moo.agent.tests.test_llm_client

Tests for moo/agent/llm_client.py.

Covers:

• make_agent: returns a PydanticAI Agent per provider

• call_llm: runs the agent and returns the validated AgentResponse

• _model_settings: provider-specific sampling-kwarg assembly

• _patch_reasoning_content: LM Studio reasoning_content -> content shim

Functions

test_call_llm_lm_studio_config()	The lm_studio config branch still produces a validated response.
test_call_llm_returns_agent_response_and_tool_calls()	call_llm runs the agent and returns (validated AgentResponse, tool_calls).
test_make_agent_anthropic(monkeypatch)
test_make_agent_lm_studio()
test_make_agent_lm_studio_default_base_url()
test_model_settings_anthropic_no_extra_body()	Non-lm_studio providers omit extra_body and unset sampling params.
test_model_settings_lm_studio_sampling()	LM Studio gets repeat_penalty/min_p/reasoning_effort in extra_body.
test_patch_reasoning_content_leaves_content_when_present()	When content is already populated, reasoning_content is ignored.
test_patch_reasoning_content_promotes_when_content_empty()	When content is empty, reasoning_content is copied into content.