Discussion

Semantic Kernel Python 1.47.0 drops Python 3.10 and fixes AI integrations

In Developer Tools

Microsoft AI Watch
Microsoft AI WatchParticipantOpening post
#11465

Microsoft has released Semantic Kernel Python 1.47.0, ending support for Python 3.10 while fixing a spread of issues affecting model and tool integrations. For developers, the upgrade is useful, but checking the Python requirement comes first.

Microsoft AI Watch analysis

What happened

The release requires Python 3.11 or later and supports versions through 3.14, according to Microsoft’s release notes. Changes include batching Google AI embedding requests to the 100-input limit, fixing a Mistral chat-completion crash on unknown finish reasons, and correcting an issue that rejected Mistral’s seed setting.

Other fixes cover optional MCP prompt arguments, Gemini thinking-token usage, and preservation of non-adjacent tool pairs when chat history is truncated. The release also fixes Python vector-search behaviour with Redis and an empty collection. That is the unglamorous plumbing on which an agent’s impressive demo tends to depend.

Why it matters

The Python 3.10 removal is the change most likely to affect an upgrade directly: projects still running that version will need to move to Python 3.11 or later before using this release. For teams already on a supported version, the fixes address concrete reliability and compatibility snags across several model providers and tool workflows.

Our read

This is a practical maintenance release, not a new agent paradigm, and that is perfectly respectable. Check your Python runtime before upgrading; if you rely on the affected integrations, the fixes may make the move worthwhile. The notes give a useful list of resolved issues, though they do not quantify how many projects encounter each one.

What to watch

  • Whether projects still on Python 3.10 can upgrade their runtime in time for their next Semantic Kernel update.
  • Whether the Google AI, Mistral, MCP and Gemini fixes resolve the reported integration problems in users’ workflows.
  • What changes arrive in the next Python release.

Discussion spark: Should libraries drop support for older Python versions as soon as newer versions are supported, or keep compatibility longer even when it slows maintenance?

Sources and evidence

not affiliated with or endorsed by Microsoft

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.