Module model 

Model and effort selection for code-agent runs (the AI PR auto-fill path).

AgentModel is a small, curated set of selectable model tiers; Effort is how hard the agent should work; AgentOptions bundles the two for one run. All three are pure data with parse/id/label/next helpers so they drive the config layer, the CLI flags, and the TUI’s live cycle keys without any process or I/O.

Structs

AgentOptions

The model and effort selected for a single agent run.

Enums

AgentModel

A selectable model tier for a code agent. The variants currently encode the Claude tiers (the only supported agent); the id doubles as the CLI --model value — a stable alias that resolves to the latest model of that tier — so labels can track the current family without breaking selection.

Effort

How much effort the agent should spend on a draft. Claude has no native headless effort flag, so wt conveys effort as a one-line directive prepended to the prompt (see Effort::directive) — a safe, never-failing lever that shapes the model’s deliberation and can map to native reasoning controls per agent in the future.