Struct ConsecutiveFailureRecovery 

pub struct ConsecutiveFailureRecovery { /* private fields */ }

Recovery strategy that tracks consecutive failures per tool and stops after a limit.

This prevents runaway retry loops where the model keeps calling the same failing tool (e.g., execute_ssh_command failing due to auth, but the model retries indefinitely). After max_consecutive_failures for the same tool name, the strategy switches from Retry to Stop with a summary message.

Failure counters are tracked per session and reset when a different tool succeeds or a new session starts.

Implementations

impl ConsecutiveFailureRecovery

pub fn new(max_consecutive_failures: usize) -> ConsecutiveFailureRecovery

pub fn reset_failures(&self, session_id: &SessionId, tool_name: &str)

Reset failure count for a specific tool in a session. Call this when a tool succeeds to avoid false positives.

pub fn reset_session(&self, session_id: &SessionId)

Reset all failure counts for a session. Call this when a new turn starts.

Trait Implementations

impl ToolErrorRecovery for ConsecutiveFailureRecovery

fn on_error( &self, session_id: &SessionId, tool_names: &[String], _error: &AgentError, ) -> Result<ToolErrorAction, AgentError>