fix(deps): upgrade dependencies to address security vulnerabilities

[dbezic (dbezic)]

Summary

Upgrade the following dependencies to patched versions:

Dependency	Previous	Updated
mcp	1.26.0	1.28.1
langsmith	0.7.31	0.8.18
pyasn1	0.6.3	0.6.4
python-multipart	0.0.26	0.0.30
langchain-core	1.2.28	1.3.3
PyJWT	2.12.0	2.14.0
anyio	4.12.1	4.14.2

• Update centralized minimum-version constraints to prevent resolution to affected versions.
• Declare security-critical transitive dependencies in the affected library manifests
  so published wheels enforce the same minimums, not only workspace installs.
• Cover LangChain, OpenAI, Semantic Kernel, Agent Framework, and Google ADK
  integrations without adding framework dependencies to unrelated core packages.
• Publish the PyJWT floor from tooling and notifications as well as runtime,
  protecting consumers that install those packages independently.
• Regenerate the lockfile, including the required transitive dependency
  langchain-protocol 0.0.19.
• Preserve all unrelated package versions and existing dependency overrides.
• Use PyJWT 2.14.0 to address the six additional advisories reported by Dependency
  Review against 2.13.0: GHSA-ffc3-869f-jxw9, GHSA-r6x4-923q-g947,
  GHSA-p4g4-x82p-q773, GHSA-9v7f-9g4p-ffgj, GHSA-w2cx-738m-mc7w,
  and GHSA-9j54-fg26-wv3r.

Validation

• 511 targeted tests passed for the initial dependency upgrades.
• 153 targeted JWT, runtime, tooling, token-cache, and dependency tests passed
  after upgrading PyJWT to 2.14.0.
• 49 packaging tests passed, including 14 regression cases that build actual
  wheels and check each distribution's published runtime dependency requirements.
• New wheel-metadata tests reproduced the publication gaps before the fixes and
  verify that unrelated framework dependencies are not added to lightweight packages.
• Lockfile consistency and centralized dependency constraint checks passed.

[github-actions]

⚠️ Deprecation Warning: The deny-licenses option is deprecated for possible removal in the next major release. For more information, see issue 997.

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

OpenSSF Scorecard

Package	Version	Score	Details
pip/pyjwt	2.14.0	Unknown	Unknown
pip/python-multipart	0.0.30	Unknown	Unknown
pip/anyio	4.14.2	Unknown	Unknown
pip/langchain-core	1.3.3	Unknown	Unknown
pip/langchain-protocol	0.0.19	Unknown	Unknown
pip/langsmith	0.8.18	Unknown	Unknown
pip/mcp	1.28.1	Unknown	Unknown
pip/pyasn1	0.6.4	Unknown	Unknown

Scanned Files

• uv.lock

[Copilot]

Copilot review overview

🟡 Changes recommended

Several transitive security floors are not propagated into published package metadata.

Review effort: Balanced
Findings: 1

Open (1)

• Security constraints missing from published package metadata · New

What changed in this PR

Upgrades vulnerable dependencies and updates centralized resolution constraints.

Changes:

• Raises minimum versions for seven dependencies.
• Regenerates the lockfile with patched versions and langchain-protocol.

File	Description
pyproject.toml	Updates centralized dependency floors.
uv.lock	Locks patched releases and transitive dependencies.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🟡 Changes recommended

Several transitive security floors are not propagated to published package metadata.

Review effort: Balanced
Findings: 2

Open (2)

• Dependency floors are not propagated to published package metadata · New
• Security constraints missing from published package metadata

[Copilot]

Copilot review overview

🟢 Approval recommended

The dependency floors, package metadata, lockfile, and regression coverage are consistent with the stated security remediation.

Review effort: Balanced
Findings: 2

Open (2)

• Dependency floors are not propagated to published package metadata
• Security constraints missing from published package metadata