Struct LoopGuard 

pub struct LoopGuard { /* private fields */ }

Implementations

impl LoopGuard

pub fn record_attempt(&mut self, tool: &str, args: &Value) -> AttemptDecision

pub fn record_outcome(&mut self, tool: &str, ok: bool) -> OutcomeDecision

pub fn reset_failures(&mut self)

Clear consecutive-failure counters so a granted continuation (e.g. a long-horizon “change approach” nudge issued after a OutcomeDecision::Halt) doesn’t immediately re-halt on the same tool. Identical-call counts are left intact, so blindly repeating the exact same call is still blocked.

pub fn note_state_changed(&mut self)

Clear identical-call counts after the workspace changed (a state-mutating tool succeeded). Re-running the exact same verify/read call after an intervening edit is legitimate work — not a redundant loop — so it must not stay blocked. Without this, an iterative edit → re-run same test loop trips the 3× block and the model is forced into meaningless arg-reordering to dodge the guard (defeating its purpose). Hammering the same call with no intervening change still blocks, because nothing calls this between those identical attempts.

pub fn is_state_mutating_tool(tool: &str) -> bool

Whether a tool’s success means the workspace materially changed, so the identical-call counter should be cleared (see Self::note_state_changed).

Trait Implementations

impl Debug for LoopGuard

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

Formats the value using the given formatter.

impl Default for LoopGuard

fn default() -> LoopGuard

Returns the “default value” for a type.