Module llm

Enums§

Effort: How hard the model should work on a request — the Anthropic output_config.effort scale.
EffortShape: What form of reasoning control a model accepts.
EffortSource: Where the effort in force came from, so a surface can say so instead of showing a bare level the user cannot account for.

LlmClient: Trait for LLM providers (Anthropic, OpenAI, Ollama). Shared between mur-core and mur-commander.

anthropic_base_url: Resolve the Anthropic API base URL from ANTHROPIC_BASE_URL env, with a trailing slash stripped. Falls back to ANTHROPIC_DEFAULT_BASE_URL.
effective_effort: The effort actually in force for model, and where it came from.
effort_shape: Which reasoning control model accepts.
is_reasoning_model: Check if a model name matches recommended reasoning models for session analysis.
openai_reasoning_effort: The reasoning_effort value to send for an OpenAI-compatible model, or None when the model takes no such parameter.
supported_effort: The effort level to actually send for model, or None when the model takes no effort parameter at all.