95d8423bd24992deea5b9df32555fa1741679e2c](https://github.com/pytest-dev/pytest/commit/95d8423bd24992deea5b9df32555fa1741679e2c) - [https://github.com/pytest-dev/pytes](https://github.com/pytest-dev/pytes) - [https://github.com/pytest-dev/pytest/releases/tag/9.0.3](https://github.com/pytest-dev/pytest/releases/tag/9.0.3) - [https://www.openwall.com/lists/oss-security/2026/01/21/5](https://www.openwall.com/lists/oss-security/2026/01/21/5) This data is provided by [OSV](https://osv.dev/vulnerability/GHSA-6w46-j5rx-g56g) and the [GitHub Advisory Database](https://github.com/github/advisory-database) ([CC-BY 4.0](https://github.com/github/advisory-database/blob/main/LICENSE.md)). </details> --- ### pytest has vulnerable tmpdir handling [CVE-2025-71176](https://nvd.nist.gov/vuln/detail/CVE-2025-71176) / [GHSA-6w46-j5rx-g56g](https://github.com/advisories/GHSA-6w46-j5rx-g56g) / PYSEC-2026-1845 <details> <summary>More information</summary> #### 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 Score: 6.8 / 10 (Medium) - Vector String: `CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:C/C:L/I:L/A:L` #### References - [https://nvd.nist.gov/vuln/detail/CVE-2025-71176](https://nvd.nist.gov/vuln/detail/CVE-2025-71176) - [https://github.com/pytest-dev/pytest/issues/13669](https://github.com/pytest-dev/pytest/issues/13669) - [https://github.com/pytest-dev/pytest/pull/14343](https://github.com/pytest-dev/pytest/pull/14343) - [https://github.com/pytest-dev/pytest/commit/95d8423bd24992deea5b9df32555fa1741679e2c](https://github.com/pytest-dev/pytest/commit/95d8423bd24992deea5b9df32555fa1741679e2c) - [https://github.com/pytest-dev/pytes](https://github.com/pytest-dev/pytes) - [https://github.com/pytest-dev/pytest/releases/tag/9.0.3](https://github.com/pytest-dev/pytest/releases/tag/9.0.3) - [https://www.openwall.com/lists/oss-security/2026/01/21/5](https://www.openwall.com/lists/oss-security/2026/01/21/5) - [https://pypi.org/project/pytest](https://pypi.org/project/pytest) - [https://github.com/advisories/GHSA-6w46-j5rx-g56g](https://github.com/advisories/GHSA-6w46-j5rx-g56g) This data is provided by [OSV](https://osv.dev/vulnerability/PYSEC-2026-1845) and the [PyPI Advisory Database](https://github.com/pypa/advisory-database) ([CC-BY 4.0](https://github.com/pypa/advisory-database/blob/main/LICENSE)). </details> --- ### Release Notes <details> <summary>langchain-ai/langgraph (langgraph)</summary> ### [`v1.0.10`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.10): langgraph==1.0.10 [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.9...1.0.10) Changes since 1.0.10rc1 - release(langgraph) 1.0.10 ([#&#8203;6967](https://github.com/langchain-ai/langgraph/issues/6967)) - release(checkpoint): 0.4.1 ([#&#8203;6966](https://github.com/langchain-ai/langgraph/issues/6966)) - chore: add serde events ([#&#8203;6954](https://github.com/langchain-ai/langgraph/issues/6954)) - chore: update defaults ([#&#8203;6953](https://github.com/langchain-ai/langgraph/issues/6953)) - release: rc2 ([#&#8203;6949](https://github.com/langchain-ai/langgraph/issues/6949)) - chore(deps-dev): bump the all-dependencies group across 1 directory with 3 updates ([#&#8203;6946](https://github.com/langchain-ai/langgraph/issues/6946)) ### [`v1.0.9`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.9): langgraph==1.0.9 [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.8...1.0.9) Changes since 1.0.8 - release: langgraph + prebuilt ([#&#8203;6875](https://github.com/langchain-ai/langgraph/issues/6875)) - fix: sequential interrupt handling w/ functional API ([#&#8203;6863](https://github.com/langchain-ai/langgraph/issues/6863)) - chore: state_updated_at sort by ([#&#8203;6857](https://github.com/langchain-ai/langgraph/issues/6857)) - chore: bump orjson ([#&#8203;6852](https://github.com/langchain-ai/langgraph/issues/6852)) - chore: conformance testing ([#&#8203;6842](https://github.com/langchain-ai/langgraph/issues/6842)) - chore(deps): bump the all-dependencies group in /libs/langgraph with 6 updates ([#&#8203;6815](https://github.com/langchain-ai/langgraph/issues/6815)) - chore(deps): bump protobuf from 6.33.4 to 6.33.5 in /libs/langgraph ([#&#8203;6833](https://github.com/langchain-ai/langgraph/issues/6833)) - chore(deps): bump cryptography from 46.0.3 to 46.0.5 in /libs/langgraph ([#&#8203;6837](https://github.com/langchain-ai/langgraph/issues/6837)) - chore(deps): bump nbconvert from 7.16.6 to 7.17.0 in /libs/langgraph ([#&#8203;6832](https://github.com/langchain-ai/langgraph/issues/6832)) - chore: server runtime type ([#&#8203;6774](https://github.com/langchain-ai/langgraph/issues/6774)) - refactor: replace bare except with BaseException in AsyncQueue ([#&#8203;6765](https://github.com/langchain-ai/langgraph/issues/6765)) ### [`v1.0.8`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.8): langgraph==1.0.8 [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.7...1.0.8) Changes since 1.0.7 - release(langgraph): 1.0.8 ([#&#8203;6757](https://github.com/langchain-ai/langgraph/issues/6757)) - chore: shallow copy futures ([#&#8203;6755](https://github.com/langchain-ai/langgraph/issues/6755)) - fix: pydantic messages double streaming ([#&#8203;6753](https://github.com/langchain-ai/langgraph/issues/6753)) - chore(deps-dev): bump ruff from 0.14.7 to 0.14.11 in /libs/sdk-py ([#&#8203;6673](https://github.com/langchain-ai/langgraph/issues/6673)) - chore: Omit lock when using connection pool ([#&#8203;6734](https://github.com/langchain-ai/langgraph/issues/6734)) - docs: enhance `Runtime` and `ToolRuntime` class descriptions for clarity ([#&#8203;6689](https://github.com/langchain-ai/langgraph/issues/6689)) - docs: add clarity to use of `thread_id` ([#&#8203;6515](https://github.com/langchain-ai/langgraph/issues/6515)) - docs: add docstrings to `add_node` overloads ([#&#8203;6514](https://github.com/langchain-ai/langgraph/issues/6514)) - docs: update notebook links and add archival notices for examples ([#&#8203;6720](https://github.com/langchain-ai/langgraph/issues/6720)) - release(cli): 0.4.12 ([#&#8203;6716](https://github.com/langchain-ai/langgraph/issues/6716)) ### [`v1.0.7`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.7): langgraph==1.0.7 [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.6...1.0.7) Changes since 1.0.6 - release: langgraph and prebuilt 1.0.7 ([#&#8203;6712](https://github.com/langchain-ai/langgraph/issues/6712)) - fix: aiosqlite's breaking change ([#&#8203;6699](https://github.com/langchain-ai/langgraph/issues/6699)) - chore(deps): upgrade dependencies with `uv lock --upgrade` ([#&#8203;6671](https://github.com/langchain-ai/langgraph/issues/6671)) - chore: update twitter URLs ([#&#8203;6683](https://github.com/langchain-ai/langgraph/issues/6683)) ### [`v1.0.6`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.6): langgraph==1.0.6 [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.5...1.0.6) Changes since 1.0.5 - fix: flip default on base cache ([#&#8203;6677](https://github.com/langchain-ai/langgraph/issues/6677)) - fix: change default recursion limit ([#&#8203;6676](https://github.com/langchain-ai/langgraph/issues/6676)) - fix: sanitize namespace for deeply nested graph jumps ([#&#8203;6665](https://github.com/langchain-ai/langgraph/issues/6665)) - fix: typo: saved the world "BaseMessge" to "BaseMessage" ([#&#8203;6639](https://github.com/langchain-ai/langgraph/issues/6639)) - release(cli): relax api bounds ([#&#8203;6606](https://github.com/langchain-ai/langgraph/issues/6606)) - feat(checkpoint): Validate checkpointer type at compile time ([#&#8203;6586](https://github.com/langchain-ai/langgraph/issues/6586)) ### [`v1.0.5`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.5): langgraph==1.0.5 [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.4...1.0.5) Changes since 1.0.4 - release(langgraph): bump to 1.0.5 ([#&#8203;6582](https://github.com/langchain-ai/langgraph/issues/6582)) - feat(sdk-py): emit id as part of stream events ([#&#8203;6581](https://github.com/langchain-ai/langgraph/issues/6581)) - fix: update readme ([#&#8203;6570](https://github.com/langchain-ai/langgraph/issues/6570)) - release(checkpoint-postgres): 3.0.1 ([#&#8203;6568](https://github.com/langchain-ai/langgraph/issues/6568)) - release(checkpoint-sqlite): 3.0.1 ([#&#8203;6566](https://github.com/langchain-ai/langgraph/issues/6566)) - chore(cli): Pass through webhook configuration in dev server ([#&#8203;6557](https://github.com/langchain-ai/langgraph/issues/6557)) - feat: custom encryption at rest ([#&#8203;6482](https://github.com/langchain-ai/langgraph/issues/6482)) - chore: fix links for docs ([#&#8203;6538](https://github.com/langchain-ai/langgraph/issues/6538)) - chore: Bump lockfile ([#&#8203;6537](https://github.com/langchain-ai/langgraph/issues/6537)) - feat: Include pagination in assistants search response ([#&#8203;6526](https://github.com/langchain-ai/langgraph/issues/6526)) ### [`v1.0.4`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.4) [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.3...1.0.4) Changes since 1.0.3 - release(langgraph): 1.0.4 ([#&#8203;6502](https://github.com/langchain-ai/langgraph/issues/6502)) - chore: pop thread ID from configurable fields in remote graph ([#&#8203;6497](https://github.com/langchain-ai/langgraph/issues/6497)) - chore: clean up some refs ([#&#8203;6487](https://github.com/langchain-ai/langgraph/issues/6487)) - chore(sdk-py): Add more type checking. ([#&#8203;6479](https://github.com/langchain-ai/langgraph/issues/6479)) - fix: interrupt stream mode values ([#&#8203;6475](https://github.com/langchain-ai/langgraph/issues/6475)) - release: langgraph-prebuilt 1.0.5 ([#&#8203;6473](https://github.com/langchain-ai/langgraph/issues/6473)) - refactor: separate prepare_push\_\* functions ([#&#8203;6450](https://github.com/langchain-ai/langgraph/issues/6450)) - fix: dep warnings in prebuilt ([#&#8203;6443](https://github.com/langchain-ai/langgraph/issues/6443)) - release: