Struct TurnOutcome 

pub struct TurnOutcome {
    pub reply: String,
    pub events: Vec<AgentEvent>,
    pub citations: Vec<Citation>,
}

The outcome of running one knowledge-grounded turn through the agent.

Carries the final assistant text plus every AgentEvent the engine emitted during the run — so callers (and tests) can inspect exactly what happened: which tools ran, what they returned, how many iterations.

Fields

reply: String

The agent’s final natural-language reply (the last assistant turn with no pending tool calls). Empty string if the agent produced no text.

events: Vec<AgentEvent>

Every event the engine emitted, in order. Inspect for AgentEvent::ToolCallStart / AgentEvent::ToolCallComplete to prove a knowledge search happened.

citations: Vec<Citation>

The sources that grounded this turn, deduplicated by id and capped at MAX_CITATIONS. Collected from the documents the turn actually retrieved — the engine’s auto-injected [Relevant knowledge] context (mirrored by the runtime) plus every knowledge_search tool result. Empty when nothing was retrieved.

Implementations

impl TurnOutcome

pub fn invoked_tool(&self, tool_name: &str) -> bool

true if the agent invoked a tool named tool_name during the turn.

pub fn tool_result(&self, tool_name: &str) -> Option<&str>

The completed result text of the first call to tool_name, if any. Sourced from AgentEvent::ToolCallComplete (truncated to ~500 chars by the engine), so a test can assert the tool returned the seeded doc.

Trait Implementations

impl Clone for TurnOutcome

fn clone(&self) -> TurnOutcome

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TurnOutcome

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

Formats the value using the given formatter.