python-1.46.0
Semantic Kernel Python 1.46.0 introduces key security and bug fixes, including path-traversal protections in SessionsPythonTool, DNS-pinned HttpPlugin requests, redaction of logged function arguments, and various connector and agent improvements.
Why this signal matters
Semantic Kernel version 1.46.0 for Python delivers a substantial set of security fixes, agent runtime improvements, and vector store patches. Key security and boundary enhancements include rejecting dot-segment paths in SessionsPythonTool remote file uploads, pinning HttpPlugin requests to validated DNS addresses to counter rebinding, omitting function argument values from logs, and documenting prompt template/tool invocation trust boundaries. On the model and connector integration front, the update fixes tool formatting across Anthropic, Bedrock, Google AI, and Vertex AI; standardizes token usage metadata (CompletionUsage) for Mistral and Anthropic; handles edge cases in Qdrant, Chroma, and Cosmos DB; and introduces typing.Literal support in KernelJsonSchemaBuilder. AgentChat lifecycle handling and termination strategies also received stability fixes.
Actionable summary
Semantic Kernel Python release 1.46.0 brings critical security safeguards (rejecting dot-segment remote upload paths, pinning HttpPlugin to validated DNS addresses, and avoiding function argument value logging), alongside connector updates (Chroma, Weaviate, Cosmos, Qdrant), multi-provider tool format fixes, and schema validation enhancements.
- Agent usefulness
- 92/100
- Confidence
- 96%
- Canonical data
- JSON + Markdown
What builders should check
- Upgrade semantic-kernel to version 1.46.0 in Python environments.
- Verify that remote file upload paths in SessionsPythonTool do not rely on dot segments (e.g., '../').
- Review custom tool schemas relying on typing.Literal or Pydantic field descriptions to ensure proper serialization.
- Inspect logging pipelines if downstream operations previously relied on function argument values being emitted in standard logs.