prebuilt 1.0.3 ([#&#8203;6441](https://github.com/langchain-ai/langgraph/issues/6441)) ### [`v1.0.3`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.3) [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.2...1.0.3) Changes since 1.0.2 - chore: langgraph patch release ([#&#8203;6429](https://github.com/langchain-ai/langgraph/issues/6429)) - feat(docs): warn that `StateGraph` is a builder class ([#&#8203;6417](https://github.com/langchain-ai/langgraph/issues/6417)) - fix(docs): `PartialState` rendering in MkDocs ([#&#8203;6416](https://github.com/langchain-ai/langgraph/issues/6416)) - fix(docs): synchronize `invoke` and `ainvoke` docstrings ([#&#8203;6415](https://github.com/langchain-ai/langgraph/issues/6415)) - fix(docs): synchronize `stream` and `astream` docstrings ([#&#8203;6414](https://github.com/langchain-ai/langgraph/issues/6414)) - docs(langgraph): Fix docstring code examples of task function ([#&#8203;6410](https://github.com/langchain-ai/langgraph/issues/6410)) - fix(langgraph): Unexpected behavior for stream_mode sequences that are not lists ([#&#8203;6354](https://github.com/langchain-ai/langgraph/issues/6354)) - fix: fix previoius edge cases such as 0 ([#&#8203;6379](https://github.com/langchain-ai/langgraph/issues/6379)) - fix(langgraph): export REMOVE_ALL_MESSAGES in **all** to fix linting ([#&#8203;6375](https://github.com/langchain-ai/langgraph/issues/6375)) - style: update docstrings to reference `StateGraph` ([#&#8203;6308](https://github.com/langchain-ai/langgraph/issues/6308)) - docs(langgraph): update streaming guide links ([#&#8203;6314](https://github.com/langchain-ai/langgraph/issues/6314)) - chore: add `pyproject.toml` links ([#&#8203;6364](https://github.com/langchain-ai/langgraph/issues/6364)) - docs(langgraph): Fix typo in docstring of PregelLoop.tick ([#&#8203;6407](https://github.com/langchain-ai/langgraph/issues/6407)) - chore(checkpoint-postgres): bump to 3.0.1 ([#&#8203;6402](https://github.com/langchain-ai/langgraph/issues/6402)) - fix(langgraph): do not apply pending writes when updating state ([#&#8203;6389](https://github.com/langchain-ai/langgraph/issues/6389)) - chore: update ormsgpack minbound and add OPT_REPLACE_SURROGATES ([#&#8203;6395](https://github.com/langchain-ai/langgraph/issues/6395)) - release(cli): 0.4.7 expand api bounds ([#&#8203;6390](https://github.com/langchain-ai/langgraph/issues/6390)) - chore: style fixes for refs ([#&#8203;6365](https://github.com/langchain-ai/langgraph/issues/6365)) ### [`v1.0.2`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.2) [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.1...1.0.2) Changes since 0.6.11 - chore: bump prebuilt dep for lg ([#&#8203;6361](https://github.com/langchain-ai/langgraph/issues/6361)) - chore: port tool node improvements back to langgraph ([#&#8203;6321](https://github.com/langchain-ai/langgraph/issues/6321)) - chore: bump core dep for prebuilt ([#&#8203;6323](https://github.com/langchain-ai/langgraph/issues/6323)) - fix(langgraph): dont persist UntrackedValue ([#&#8203;6316](https://github.com/langchain-ai/langgraph/issues/6316)) - feat(langgraph): add Overwrite to bypass reducer ([#&#8203;6286](https://github.com/langchain-ai/langgraph/issues/6286)) - fix: replace `python.langchain` links with new `docs.langchain` ([#&#8203;6352](https://github.com/langchain-ai/langgraph/issues/6352)) - release: Checkpointers 3.0 ([#&#8203;6313](https://github.com/langchain-ai/langgraph/issues/6313)) - chore: Restrict "json" type deserialization ([#&#8203;6269](https://github.com/langchain-ai/langgraph/issues/6269)) - release: langgraph + langgraph-prebuilt v1.0.0 ([#&#8203;6300](https://github.com/langchain-ai/langgraph/issues/6300)) - feat: adding cursory Python 3.14 support ([#&#8203;6298](https://github.com/langchain-ai/langgraph/issues/6298)) - style: fixes for ref docs ([#&#8203;6297](https://github.com/langchain-ai/langgraph/issues/6297)) - chore: release rcs for prebuilt + langgraph ([#&#8203;6296](https://github.com/langchain-ai/langgraph/issues/6296)) - chore: drop Python 3.9 (and syntax) ([#&#8203;6289](https://github.com/langchain-ai/langgraph/issues/6289)) - docs: style linting ([#&#8203;6260](https://github.com/langchain-ai/langgraph/issues/6260)) - release(langgraph): v1 working branch ([#&#8203;6093](https://github.com/langchain-ai/langgraph/issues/6093)) - fix: rename away from LangGraph Platform ([#&#8203;6281](https://github.com/langchain-ai/langgraph/issues/6281)) ### [`v1.0.1`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.1) [Compare Source](https://github.com/langchain-ai/langgraph/compare/1.0.0...1.0.1) Changes since 1.0.0 - release: Checkpointers 3.0 ([#&#8203;6313](https://github.com/langchain-ai/langgraph/issues/6313)) - chore: Restrict "json" type deserialization ([#&#8203;6269](https://github.com/langchain-ai/langgraph/issues/6269)) ### [`v1.0.0`](https://github.com/langchain-ai/langgraph/releases/tag/1.0.0) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.11...1.0.0) Changes since 1.0.0rc1 - release: langgraph + langgraph-prebuilt v1.0.0 ([#&#8203;6300](https://github.com/langchain-ai/langgraph/issues/6300)) - feat: adding cursory Python 3.14 support ([#&#8203;6298](https://github.com/langchain-ai/langgraph/issues/6298)) - style: fixes for ref docs ([#&#8203;6297](https://github.com/langchain-ai/langgraph/issues/6297)) ### [`v0.6.11`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.11) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.10...0.6.11) Changes since 0.6.10 - chore: Allow checkpoint 3.0 in 0.6.\* ([#&#8203;6315](https://github.com/langchain-ai/langgraph/issues/6315)) ### [`v0.6.10`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.10) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.9...0.6.10) Changes since 0.6.9 - chore(langgraph): bump langgraph version ([#&#8203;6257](https://github.com/langchain-ai/langgraph/issues/6257)) - fix(langgraph): revert selective interrupt task scheduling ([#&#8203;6252](https://github.com/langchain-ai/langgraph/issues/6252)) ### [`v0.6.9`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.9) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.8...0.6.9) Changes since 1.0.0a4 - chore(langgraph): bump version ([#&#8203;6245](https://github.com/langchain-ai/langgraph/issues/6245)) - chore(checkpoint): bump patch version ([#&#8203;6244](https://github.com/langchain-ai/langgraph/issues/6244)) - fix(langgraph): selective interrupt task scheduling ([#&#8203;6158](https://github.com/langchain-ai/langgraph/issues/6158)) - fix(langgraph): task result from stream mode debug / tasks should match format from get_state_history / get_state ([#&#8203;6233](https://github.com/langchain-ai/langgraph/issues/6233)) - fix(langgraph): don't use rst code blocks in docstrings ([#&#8203;6231](https://github.com/langchain-ai/langgraph/issues/6231)) - docs(langgraph): standardize version-added admonitions ([#&#8203;6230](https://github.com/langchain-ai/langgraph/issues/6230)) - fix(langgraph): fix supersteps not populating task.result field ([#&#8203;6195](https://github.com/langchain-ai/langgraph/issues/6195)) - fix(langgraph): revert -- reuse cached writes on nested resume to prevent task re-execution ([#&#8203;6227](https://github.com/langchain-ai/langgraph/issues/6227)) - chore(checkpoint-postgres): bump version ([#&#8203;6222](https://github.com/langchain-ai/langgraph/issues/6222)) ### [`v0.6.8`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.8) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.7...0.6.8) Changes since 1.0.0a3 - release(langgraph): 0.6.8 ([#&#8203;6215](https://github.com/langchain-ai/langgraph/issues/6215)) - fix(langgraph): handle multiple annotations w/ `BaseChannel` detection ([#&#8203;6210](https://github.com/langchain-ai/langgraph/issues/6210)) - fix(langgraph): `CheckpointTask.state` can be a `StateSnapshot` ([#&#8203;6201](https://github.com/langchain-ai/langgraph/issues/6201)) - chore(langgraph): clean up ruff format config ([#&#8203;6188](https://github.com/langchain-ai/langgraph/issues/6188)) - fix(langgraph): cleanup orphaned waiter task in `AsyncPregelLoop` ([#&#8203;6167](https://github.com/langchain-ai/langgraph/issues/6167)) - fix(langgraph): fix graph rendering for `defer=True` ([#&#8203;6130](https://github.com/langchain-ai/langgraph/issues/6130)) - fix(langgraph): reuse cached writes on nested resume to prevent task re-execution ([#&#8203;6161](https://github.com/langchain-ai/langgraph/issues/6161)) - chore(sdk-py): allow UUIDs in config ([#&#8203;6151](https://github.com/langchain-ai/langgraph/issues/6151)) - revert(langgraph): restore logic to surface interrupts for stream_mod… ([#&#8203;6141](https://github.com/langchain-ai/langgraph/issues/6141)) - chore(langgraph): Log when no values event is emitted from `RemoteGraph` ([#&#8203;6140](https://github.com/langchain-ai/langgraph/issues/6140)) - chore(cli): Add config schema ([#&#8203;6142](https://github.com/langchain-ai/langgraph/issues/6142)) - fix(langgraph): `get_graph` generates unexpected conditional edge ([#&#8203;6122](https://github.com/langchain-ai/langgraph/issues/6122)) - fix(langgraph): type checking for async w/ functional API ([#&#8203;6126](https://github.com/langchain-ai/langgraph/issues/6126)) - feat(langgraph): prevent arbitrary resumes w/ multiple pending interrupts ([#&#8203;6108](https://github.com/langchain-ai/langgraph/issues/6108)) - fix(langgraph): key error on runtime for config w/o configurable ([#&#8203;6106](https://github.com/langchain-ai/langgraph/issues/6106)) ### [`v0.6.7`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.7) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.6...0.6.7) Changes since 1.0.0a2 - chore: update emphemeral local ([#&#8203;6091](https://github.com/langchain-ai/langgraph/issues/6091)) - fix: Unwrap Required/NotRequired special forms before resolving channel/reducer annotations ([#&#8203;6080](https://github.com/langchain-ai/langgraph/issues/6080)) - monorepo