Struct BeforeTurnEnd 

pub struct BeforeTurnEnd {
    pub stop_reason: StopReason,
    pub continues_so_far: u32,
    pub voluntary: bool,
    pub feedback: Vec<ContentBlock>,
}

before turn-end: the turn’s only voluntary exit point. Defaults to Break — “do nothing” = let it stop.

A hook returning HookControl::Continue extends the turn: it injects Self::feedback into history (appended as a user message when committed), does not end the turn, and loops back to the top for another round. Continue only takes effect when Self::voluntary is true — involuntary stops (Refusal / MaxTokens / Cancelled / MaxTurnRequests) ignore it; otherwise the hook could bypass the request cap and extend indefinitely.

Fields

stop_reason: StopReason

The reason this turn stopped.

continues_so_far: u32

How many times this turn has been extended by a hook (the hook decides when to stop; a hard cap in the loop provides a safety net).

voluntary: bool

Whether the stop is voluntary (LLM said EndTurn or returned empty tool_use). Continue only takes effect when voluntary.

feedback: Vec<ContentBlock>

Feedback to inject into history when continuing the turn. apply_verdict fills this from the verdict’s additional_context; internal Rust hooks push directly. On finalization, the loop appends it as a user message.

Trait Implementations

impl Clone for BeforeTurnEnd

fn clone(&self) -> BeforeTurnEnd

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for BeforeTurnEnd

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

Formats the value using the given formatter.

impl HookStep for BeforeTurnEnd

fn event_name(&self) -> &'static str

Event name (snake_case). Used in envelope headers and matchers.

fn to_envelope(&self) -> Value

Projects the step into an input envelope — fed to command stdin / prompt templates. Contains a common header plus step-specific fields.

fn apply_verdict( &mut self, verdict: &Value, ) -> Result<HookControl, VerdictError>

Apply the handler’s output verdict (JSON) back to this step: parse the common control / additional_context fields, then handle the step-specific “fill output” fields. Returns a control directive.