Struct StopCheckContext 

pub struct StopCheckContext<'a> {
    pub rounds: usize,
    pub total_input_tokens: usize,
    pub total_output_tokens: usize,
    pub consecutive_errors: usize,
    pub elapsed: Duration,
    pub last_tool_calls: &'a [ToolCall],
    pub last_text: &'a str,
    pub tool_call_history: &'a VecDeque<Vec<String>>,
    pub run_ctx: &'a RunContext,
}

Snapshot of loop state provided to stop checks.

Fields

rounds: usize

Number of completed tool-call rounds.

total_input_tokens: usize

Cumulative input tokens across all LLM calls.

total_output_tokens: usize

Cumulative output tokens across all LLM calls.

consecutive_errors: usize

Number of consecutive rounds where all tools failed.

elapsed: Duration

Time elapsed since the loop started.

last_tool_calls: &'a [ToolCall]

Tool calls from the most recent LLM response.

last_text: &'a str

Text from the most recent LLM response.

tool_call_history: &'a VecDeque<Vec<String>>

History of tool call names per round (most recent last), for loop detection.

run_ctx: &'a RunContext

The current run context, providing access to conversation history and state.

Custom stop conditions can inspect messages for patterns or examine the accumulated state via rebuild_state().

Implementations

impl<'a> StopCheckContext<'a>

pub fn as_policy_input(&'a self) -> StopPolicyInput<'a>

Convert legacy context shape to canonical stop-policy input.