support in CLI ([#&#8203;6028](https://github.com/langchain-ai/langgraph/issues/6028)) ### [`v0.6.6`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.6) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.5...0.6.6) Changes since 0.6.5 - fix(langgraph): Remote Baggage ([#&#8203;5964](https://github.com/langchain-ai/langgraph/issues/5964)) - chore(langgraph): Add passthrough params/headers to invoke/stream/etc. ([#&#8203;5940](https://github.com/langchain-ai/langgraph/issues/5940)) - feat(sdk-py): client qparams ([#&#8203;5918](https://github.com/langchain-ai/langgraph/issues/5918)) ### [`v0.6.5`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.5) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.4...0.6.5) Changes since 0.6.4 - release(langgraph): 0.6.5 ([#&#8203;5901](https://github.com/langchain-ai/langgraph/issues/5901)) - fix: Persist resume_map values ([#&#8203;5898](https://github.com/langchain-ai/langgraph/issues/5898)) - feat(langgraph): implement redis node level cache ([#&#8203;5834](https://github.com/langchain-ai/langgraph/issues/5834)) ### [`v0.6.4`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.4) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.3...0.6.4) Changes since 0.6.3 - release: langgraph + prebuilt 0.6.4 ([#&#8203;5854](https://github.com/langchain-ai/langgraph/issues/5854)) - fix: mypy issue with conditional edges ([#&#8203;5851](https://github.com/langchain-ai/langgraph/issues/5851)) - fix(langgraph): support emitting messages from subgraphs when messages mode explicitly requested ([#&#8203;5836](https://github.com/langchain-ai/langgraph/issues/5836)) - perf(langgraph): Save `updated_channels` to checkpoint ([#&#8203;5828](https://github.com/langchain-ai/langgraph/issues/5828)) - chore(langgraph): deprecate `MessageGraph` ([#&#8203;5843](https://github.com/langchain-ai/langgraph/issues/5843)) ### [`v0.6.3`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.3) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.2...0.6.3) Changes since 0.6.2 - fix(langgraph): Tidy up `AgentState` ([#&#8203;5801](https://github.com/langchain-ai/langgraph/issues/5801)) - release: langgraph + prebuilt 0.6.3 ([#&#8203;5799](https://github.com/langchain-ai/langgraph/issues/5799)) - fix(langgraph): Add warning for incorrect node signature with mistyped `config` param ([#&#8203;5798](https://github.com/langchain-ai/langgraph/issues/5798)) - fix(langgraph): fix up deprecation warnings ([#&#8203;5796](https://github.com/langchain-ai/langgraph/issues/5796)) - feat(langgraph): add durability mode for `invoke` and `ainvoke` ([#&#8203;5771](https://github.com/langchain-ai/langgraph/issues/5771)) - fix(docs): Add missing imports to make examples runnable ([#&#8203;5477](https://github.com/langchain-ai/langgraph/issues/5477)) - fix(langgraph): Remove duplicate call to ensure_config ([#&#8203;5768](https://github.com/langchain-ai/langgraph/issues/5768)) ### [`v0.6.2`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.2) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.1...0.6.2) Changes since 0.6.1 - fix(prebuilt): assign `context_schema` to `config_schema` with correct condition ([#&#8203;5746](https://github.com/langchain-ai/langgraph/issues/5746)) - feat(langgraph): Add context coercion for LangGraph runtime ([#&#8203;5736](https://github.com/langchain-ai/langgraph/issues/5736)) ### [`v0.6.1`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.1) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.6.0...0.6.1) Changes since 0.6.0 - fix(langgraph): use parent runtime when available ([#&#8203;5707](https://github.com/langchain-ai/langgraph/issues/5707)) - fix(langgraph): inject config even when optional ([#&#8203;5708](https://github.com/langchain-ai/langgraph/issues/5708)) ### [`v0.6.0`](https://github.com/langchain-ai/langgraph/releases/tag/0.6.0) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.5.4...0.6.0) ### 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): ```python from langchain_core.runnables import RunnableConfig from langgraph.graph import StateGraph def node(state: State, config: RunnableConfig): # verbose .get() access pattern for nested dicts user_id = config.get("configurable", {}).get("user_id") db_conn = config.get("configurable", {}).get("db_connection") ... builder = StateGraph(state_schema=State, config_schema=Config) # add nodes, edges, compile the graph... # nested runtime context passed to config's configurable key result = graph.invoke( {'input': 'abc'}, config={'configurable': {'user_id': '123', 'db_connection': 'conn_mock'}} ) ``` ##### After (v0.6): ```python from dataclasses import dataclass from langgraph.graph import StateGraph from langgraph.runtime import Runtime @dataclass class Context: """Context schema defined by the developer.""" user_id: str db_connection: str def node(state: State, runtime: Runtime[Context]): # type safe access to context attributes user_id = runtime.context.user_id db_conn = runtime.context.db_connection ... builder = StateGraph(state_schema=State, context_schema=Context) # add nodes, edges, compile the graph... # top level context arg is typed as Context for autocomplete and type checking result = graph.invoke( {'input': 'abc'}, context=Context(user_id='123', db_conn='conn_mock') ) ``` The [`Runtime`](https://langchain-ai.github.io/langgraph/reference/runtime/#runtime) class provides a single interface for accessing information like: - context: static data passed at the start of a run - store: storage mechanism for long term memories - stream_writer: custom function for writing to the graph’s output stream - for functional API users, `previous` is also available: the previous return value for the given thread Now, instead of injecting all of the above as separate parameters to node functions, developers can access them all through a single `runtime` parameter. ##### Migration Path - `config_schema` is deprecated in favor of `context_schema`, and will be removed in v2.0.0 - The new API maintains backward compatibility for existing code - Gradual migration is supported with deprecation warnings for `config_schema` - `get_config_jsonschema` is deprecated in favor of `get_context_jsonschema` (though this is generally only used for graph introspection and not by most langgraph users) #### 🔀 Dynamic model & tool selection `create_react_agent` can now dynamically choose both the model and tools at runtime using a custom context object: ```python from langgraph.prebuilt import create_react_agent @dataclass class CustomContext: provider: Literal["anthropic", "openai"] tools: list[str] def select_model(state, Runtime[Context]): model = { "openai": openai_model, "anthropic": anthropic_model, }[runtime.context.provider] selected_tools = [ tool for tool in [weather, compass] if tool.name in runtime.context.tools ] return model.bind_tools(selected_tools) agent = create_react_agent( select_model, # Initialize the agent with all known tools tools=[weather, compass] ) ``` Then invoke the agent with your desired settings: ```python agent.invoke( some_input, context=CustomContext(provider="openai", tools=["compass"]) ) ``` 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 - Choose between three durability modes: - `"exit"` - Save checkpoint only when the graph exits - equivalent to `checkpoint_during=False` - Least durable, fastest - `"async"` - Save checkpoint asynchronously while next step executes - equivalent to `checkpoint_during=True` - Moderately durable, mid speed - `"sync"` - Save checkpoint synchronously before next step - New! - Highly durable, slowest ##### Migration Path `checkpoint_during` is now deprecated in favor of the new `durability` argument. 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 `StateGraph` and `Pregel` interfaces are now generic over a graph’s: - `state_schema` - `context_schema` - `input_schema` - `output_schema` This means that: - Node signatures are type checked at graph creation time - Input to `invoke` / `stream` is type checked against the relevant schema - `context` available via the aforementioned `Runtime` class matches the `context_schema` #### 🔧 A Refined `Interrupt` Interface In preparation for v1.0, we’ve made a few changes to the `Interrupt` interface. Interrupts now have two attributes: - `id` - a unique identifier for the interrupt - `value` - the interrupt value In v0.6, we’ve removed the following attributes from the `Interrupt` class: - `when` - this was always `"during"` and offered no practical value - `resumable` - functionally, this is always `True` - `ns` - this information is now stored in a condensed format in the `id` attribute - `interrupt_id` has been deprecated in favor of `id`, 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: | Old Import | New Import | Status | | -------------------------------------------------- | ------------------------------------------- | ------------------------------------------------------------ | | `from langgraph.pregel.types import ...` | `from langgraph.types import ...