Struct ModelResponse 

pub struct ModelResponse {
    pub content: String,
    pub usage: Option<TokenUsage>,
    pub model_name: String,
    pub thinking: Option<String>,
    pub tool_calls: Option<Vec<ToolCall>>,
    pub stop_reason: Option<FinishReason>,
    pub thinking_signature: Option<String>,
}

Response from a model

Fields

content: String

The actual response text

usage: Option<TokenUsage>

Usage statistics if available

model_name: String

Model that generated the response

thinking: Option<String>

Thinking/reasoning content (for models that expose their thought process)

tool_calls: Option<Vec<ToolCall>>

Tool calls from the model (Ollama native function calling)

stop_reason: Option<FinishReason>

Why generation stopped, when the provider reported it. None if the provider didn’t say (or the adapter doesn’t yet map it).

thinking_signature: Option<String>

Anthropic thinking-block signature (encrypted server state). Set only by AnthropicAdapter; other adapters leave it None. The agent loop’s commit step copies this onto the resulting assistant ChatMessage::thinking_signature so it round-trips on the next turn — without this, multi-turn Claude conversations with extended thinking 400 with invalid_request_error.

Trait Implementations

impl Clone for ModelResponse

fn clone(&self) -> ModelResponse

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelResponse

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

Formats the value using the given formatter.