Python: Fix Ollama settings issue with structured outputs - #12355

[xyang2013]

Issue heading:
Python: New Feature: OllamaChatPromptExecutionSettings does not support structured outputs #12355

Link:
#12355

[moonbox3]

Hi @xyang2013, thanks for creating a PR for this. You'll want to make sure you properly set up your dev env so that you can install the pre-commit hooks. This will ensure that all checks have passed locally - like ruff and others. You can follow our setup guide here: https://github.com/microsoft/semantic-kernel/blob/main/python/DEV_SETUP.md#using-uv.

There are also some unit tests failing: https://github.com/microsoft/semantic-kernel/actions/runs/15460655067/job/43533345701?pr=12386#step:5:604. Please have a look, thank you.

[xyang2013]

Thanks, @moonbox3. I’ll take a look at the documentation later.

Does that mean I need to update not only the Ollama settings file but also the associated unit tests?

[moonbox3]

Thanks, @moonbox3. I’ll take a look at the documentation later.

Does that mean I need to update not only the Ollama settings file but also the associated unit tests?

If unit tests are not passing because of this change, then yes, they will need an update.

[xyang2013]

Thanks @moonbox3. I spent some time setting up the environment and ran the following command:

uv run pytest tests/integration

It showed up many unrelated errors (not related to Ollama). Should I just focus on the ones related to the change?

[moonbox3]

Thanks @moonbox3. I spent some time setting up the environment and ran the following command:

uv run pytest tests/integration

It showed up many unrelated errors (not related to Ollama). Should I just focus on the ones related to the change?

Please run the unit tests, not the integration tests.

[moonbox3]

Hi @xyang2013, haven't seen a response from you in a bit. I've got a PR open that fixes this - we should still allow for Literal["json"] therefore:

format: Literal["json"] | dict[str, Any] | None = None

No tests are breaking.