` | ⚠️ Deprecated - Will be removed in V2 | | `from langgraph.constants import Send` | `from langgraph.types import Send` | ⚠️ Deprecated - Will be removed in V2 | | `from langgraph.constants import Interrupt` | `from langgraph.types import Interrupt` | ⚠️ Deprecated - Will be removed in V2 | | `from langgraph.channels import <ErrorClass>` | `from langgraph.errors import <ErrorClass>` | ❌ Removed - All errors now centralized in `langgraph.errors` | | `from langgraph.constants import TAG_NOSTREAM_ALT` | `from langgraph.constants import NOSTREAM` | ❌ Removed - Deprecated constant removed | #### 🎯 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](https://forum.langchain.com/) 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 - feat(sdk-py): sdk support for `context` API ([#&#8203;5566](https://github.com/langchain-ai/langgraph/issues/5566)) - feat(langgraph): Implement durability mode argument ([#&#8203;5432](https://github.com/langchain-ai/langgraph/issues/5432)) - refactor(langgraph): improve `Runtime` interface re patch/overrides ([#&#8203;5546](https://github.com/langchain-ai/langgraph/issues/5546)) - refactor(langgraph): make constants generally private with a few select exports ([#&#8203;5529](https://github.com/langchain-ai/langgraph/issues/5529)) - refactor(langgraph): move private typing constructs in `constants.py` -> `_internal/_typing.py` ([#&#8203;5518](https://github.com/langchain-ai/langgraph/issues/5518)) - feat(langgraph): new context api (replacing `config['configurable']` and `config_schema`) ([#&#8203;5243](https://github.com/langchain-ai/langgraph/issues/5243)) - feat(langgraph): add type checking for matching `node` signatures vs `input_schema` for `add_node` ([#&#8203;5424](https://github.com/langchain-ai/langgraph/issues/5424)) - change\[langgraph]: clean up `Interrupt` interface for v1 ([#&#8203;5405](https://github.com/langchain-ai/langgraph/issues/5405)) - langgraph\[change]: solidify public/private differentiations ([#&#8203;5252](https://github.com/langchain-ai/langgraph/issues/5252)) ### [`v0.5.4`](https://github.com/langchain-ai/langgraph/releases/tag/0.5.4) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.5.3...0.5.4) Changes since 0.5.3 - feat(langgraph): Handle ParentCommand in RemoteGraph ([#&#8203;5600](https://github.com/langchain-ai/langgraph/issues/5600)) - fix(langgraph): ignore write to `END` with `Command` ([#&#8203;5601](https://github.com/langchain-ai/langgraph/issues/5601)) - fix(langgraph): add `stacklevel=2` to the warnings to point to the caller’s codes ([#&#8203;5457](https://github.com/langchain-ai/langgraph/issues/5457)) ### [`v0.5.3`](https://github.com/langchain-ai/langgraph/releases/tag/0.5.3) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.5.2...0.5.3) Changes since 0.5.2 - release(langgraph): v0.5.3 ([#&#8203;5498](https://github.com/langchain-ai/langgraph/issues/5498)) - chore\[deps]: upgrade dependencies with `uv lock --upgrade` ([#&#8203;5471](https://github.com/langchain-ai/langgraph/issues/5471)) - docs(checkpoint-postgres): fix typo in comment ([#&#8203;5486](https://github.com/langchain-ai/langgraph/issues/5486)) - chore: add forum to readme ([#&#8203;5488](https://github.com/langchain-ai/langgraph/issues/5488)) - fix(langgraph): remove `ABC` spec for `PregelProtocol` ([#&#8203;5485](https://github.com/langchain-ai/langgraph/issues/5485)) - fix(langgraph): replace \_state_schema to state_schema when accessing StateGraph ([#&#8203;5436](https://github.com/langchain-ai/langgraph/issues/5436)) - fix(checkpoint-postgres): Remove python invalid escape warning ([#&#8203;5441](https://github.com/langchain-ai/langgraph/issues/5441)) ### [`v0.5.2`](https://github.com/langchain-ai/langgraph/releases/tag/0.5.2) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.5.1...0.5.2) Changes since 0.5.1 - patch\[langgraph]: Fix hint for `invoke`/`stream` to allow for `Command` and `None` ([#&#8203;5414](https://github.com/langchain-ai/langgraph/issues/5414)) ### [`v0.5.1`](https://github.com/langchain-ai/langgraph/releases/tag/0.5.1) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.5.0...0.5.1) Changes since 0.5.0 - langgraph\[fix]: remove deprecated pydantic logic + fix schema gen behavior for typed dicts ([#&#8203;5296](https://github.com/langchain-ai/langgraph/issues/5296)) - prebuilt\[patch]: import recognized tool message content block types from langchain-core ([#&#8203;5275](https://github.com/langchain-ai/langgraph/issues/5275)) ### [`v0.5.0`](https://github.com/langchain-ai/langgraph/releases/tag/0.5.0) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.10...0.5.0) #### 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: - lock in the few API surfaces that still felt “fuzzy” - prune internal code paths that made distributed execution harder - land long-queued performance work 0.5 is that housekeeping release. *** #### Headline changes ##### 1. A leaner, stricter StateGraph 1. **`state_schema` is 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. 2. **`input`/`output` → `input_schema`/`output_schema`** The old names still work but raise a deprecation warning. ```python graph = StateGraph( state_schema=MyState, input_schema=UserQuery, output_schema=AssistantResponse, ) ``` 3. **New `NodeBuilder` utility** A simpler, declarative way to create nodes and attach them to channels. The old `Channel.subscribe_to` helper 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: ```python graph.stream(stream_mode="tasks") # only task-level updates graph.stream(stream_mode="checkpoints") # only checkpoint deltas ``` This makes it cheaper to subscribe only to the information you need. ##### 3. Checkpointing overhaul - **Smaller blobs & faster restores** Redundant keys have been dropped and per-task writes are stored directly. - **Seamless migrations** Legacy “pending \_sends” data is auto-migrated the first time it is loaded. Custom checkpointers continue to work unchanged. ##### 4. Better serialization - `JsonPlusSerializer` now handles NumPy arrays stored in your state (including Fortran-ordered ones) without falling back to `pickle`. *** #### Minor breaking changes you might notice 1. **`state_schema` required** – add it if you were passing only input and output schemas instead (very rare). 2. **`input` / `output` renaming** – rename to `input_schema` / `output_schema`. 3. **Subclassing internals** – if you subclassed `PregelNode` *and* `Runnable`, drop the latter. Nothing else should require code changes. *** #### How to upgrade ```bash pip install -U "langgraph>=0.5" ``` 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](https://github.com/langchain-ai/langgraph/issues/4973). #### Detailed Changelog - feat: crons sorting sdk ([#&#8203;5197](https://github.com/langchain-ai/langgraph/issues/5197)) - feat(langgraph): task masquerading with update state - Add print_mode= arg to invoke/stream ([#&#8203;5201](https://github.com/langchain-ai/langgraph/issues/5201)) - Fix bug where Command(update=) could be ignored if there was a 2nd interrupt after it - Reduce extraneous keys in checkpoint.metadata - If FuturesDict callback has been GCed, don't call it - Add migration for pending_sends ([#&#8203;5106](https://github.com/langchain-ai/langgraph/issues/5106)) - Introduce "tasks" and "checkpoints" stream modes - Add migration for pending_sends - Restore compatibility with custom checkpointer classes created in prior versions - Revert "Remove UntrackedValue channel" - Revert "Remove `MessageGraph` ([#&#8203;4875](https://github.com/langchain-ai/langgraph/issues/4875))" - fix(langgraph): remove deprecated `output` usage in favor of `output_schema` ([#&#8203;5095](https://github.com/langchain-ai/langgraph/issues/5095)) - refactor(langgraph): Remove `PregelNode`'s inheritance from `Runnable` ([#&#8203;5093](https://github.com/langchain-ai/langgraph/issues/5093)) - Remove support for node reading a single managed value - PregelLoop: Simplify tick() method ([#&#8203;5080](https://github.com/langchain-ai/langgraph/issues/5080)) - serialize/deserialize pandas with pickle fallback ([#&#8203;5057](https://github.com/langchain-ai/langgraph/issues/5057)) - Remove code paths no longer needed - Avoid saving checkpoints for subgraphs when checkpoint_during=False ([#&#8203;5051](https://github.com/langchain-ai/langgraph/issues/5051)) - Remove gitmcp badge ([#&#8203;5055](https://github.com/langchain-ai/langgraph/issues/5055)) - Avoid saving checkpoints for subgraphs when checkpoint_during=False - Clean up things for Matt! - Support numpy array serialization in JsonPlusSerializer ([#&#8203;5035](https://github.com/langchain-ai/langgraph/issues/5035)) - Update ormsgpack ([#&#8203;5034](https://github.com/langchain-ai/langgraph/issues/5034)) - deprecate `input` and `output` in favor of `input_schema` and `output_schema` ([#&#8203;4983](https://github.com/langchain-ai/langgraph/issues/4983)) - using StateT as default for InputT - lint: use `pep 604` union syntax and `pep 585` generic syntax ([#&#8203;4963](https://github.com/langchain-ai/langgraph/issues/4963)) - docs: remove references to `StateGraph(dict)` ([#&#8203;4964](https://github.com/langchain-ai/langgraph/issues/4964)) - rename `retry` -> `retry_policy` ([#&#8203;4957](https://github.com/langchain-ai/langgraph/issues/4957)) - Fix step_timeout causing ParentCommand/GraphInterrupt exception to bubble up - docs: fix a grammar issue in multiple files ([#&#8203;4935](https://github.com/langchain-ai/langgraph/issues/4935)) - Make it possible to run test command without docker installed ([#&#8203;4948](https://github.com/langchain-ai/langgraph/issues/4948)) - Fix async callback manager tag handling ([#&#8203;4949](https://github.com/langchain-ai/langgraph/issues/4949)) - Pass tags when configuring async callback manager - Allow same-name channels and nodes in StateGraph ([#&#8203;4944](https://github.com/langchain-ai/langgraph/issues/4944)) - Make it possible to run test command without docker installed - docs: fix ensure_config docstring - Allow same-name channels and nodes in StateGraph - Improve type checking on graph `init` and `invoke`/`stream` ([#&#8203;4932](https://github.com/langchain-ai/langgraph/issues/4932)) - docs: fix a grammar issue in multiple files - Fix makefile command file for dev server - pidfile was always empty - Fix Command(graph=PARENT) when used together w checkpointer=True - Remove unused deprecation decorator/warning ([#&#8203;4917](https://github.com/langchain-ai/langgraph/issues/4917)) - Require `state_schema` in `StateGraph.