Struct UtilityContext 

pub struct UtilityContext {
    pub tool_calls_this_turn: usize,
    pub tokens_consumed: usize,
    pub token_budget: usize,
    pub user_requested: bool,
    pub mandated_retry: bool,
}

Context required to compute utility — provided by the agent loop.

Fields

tool_calls_this_turn: usize

Number of tool calls already dispatched in the current LLM turn.

tokens_consumed: usize

Tokens consumed so far in this turn.

token_budget: usize

Token budget for the current turn. 0 = budget unknown (cost component treated as 0).

user_requested: bool

True when the user explicitly requested tool invocation — either via a /tool slash command or when the user message contains an unambiguous tool-invocation phrase detected by has_explicit_tool_request. Must NOT be set from LLM call content or tool outputs.

mandated_retry: bool

True when this exact call is the mandated retry the Retrieve rule itself asked for: the same (tool_id, params) call was vetoed by rule 8 earlier this turn, and the injected system hint explicitly instructed the LLM to call it again with the same arguments. Set via UtilityScorer::take_mandated_retry.

When true, recommend_action must not re-veto the retry through the redundancy (Respond) rule — doing so fabricates a rejection despite the model correctly complying with the gate’s own hint, stalling tools like find_path/list_directory in a doomed Retrieve-then-redundant-Respond cycle (#5719).

Trait Implementations

impl Clone for UtilityContext

fn clone(&self) -> UtilityContext

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for UtilityContext

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.