llm -m gpt-5 -o reasoning_effort minimal --save gm saves bad YAML

[simonw]

Just noticed this:

llm -m gpt-5 -o reasoning_effort minimal --save gm

Then:

llm -t gm hi

Outputs:

Error: Invalid YAML: could not determine a constructor for the tag 'tag:yaml.org,2002:python/object/apply:llm.default_plugins.openai_models.ReasoningEffortEnum'
  in "<unicode string>", line 3, column 23:
        reasoning_effort: !!python/object/apply:llm.defaul ...     

Because the YAML file that was saved as gm.yaml is bad:

model: gpt-5
options:
    reasoning_effort: !!python/object/apply:llm.default_plugins.openai_models.ReasoningEffortEnum
    - minimal

[simonw]

Here's why:

llm/llm/cli.py

Lines 656 to 664 in e15e1ad

	path.write_text(
	yaml.dump(
	to_save,
	indent=4,
	default_flow_style=False,
	sort_keys=False,
	),
	"utf-8",
	)

safe_dump() should fix this.

[simonw]

Manually tested and it works now.