__init__` ([#&#8203;4897](https://github.com/langchain-ai/langgraph/issues/4897)) - Remove add_conditional_edge(..., then=) ([#&#8203;4893](https://github.com/langchain-ai/langgraph/issues/4893)) - Avoid repeated runtime calls to get_type_hints - Remove `MessageGraph` ([#&#8203;4875](https://github.com/langchain-ai/langgraph/issues/4875)) - Remove non-state Graph - Remove UntrackedValue channel - Remove UntrackedValue channel ([#&#8203;4859](https://github.com/langchain-ai/langgraph/issues/4859)) - Flip default for checkpoint_during - Remove Channel node builder - Update managed value usage in local_read ([#&#8203;4854](https://github.com/langchain-ai/langgraph/issues/4854)) - Modify stream mode messages and custom to respect subgraphs=False ([#&#8203;4843](https://github.com/langchain-ai/langgraph/issues/4843)) - Remove Checkpoint.writes ([#&#8203;4822](https://github.com/langchain-ai/langgraph/issues/4822)) - Remove Checkpoint.pending_sends ([#&#8203;4820](https://github.com/langchain-ai/langgraph/issues/4820)) - Pregel: Add NodeBuilder class to replace Channel.subscribe_to ([#&#8203;4819](https://github.com/langchain-ai/langgraph/issues/4819)) - Remove dict subclasses used for values/updates stream chunks ([#&#8203;4816](https://github.com/langchain-ai/langgraph/issues/4816)) - Remove old checkpoint test fixtures ([#&#8203;4814](https://github.com/langchain-ai/langgraph/issues/4814)) - Remove postgres shallow checkpointer ([#&#8203;4813](https://github.com/langchain-ai/langgraph/issues/4813)) - Remove ChannelsManager, ManagedValues are now static classes and never instantiated ([#&#8203;4812](https://github.com/langchain-ai/langgraph/issues/4812)) - Remove Context channel / managed value, Remove SharedValue ([#&#8203;4857](https://github.com/langchain-ai/langgraph/issues/4857)) - Pregel: Add NodeBuilder class to replace Channel.subscribe_to - Remove old checkpoint test fixtures - Remove postgres shallow checkpointer - Remove ChannelsManager, ManagedValues are now static classes and never instantiated - Remove Context channel / managed value, Remove SharedValue ### [`v0.4.10`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.10) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.9...0.4.10) Changes since 0.4.9 - Revert change to default value of checkpoint_during arg ([#&#8203;5177](https://github.com/langchain-ai/langgraph/issues/5177)) - Revert change to default value of checkpoint_during arg - Fix bug where Command(update=) could be ignored if there was a 2nd interrupt after it - Reduce extraneous keys in checkpoint.metadata - If FuturesDict callback has been GCed, don't call it - langgraph 0.5.0rc1 - langgraph 0.5.0rc0 - Preparation for 0.5 release: langgraph-checkpoint ([#&#8203;5124](https://github.com/langchain-ai/langgraph/issues/5124)) - Preparation for 0.5 release - Add migration for pending_sends ([#&#8203;5106](https://github.com/langchain-ai/langgraph/issues/5106)) - Lint - Introduce "tasks" and "checkpoints" stream modes - Add migration for pending_sends - Restore compatibility with custom checkpointer classes created in prior versions - Revert "Remove UntrackedValue channel" - Revert "Remove `MessageGraph` ([#&#8203;4875](https://github.com/langchain-ai/langgraph/issues/4875))" - fix(langgraph): remove deprecated `output` usage in favor of `output_schema` ([#&#8203;5095](https://github.com/langchain-ai/langgraph/issues/5095)) - refactor(langgraph): Remove `PregelNode`'s inheritance from `Runnable` ([#&#8203;5093](https://github.com/langchain-ai/langgraph/issues/5093)) - Remove support for node reading a single managed value - PregelLoop: Simplify tick() method ([#&#8203;5080](https://github.com/langchain-ai/langgraph/issues/5080)) - serialize/deserialize pandas with pickle fallback ([#&#8203;5057](https://github.com/langchain-ai/langgraph/issues/5057)) - Remove code paths no longer needed - Avoid saving checkpoints for subgraphs when checkpoint_during=False ([#&#8203;5051](https://github.com/langchain-ai/langgraph/issues/5051)) - Remove gitmcp badge ([#&#8203;5055](https://github.com/langchain-ai/langgraph/issues/5055)) - Update existing - Add tests - Avoid saving checkpoints for subgraphs when checkpoint_during=False - Clean up things for Matt! - Support numpy array serialization in JsonPlusSerializer ([#&#8203;5035](https://github.com/langchain-ai/langgraph/issues/5035)) - Update ormsgpack ([#&#8203;5034](https://github.com/langchain-ai/langgraph/issues/5034)) - deprecate `input` and `output` in favor of `input_schema` and `output_schema` ([#&#8203;4983](https://github.com/langchain-ai/langgraph/issues/4983)) - using StateT as default for InputT - lint: use `pep 604` union syntax and `pep 585` generic syntax ([#&#8203;4963](https://github.com/langchain-ai/langgraph/issues/4963)) - docs: remove references to `StateGraph(dict)` ([#&#8203;4964](https://github.com/langchain-ai/langgraph/issues/4964)) - rename `retry` -> `retry_policy` ([#&#8203;4957](https://github.com/langchain-ai/langgraph/issues/4957)) - Fix step_timeout causing ParentCommand/GraphInterrupt exception to bubble up - docs: fix a grammar issue in multiple files ([#&#8203;4935](https://github.com/langchain-ai/langgraph/issues/4935)) - Make it possible to run test command without docker installed ([#&#8203;4948](https://github.com/langchain-ai/langgraph/issues/4948)) - Fix async callback manager tag handling ([#&#8203;4949](https://github.com/langchain-ai/langgraph/issues/4949)) - Pass tags when configuring async callback manager - Allow same-name channels and nodes in StateGraph ([#&#8203;4944](https://github.com/langchain-ai/langgraph/issues/4944)) - Make it possible to run test command without docker installed - docs: fix ensure_config docstring - Allow same-name channels and nodes in StateGraph - Improve type checking on graph `init` and `invoke`/`stream` ([#&#8203;4932](https://github.com/langchain-ai/langgraph/issues/4932)) - docs: fix a grammar issue in multiple files - Fix makefile command file for dev server - pidfile was always empty - Fix Command(graph=PARENT) when used together w checkpointer=True - Remove unused deprecation decorator/warning ([#&#8203;4917](https://github.com/langchain-ai/langgraph/issues/4917)) - Require `state_schema` in `StateGraph.__init__` ([#&#8203;4897](https://github.com/langchain-ai/langgraph/issues/4897)) - Remove add_conditional_edge(..., then=) ([#&#8203;4893](https://github.com/langchain-ai/langgraph/issues/4893)) - Avoid repeated runtime calls to get_type_hints - Remove `MessageGraph` ([#&#8203;4875](https://github.com/langchain-ai/langgraph/issues/4875)) - Remove non-state Graph - Remove UntrackedValue channel - Remove UntrackedValue channel ([#&#8203;4859](https://github.com/langchain-ai/langgraph/issues/4859)) - Flip default for checkpoint_during - Remove Channel node builder - Update managed value usage in local_read ([#&#8203;4854](https://github.com/langchain-ai/langgraph/issues/4854)) - Modify stream mode messages and custom to respect subgraphs=False ([#&#8203;4843](https://github.com/langchain-ai/langgraph/issues/4843)) - Remove Checkpoint.writes ([#&#8203;4822](https://github.com/langchain-ai/langgraph/issues/4822)) - Remove Checkpoint.pending_sends ([#&#8203;4820](https://github.com/langchain-ai/langgraph/issues/4820)) - Pregel: Add NodeBuilder class to replace Channel.subscribe_to ([#&#8203;4819](https://github.com/langchain-ai/langgraph/issues/4819)) - Remove dict subclasses used for values/updates stream chunks ([#&#8203;4816](https://github.com/langchain-ai/langgraph/issues/4816)) - Remove old checkpoint test fixtures ([#&#8203;4814](https://github.com/langchain-ai/langgraph/issues/4814)) - Remove postgres shallow checkpointer ([#&#8203;4813](https://github.com/langchain-ai/langgraph/issues/4813)) - Remove ChannelsManager, ManagedValues are now static classes and never instantiated ([#&#8203;4812](https://github.com/langchain-ai/langgraph/issues/4812)) - Remove Context channel / managed value, Remove SharedValue ([#&#8203;4857](https://github.com/langchain-ai/langgraph/issues/4857)) - Update managed value usage in local_read - Format - Remove Checkpoint.writes - Remove Checkpoint.pending_sends - Pregel: Add NodeBuilder class to replace Channel.subscribe_to - Lint - Remove old checkpoint test fixtures - Remove postgres shallow checkpointer - Remove ChannelsManager, ManagedValues are now static classes and never instantiated - Remove Context channel / managed value, Remove SharedValue ### [`v0.4.9`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.9) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.8...0.4.9) Changes since 0.5.0rc1 - Revert change to default value of checkpoint_during arg ([#&#8203;5177](https://github.com/langchain-ai/langgraph/issues/5177)) - Revert change to default value of checkpoint_during arg - Fix bug where Command(update=) could be ignored if there was a 2nd interrupt after it - Reduce extraneous keys in checkpoint.metadata - If FuturesDict callback has been GCed, don't call it ### [`v0.4.8`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.8) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.7...0.4.8) Changes since 0.4.7 - Remove unused deprecation decorator/warning ([#&#8203;4917](https://github.com/langchain-ai/langgraph/issues/4917)) - Require `state_schema` in `StateGraph.