Struct NonStreamingResponse 

pub struct NonStreamingResponse {
    pub message: Message,
    pub stop_reason: StreamStopReason,
    pub usage: Option<Usage>,
}

A completed, non-streaming LLM response.

The typed counterpart to a streamed response: instead of a sequence of StreamEvents, the complete assistant Message, the reason generation stopped, and the token Usage are delivered in one shot. Each provider builds this from its own native JSON envelope, so callers never see provider-specific shapes — exactly mirroring how the streaming path emits typed events.

Produced by create_message and create_message_with_options.

Fields

message: Message

The fully assembled assistant message.

Contains the same MessagePart sequence a stream would accumulate — text blocks, tool calls, etc. Built by the provider from its native response shape.

stop_reason: StreamStopReason

Why the model stopped generating.

Mapped by the provider from its native finish/stop field. Drives the agent loop’s decision to continue to tool execution or end the turn.

usage: Option<Usage>

Token counts for the request, as reported by the provider.

Extracted from the provider’s native usage field (usage on OpenAI and Anthropic, usageMetadata on Gemini). None when the provider omits usage from the response — symmetric with the Option<Usage> carried by the final MessageDelta event on the streaming path.

Trait Implementations

impl Clone for NonStreamingResponse

fn clone(&self) -> NonStreamingResponse

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for NonStreamingResponse

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

Formats the value using the given formatter.