__init__` ([#&#8203;4897](https://github.com/langchain-ai/langgraph/issues/4897)) - Remove add_conditional_edge(..., then=) ([#&#8203;4893](https://github.com/langchain-ai/langgraph/issues/4893)) - Avoid repeated runtime calls to get_type_hints - Remove `MessageGraph` ([#&#8203;4875](https://github.com/langchain-ai/langgraph/issues/4875)) - Remove non-state Graph - Remove UntrackedValue channel - Remove UntrackedValue channel ([#&#8203;4859](https://github.com/langchain-ai/langgraph/issues/4859)) - Flip default for checkpoint_during - Remove Channel node builder - Update managed value usage in local_read ([#&#8203;4854](https://github.com/langchain-ai/langgraph/issues/4854)) - Modify stream mode messages and custom to respect subgraphs=False ([#&#8203;4843](https://github.com/langchain-ai/langgraph/issues/4843)) - Remove Checkpoint.writes ([#&#8203;4822](https://github.com/langchain-ai/langgraph/issues/4822)) - Remove Checkpoint.pending_sends ([#&#8203;4820](https://github.com/langchain-ai/langgraph/issues/4820)) - Pregel: Add NodeBuilder class to replace Channel.subscribe_to ([#&#8203;4819](https://github.com/langchain-ai/langgraph/issues/4819)) - Remove dict subclasses used for values/updates stream chunks ([#&#8203;4816](https://github.com/langchain-ai/langgraph/issues/4816)) - Remove old checkpoint test fixtures ([#&#8203;4814](https://github.com/langchain-ai/langgraph/issues/4814)) - Remove postgres shallow checkpointer ([#&#8203;4813](https://github.com/langchain-ai/langgraph/issues/4813)) - Remove ChannelsManager, ManagedValues are now static classes and never instantiated ([#&#8203;4812](https://github.com/langchain-ai/langgraph/issues/4812)) - Remove Context channel / managed value, Remove SharedValue ([#&#8203;4857](https://github.com/langchain-ai/langgraph/issues/4857)) - Remove SchemaCoercionMapper - Update managed value usage in local_read - prebuilt: release 0.2.2 ([#&#8203;4851](https://github.com/langchain-ai/langgraph/issues/4851)) - Format - Remove Checkpoint.writes - Remove Checkpoint.pending_sends - Pregel: Add NodeBuilder class to replace Channel.subscribe_to - Lint - Remove old checkpoint test fixtures - Remove postgres shallow checkpointer - Remove ChannelsManager, ManagedValues are now static classes and never instantiated - Remove Context channel / managed value, Remove SharedValue - Merge branch 'main' into fix-examples-link-readme - Update libs/langgraph/README.md - Update libs/langgraph/README.md - Merge branch 'main' into fix-examples-link-readme - Merge branch 'main' into fix-examples-link-readme - Merge branch 'main' into fix-examples-link-readme - Merge branch 'langchain-ai:main' into fix-examples-link-readme - fix link to examples page in readme.md ### [`v0.4.7`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.7) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.6...0.4.7) Changes since 0.4.6 - 0.4.7 - Fix stream mode not respected in subgraphs - Try to make test less flaky - Add tests for stream_events when using imperative api ### [`v0.4.6`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.6) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.5...0.4.6) Changes since 0.4.5 - 0.4.6 - Fix - Lint - Add sync test - Fix - Fix exception handling for imperative tasks - prebuilt: release 0.2.1 ([#&#8203;4801](https://github.com/langchain-ai/langgraph/issues/4801)) - prebuilt: release 0.2.0 ([#&#8203;4793](https://github.com/langchain-ai/langgraph/issues/4793)) - Code review - feat(langgraph): push_messages should directly write to the state - feat(graph): add push_message method to push manually to `messages` / `message-tuple` stream ([#&#8203;4722](https://github.com/langchain-ai/langgraph/issues/4722)) - Add message state test - Apply same condition for stream_mode=values in suppress interrupt - Remove redundant cast - feat(graph): add push_message method to push manually to `messages` / `message-tuple` stream - Last uv.lock? - Only emit stream values chunks when the output channels have changed ([#&#8203;4774](https://github.com/langchain-ai/langgraph/issues/4774)) - docs: fix example pregel reducer ([#&#8203;4740](https://github.com/langchain-ai/langgraph/issues/4740)) - prebuilts hitl: fix branching logic + add structural snapshot tests ([#&#8203;4767](https://github.com/langchain-ai/langgraph/issues/4767)) - docs: deferred nodes ([#&#8203;4759](https://github.com/langchain-ai/langgraph/issues/4759)) - Remove local_write utility ([#&#8203;4751](https://github.com/langchain-ai/langgraph/issues/4751)) - Print output for cached [@&#8203;task](https://github.com/task) functions ([#&#8203;4750](https://github.com/langchain-ai/langgraph/issues/4750)) - Print output for cached [@&#8203;task](https://github.com/task) functions - sqlite: Add test for search with list filters ([#&#8203;4747](https://github.com/langchain-ai/langgraph/issues/4747)) - sqlite: update list_namespaces with max_depth ([#&#8203;4746](https://github.com/langchain-ai/langgraph/issues/4746)) - Update lockfile - docs: Clean up adjective use in readme ([#&#8203;4734](https://github.com/langchain-ai/langgraph/issues/4734)) - SqliteStore ([#&#8203;3608](https://github.com/langchain-ai/langgraph/issues/3608)) - ci: fix benchmark command ([#&#8203;4729](https://github.com/langchain-ai/langgraph/issues/4729)) - docs: document tuples in streaming guides ([#&#8203;4690](https://github.com/langchain-ai/langgraph/issues/4690)) - ci: migrate to uv! ([#&#8203;4698](https://github.com/langchain-ai/langgraph/issues/4698)) ### [`v0.4.5`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.5) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.4...0.4.5) Changes since 0.4.4 - langgraph: release 0.4.5 ([#&#8203;4709](https://github.com/langchain-ai/langgraph/issues/4709)) - Improve how we match cached writes for async imperative tasks ([#&#8203;4691](https://github.com/langchain-ai/langgraph/issues/4691)) - Lint - Improve how we match cached writes for async imperative tasks ### [`v0.4.4`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.4) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.3...0.4.4) Changes since 0.4.3 - langgraph: release 0.4.4 ([#&#8203;4704](https://github.com/langchain-ai/langgraph/issues/4704)) - update for consistency - lint again - use list - lint + update - update - update - update - langgraph: fix drawing graph with **root** channel - langgraph: fix graph drawing for self-loops - Merge branch 'main' into clean - \[Proposal] add gitmcp badge for simple LLM-accessible documentation ([#&#8203;4502](https://github.com/langchain-ai/langgraph/issues/4502)) - Change LG Cloud > Platform - Fix invalid channels error message ([#&#8203;4357](https://github.com/langchain-ai/langgraph/issues/4357)) - docs: remove unused annotation comments in readmes ([#&#8203;4656](https://github.com/langchain-ai/langgraph/issues/4656)) - docs: General language cleanup ([#&#8203;4649](https://github.com/langchain-ai/langgraph/issues/4649)) - \[docs] LangGraph / LangGraph Platform docs updates ([#&#8203;4479](https://github.com/langchain-ai/langgraph/issues/4479)) - Port checkpointer/store fixtures in langgraph-prebuilt to idiomatic pattern ([#&#8203;4629](https://github.com/langchain-ai/langgraph/issues/4629)) - Extend all tests using InMemorySaver to use all available checkpointers ([#&#8203;4627](https://github.com/langchain-ai/langgraph/issues/4627)) - Implement update_state for functional api ([#&#8203;4626](https://github.com/langchain-ai/langgraph/issues/4626)) - More idiomatic sync/async checkpointer fixtures in pytest ([#&#8203;4624](https://github.com/langchain-ai/langgraph/issues/4624)) - More idiomatic sync/async store fixtures in pytest ([#&#8203;4617](https://github.com/langchain-ai/langgraph/issues/4617)) - Overload clear method to delete all when called without args - Limit depth - Extend all tests using InMemorySaver to use all available checkpointers - Implement update_state for functional api - Fix - More idiomatic sync/async checkpointer fixtures in pytest - Move to sep file - More idiomatic sync/async store fixtures in pytest - Lint - Lint - Move FileCache to sqlite package, add InMemoryCache - Add clear cache methods - Remove refresh - Add namespace to cache keys - Lint - Lint - Lint - Lint - Lint - Lint - Lint - Finish implementation, add tests - Lint - Lint - Accept default cache_policy for graph/entrypoint/pregel - Fix - Fix - Output cached writes - Lint - Rename - Fix type annotation - Fixes - Re-do with separate cache interface - WIP - Deferred Node ([#&#8203;4269](https://github.com/langchain-ai/langgraph/issues/4269)) - Start dev server externally ([#&#8203;4604](https://github.com/langchain-ai/langgraph/issues/4604)) ### [`v0.4.3`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.3) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.2...0.4.3) Changes since 0.4.2 - release: 0.4.3 ([#&#8203;4592](https://github.com/langchain-ai/langgraph/issues/4592)) - langgraph: use tuples for streamed message events in RemoteGraph ([#&#8203;4589](https://github.com/langchain-ai/langgraph/issues/4589)) - Fix remote streaming of subgraphs ([#&#8203;4590](https://github.com/langchain-ai/langgraph/issues/4590)) - Add a limit to Pregel.draw ([#&#8203;4575](https://github.com/langchain-ai/langgraph/issues/4575)) ### [`v0.4.2`](https://github.com/langchain-ai/langgraph/releases/tag/0.4.2) [Compare Source](https://github.com/langchain-ai/langgraph/compare/0.4.1...0.4.2) Changes since 0.4.1 - 0.4.2 ([#&#8203;4570](https://github.com/langchain-ai/langgraph/issues/4570)) - update - langgraph: decouple name from assistant ID in RemoteGraph - prebuilt: switch to executing parallel tool calls via Send by default ([#&#8203;4438](https://github.com/langchain-ai/langgraph/issues/4438)) - fix for langgraph - docstring fixes for libs/langgraph - docs: update add_messages API ref ([#&#8203;4495](https://github.com/langchain-ai/langgraph/issues/4495)) - update other tests </details> <details> <summary>pytest-dev/pytest (pytest)</summary> ### [`v9.0.3`](https://github.com/pytest-dev/pytest/releases/tag/9.0.3) [Compare Source](https://github.com/pytest-dev/pytest/compare/9.0.2...9.0.3) ### pytest 9.0.3 (2026-04-07) #### Bug fixes - [#&#8203;12444](https://github.com/pytest-dev/pytest/issues/12444): Fixed `pytest.approx` which now correctly takes into account `~collections.abc.Mapping` keys order to compare them. - [#&#8203;13634](https://github.com/pytest-dev/pytest/issues/13634): Blocking a `conftest.py` file 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 `UsageError` explaining that conftest files are not plugins and cannot be disabled via `-p`. - [#&#8203;13734](https://github.com/pytest-dev/pytest/issues/13734): Fixed crash when a test raises an exceptiongroup with `__tracebackhide__ = True`. - [#&#8203;14195](https://github.com/pytest-dev/pytest/issues/14195): Fixed an issue where non-string messages passed to <span class="title-ref">unittest.TestCase.subTest()</span> were not printed. - [#&#8203;14343](https://github.com/pytest-dev/pytest/issues/14343): Fixed use of insecure temporary directory (CVE-2025-71176). #### Improved documentation - [#&#8203;13388](https://github.com/pytest-dev/pytest/issues/13388): Clarified documentation for `-p` vs `PYTEST_PLUGINS` plugin loading and fixed an incorrect `-p` example. - [#&#8203;13731](https://github.com/pytest-dev/pytest/issues/13731): Clarified that capture fixtures (e.g. `capsys` and `capfd`) take precedence over the `-s` / `--capture=no` command-line options in `Accessing captured output from a test function <accessing-captured-output>`. - [#&#8203;14088](https://github.com/pytest-dev/pytest/issues/14088): Clarified that the default `pytest_collection` hook sets `session.items` before it calls `pytest_collection_finish`, not after. - [#&#8203;14255](https://github.com/pytest-dev/pytest/issues/14255): TOML integer log levels must be quoted: Updating reference documentation. #### Contributor-facing changes - [#&#8203;12689](https://github.com/pytest-dev/pytest/issues/12689): The test reports are now published to Codecov from GitHub Actions. The test statistics is visible [on the web interface](https://app.codecov.io/gh/pytest-dev/pytest/tests). \-- by `aleguy02` ### [`v9.0.2`](https://github.com/pytest-dev/pytest/releases/tag/9.0.2) [Compare Source](https://github.com/pytest-dev/pytest/compare/9.0.1...9.0.2) ### pytest 9.0.2 (2025-12-06) #### Bug fixes - [#&#8203;13896](https://github.com/pytest-dev/pytest/issues/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 `TERM` is `dumb`, the escape codes are no longer emitted, even if the plugin is enabled. - [#&#8203;13904](https://github.com/pytest-dev/pytest/issues/13904): Fixed the TOML type of the `tmp_path_retention_count` settings in the API reference from number to string. - [#&#8203;13946](https://github.com/pytest-dev/pytest/issues/13946): The private `config.inicfg` attribute 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. - [#&#8203;13965](https://github.com/pytest-dev/pytest/issues/13965): Fixed quadratic-time behavior when handling `unittest` subtests in Python 3.10. #### Improved documentation - [#&#8203;4492](https://github.com/pytest-dev/pytest/issues/4492): The API Reference now contains cross-reference-able documentation of `pytest's command-line flags <command-line-flags>`. ### [`v9.0.1`](https://github.com/pytest-dev/pytest/releases/tag/9.0.1) [Compare Source](https://github.com/pytest-dev/pytest/compare/9.0.0...9.0.1) ### pytest 9.0.1 (2025-11-12) #### Bug fixes - [#&#8203;13895](https://github.com/pytest-dev/pytest/issues/13895): Restore support for skipping tests via `raise unittest.SkipTest`. - [#&#8203;13896](https://github.com/pytest-dev/pytest/issues/13896): The terminal progress plugin added in pytest 9.0 is now automatically disabled when iTerm2 is detected, it generated desktop notifications instead of the desired functionality. - [#&#8203;13904](https://github.com/pytest-dev/pytest/issues/13904): Fixed the TOML type of the verbosity settings in the API reference from number to string. - [#&#8203;13910](https://github.com/pytest-dev/pytest/issues/13910): Fixed <span class="title-ref">UserWarning: Do not expect file_or_dir</span> on some earlier Python 3.12 and 3.13 point versions. #### Packaging updates and notes for downstreams - [#&#8203;13933](https://github.com/pytest-dev/pytest/issues/13933): The tox configuration has been adjusted to make sure the desired version string can be passed into its `package_env` through the `SETUPTOOLS_SCM_PRETEND_VERSION_FOR_PYTEST` environment variable as a part of the release process -- by `webknjaz`. #### Contributor-facing changes - [#&#8203;13891](https://github.com/pytest-dev/pytest/issues/13891), [#&#8203;13942](https://github.com/pytest-dev/pytest/issues/13942): The CI/CD part of the release automation is now capable of creating GitHub Releases without having a Git checkout on disk -- by `bluetech` and `webknjaz`. - [#&#8203;13933](https://github.com/pytest-dev/pytest/issues/13933): The tox configuration has been adjusted to make sure the desired version string can be passed into its `package_env` through the `SETUPTOOLS_SCM_PRETEND_VERSION_FOR_PYTEST` environment variable as a part of the release process -- by `webknjaz`. ### [`v9.0.0`](https://github.com/pytest-dev/pytest/releases/tag/9.0.0) [Compare Source](https://github.com/pytest-dev/pytest/compare/8.4.2...9.0.0) ### pytest 9.0.0 (2025-11-05) #### New features - [#&#8203;1367](https://github.com/pytest-dev/pytest/issues/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: ```python def contains_docstring(p: Path) -> bool: """Return True if the given Python file contains a top-level docstring.""" ... def test_py_files_contain_docstring(subtests: pytest.Subtests) -> None: for path in Path.cwd().glob("*.py"): with subtests.test(path=str(path)): assert contains_docstring(path) ``` 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.subTest` is now also supported. This feature was originally implemented as a separate plugin in [pytest-subtests](https://github.com/pytest-dev/pytest-subtests), but since then has been merged into the core. > \[!NOTE] > This feature is experimental and will likely evolve in future releases. By that we mean that we might change how subtests are reported on failure, but the functionality and how to use it are stable. - [#&#8203;13743](https://github.com/pytest-dev/pytest/issues/13743): Added support for **native TOML configuration files**. While pytest, since version 6, supports configuration in `pyproject.toml` files 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. ```toml # pyproject.toml [tool.pytest] minversion = "9.0" addopts = ["-ra", "-q"] testpaths = [ "tests", "integration", ] ``` 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 use `pytest.toml` or `.pytest.toml`: ```toml # pytest.toml or .pytest.toml [pytest] minversion = "9.0" addopts = ["-ra", "-q"] testpaths = [ "tests", "integration", ] ``` The documentation now (sometimes) shows configuration snippets in both TOML and INI formats, in a tabbed interface. See `config file formats` for full details. - [#&#8203;13823](https://github.com/pytest-dev/pytest/issues/13823): Added a **"strict mode"** enabled by the `strict` configuration option. When set to `true`, the `strict` option currently enables - `strict_config` - `strict_markers` - `strict_parametrization_ids` - `strict_xfail` The individual strictness options can be explicitly set to override the global `strict` setting. The previously-deprecated `--strict` command-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 mode` for more details. - [#&#8203;13737](https://github.com/pytest-dev/pytest/issues/13737): Added the `strict_parametrization_ids` configuration option. When set, pytest emits an error if it detects non-unique parameter set IDs, rather than automatically making the IDs unique by adding <span class="title-ref">0</span>, <span class="title-ref">1</span>, ... to them. This can be particularly useful for catching unintended duplicates. - [#&#8203;13072](https://github.com/pytest-dev/pytest/issues/13072): Added support for displaying test session **progress in the terminal tab** using the [OSC 9;4;](https://conemu.github.io/en/AnsiEscapeCodes.html#ConEmu_specific_OSC) 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: - On a user level, using `-p no:terminalprogress` on the command line or via an environment variable `PYTEST_ADDOPTS='-p no:terminalprogress'`. - On a project configuration level, using `addopts = "-p no:terminalprogress"`. - [#&#8203;478](https://github.com/pytest-dev/pytest/issues/478): Support PEP420 (implicit namespace packages) as <span class="title-ref">--pyargs</span> target when `consider_namespace_packages` is <span class="title-ref">true</span> in the config. Previously, this option only impacted package imports, now it also impacts tests discovery. - [#&#8203;13678](https://github.com/pytest-dev/pytest/issues/13678): Added a new `faulthandler_exit_on_timeout` configuration option set to "false" by default to let <span class="title-ref">faulthandler</span> interrupt the <span class="title-ref">pytest</span> process after a timeout in case of deadlock. Previously, a <span class="title-ref">faulthandler</span> timeout would only dump the traceback of all threads to stderr, but would not interrupt the <span class="title-ref">pytest</span> process. \-- by `ogrisel`. - [#&#8203;13829](https://github.com/pytest-dev/pytest/issues/13829): Added support for configuration option aliases via the `aliases` parameter in `Parser.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 - [#&#8203;13330](https://github.com/pytest-dev/pytest/issues/13330): Having pytest configuration spread over more than one file (for example having both a `pytest.ini` file and `pyproject.toml` with 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` - [#&#8203;13574](https://github.com/pytest-dev/pytest/issues/13574): The single argument `--version` no longer loads the entire plugin infrastructure, making it faster and more reliable when displaying only the pytest version. Passing `--version` twice (e.g., `pytest --version --version`) retains the original behavior, showing both the pytest version and plugin information. > \[!NOTE] > Since `--version` is now processed early, it only takes effect when passed directly via the command line. It will not work if set through other mechanisms, such as `PYTEST_ADDOPTS` or `addopts`. - [#&#8203;13823](https://github.com/pytest-dev/pytest/issues/13823): Added `strict_xfail` as an alias to the `xfail_strict` option, `strict_config` as an alias to the `--strict-config` flag, and `strict_markers` as an alias to the `--strict-markers` flag. This makes all strictness options consistently have configuration options with the prefix `strict_`. - [#&#8203;13700](https://github.com/pytest-dev/pytest/issues/13700): <span class="title-ref">--junitxml</span> no longer prints the <span class="title-ref">generated xml file</span> summary at the end of the pytest session when <span class="title-ref">--quiet</span> is given. - [#&#8203;13732](https://github.com/pytest-dev/pytest/issues/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. - [#&#8203;13859](https://github.com/pytest-dev/pytest/issues/13859): Clarify the error message for <span class="title-ref">pytest.raises()</span> when a regex <span class="title-ref">match</span> fails. - [#&#8203;13861](https://github.com/pytest-dev/pytest/issues/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 - [#&#8203;12083](https://github.com/pytest-dev/pytest/issues/12083): Fixed a bug where an invocation such as <span class="title-ref">pytest a/ a/b</span> would cause only tests from <span class="title-ref">a/b</span> to run, and not other tests under <span class="title-ref">a/</span>. The fix entails a few breaking changes to how such overlapping arguments and duplicates are handled: 1. <span class="title-ref">pytest a/b a/</span> or <span class="title-ref">pytest a/ a/b</span> are equivalent to <span class="title-ref">pytest a</span>; if an argument overlaps another arguments, only the prefix remains. 2. <span class="title-ref">pytest x.py x.py</span> is equivalent to <span class="title-ref">pytest x.py</span>; previously such an invocation was taken as an explicit request to run the tests from the file twice. If you rely on these behaviors, consider using `--keep-duplicates <duplicate-paths>`, which retains its existing behavior (including the bug). - [#&#8203;13719](https://github.com/pytest-dev/pytest/issues/13719): Support for Python 3.9 is dropped following its end of life. - [#&#8203;13766](https://github.com/pytest-dev/pytest/issues/13766): Previously, pytest would assume it was running in a CI/CD environment if either of the environment variables <span class="title-ref">$CI</span> or <span class="title-ref">$BUILD_NUMBER</span> was defined; now, CI mode is only activated if at least one of those variables is defined and set to a *non-empty* value. - [#&#8203;13779](https://github.com/pytest-dev/pytest/issues/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 `PytestRemovedIn9Warning` now generate errors instead of warning messages by default. **The affected features will be effectively removed in pytest 9.1**, so please consult the `deprecations` section in the docs for directions on how to update existing code. In the pytest `9.0.X` series, it is possible to change the errors back into warnings as a stopgap measure by adding this to your `pytest.ini` file: ```ini [pytest] filterwarnings = ignore::pytest.PytestRemovedIn9Warning ``` But this will stop working when pytest `9.1` is released. **If you have concerns** about the removal of a specific feature, please add a comment to `13779`. #### Deprecations (removal in next major release) - [#&#8203;13807](https://github.com/pytest-dev/pytest/issues/13807): `monkeypatch.syspath_prepend() <pytest.MonkeyPatch.syspath_prepend>` now issues a deprecation warning when the prepended path contains legacy namespace packages (those using `pkg_resources.declare_namespace()`). Users should migrate to native namespace packages (`420`). See `monkeypatch-fixup-namespace-packages` for details. #### Bug fixes - [#&#8203;13445](https://github.com/pytest-dev/pytest/issues/13445): Made the type annotations of `pytest.skip` and friends more spec-complaint to have them work across more type checkers. - [#&#8203;13537](https://github.com/pytest-dev/pytest/issues/13537): Fixed a bug in which `ExceptionGroup` with only `Skipped` exceptions in teardown was not handled correctly and showed as error. - [#&#8203;13598](https://github.com/pytest-dev/pytest/issues/13598): Fixed possible collection confusion on Windows when short paths and symlinks are involved. - [#&#8203;13716](https://github.com/pytest-dev/pytest/issues/13716): Fixed a bug where a nonsensical invocation like `pytest x.py[a]` (a file cannot be parametrized) was silently treated as `pytest x.py`. This is now a usage error. - [#&#8203;13722](https://github.com/pytest-dev/pytest/issues/13722): Fixed a misleading assertion failure message when using `pytest.approx` on mappings with differing lengths. - [#&#8203;13773](https://github.com/pytest-dev/pytest/issues/13773): Fixed the static fixture closure calculation to properly consider transitive dependencies requested by overridden fixtures. - [#&#8203;13816](https://github.com/pytest-dev/pytest/issues/13816): Fixed `pytest.approx` which now returns a clearer error message when comparing mappings with different keys. - [#&#8203;13849](https://github.com/pytest-dev/pytest/issues/13849): Hidden `.pytest.ini` files are now picked up as the config file even if empty. This was an inconsistency with non-hidden `pytest.ini`. - [#&#8203;13865](https://github.com/pytest-dev/pytest/issues/13865): Fixed <span class="title-ref">--show-capture</span> with <span class="title-ref">--tb=line</span>. - [#&#8203;13522](https://github.com/pytest-dev/pytest/issues/13522): Fixed `pytester` in subprocess mode ignored all :attr\`pytester.plugins \<pytest.Pytester.plugins>\` except the first. Fixed `pytester` in subprocess mode silently ignored non-str `pytester.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 - [#&#8203;13791](https://github.com/pytest-dev/pytest/issues/13791): Minimum requirements on `iniconfig` and `packaging` were bumped to `1.0.1` and `22.0.0`, respectively. #### Contributor-facing changes - [#&#8203;12244](https://github.com/pytest-dev/pytest/issues/12244): Fixed self-test failures when <span class="title-ref">TERM=dumb</span>. - [#&#8203;12474](https://github.com/pytest-dev/pytest/issues/12474): Added scheduled GitHub Action Workflow to run Sphinx linkchecks in repo documentation. - [#&#8203;13621](https://github.com/pytest-dev/pytest/issues/13621): pytest's own testsuite now handles the `lsof` command hanging (e.g. due to unreachable network filesystems), with the affected selftests being skipped after 10 seconds. - [#&#8203;13638](https://github.com/pytest-dev/pytest/issues/13638): Fixed deprecated `gh pr new` command in `scripts/prepare-release-pr.py`. The script now uses `gh pr create` which is compatible with GitHub CLI v2.0+. - [#&#8203;13695](https://github.com/pytest-dev/pytest/issues/13695): Flush <span class="title-ref">stdout</span> and <span class="title-ref">stderr</span> in <span class="title-ref">Pytester.run</span> to avoid truncated outputs in <span class="title-ref">test_faulthandler.py::test_timeout</span> on CI -- by `ogrisel`. - [#&#8203;13771](https://github.com/pytest-dev/pytest/issues/13771): Skip <span class="title-ref">test_do_not_collect_symlink_siblings</span> on Windows environments without symlink support to avoid false negatives. - [#&#8203;13841](https://github.com/pytest-dev/pytest/issues/13841): `tox>=4` is now required when contributing to pytest. - [#&#8203;13625](https://github.com/pytest-dev/pytest/issues/13625): Added missing docstrings to `pytest_addoption()`, `pytest_configure()`, and `cacheshow()` functions in `cacheprovider.py`. #### Miscellaneous internal changes - [#&#8203;13830](https://github.com/pytest-dev/pytest/issues/13830): Configuration overrides (`-o`/`--override-ini`) are now processed during startup rather than during `config.getini() <pytest.Config.getini>`. </details> --- ### Configuration 📅 **Schedule**: (in timezone Europe/Berlin) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **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](https://github.com/renovatebot/renovate/discussions) if that's undesired. --- - [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate CLI](https://github.com/renovatebot/renovate). <!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0NC4xNzguNSIsInVwZGF0ZWRJblZlciI6IjQ0LjE3OC41IiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6W119-->
fix(deps): update security fixes [security]
Some checks failed
Test / test (pull_request) Failing after 21s
3cb6d7ff81
renovate-bot force-pushed renovate/major-security from 3cb6d7ff81
Some checks failed
Test / test (pull_request) Failing after 21s
to ab4e74f7b3
Some checks failed
Test / test (pull_request) Failing after 30s
2026-11-09 17:30:45 +00:00
Compare
fix(deps): a consistent langchain 1.x set for the security update
All checks were successful
Test / test (pull_request) Successful in 3m47s
6f6d60a0b8
Renovate's security update raised langchain, langchain-core, langchain-text-splitters and
langgraph to 1.x but left langchain-community 0.3.31, langchain-chroma 0.2.6,
langchain-huggingface 0.3.1 and langchain-ollama 0.3.10, which require langchain-core < 1.0;
its picks also contradict each other (langchain 1.3.9 needs langchain-core >= 1.4.6 and
langgraph >= 1.2.4, langchain-text-splitters 1.1.2 langchain-core >= 1.2.31), so pip could not
resolve the install and the test job failed. Pin the whole family to its current releases:
langchain 1.4.9, langchain-core 1.7.7, langchain-text-splitters 1.1.3, langgraph 1.2.20,
langchain-community 0.4.2, langchain-chroma 1.1.0, langchain-huggingface 1.2.3 and
langchain-ollama 1.1.0. pytest 9 needs pytest-asyncio 1.x (0.26 requires pytest < 9).
langchain 1.0 dropped the langchain.text_splitter module: import the splitter from
langchain_text_splitters.
Author
Collaborator

Edited/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.

### Edited/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.
dawid force-pushed renovate/major-security from 6f6d60a0b8
All checks were successful
Test / test (pull_request) Successful in 3m47s
to 539b23dd71
All checks were successful
Test / test (pull_request) Successful in 3m45s
2026-11-10 10:09:21 +00:00
Compare
dawid merged commit e8392b41be into main 2026-11-10 10:09:47 +00:00
dawid deleted branch renovate/major-security 2026-11-10 10:09:48 +00:00
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
dawid/rag-agent!112
No description provided.