Struct ModelMessage 

pub struct ModelMessage {
    pub role: ModelRole,
    pub content: String,
    pub content_parts: Vec<ContentPart>,
    pub tool_call_id: Option<String>,
    pub tool_name: Option<String>,
    pub tool_calls: Vec<ToolInvocation>,
    pub created_at: Option<u64>,
    pub thinking_content: Option<String>,
}

A single message in a model conversation.

Messages carry role, content, and optional tool-related metadata so the same type can represent system prompts, user input, assistant responses, tool calls, and tool results.

Fields

role: ModelRole

The conversational role of this message.

content: String

The text content of the message.

content_parts: Vec<ContentPart>

Multimodal content parts (text + images).

When non-empty, providers use these parts instead of the plain content field to build the wire-format message. This allows attaching images alongside text in user messages.

tool_call_id: Option<String>

For tool-result messages, the id of the tool call being answered.

tool_name: Option<String>

For tool-result messages, the name of the tool that produced this result.

tool_calls: Vec<ToolInvocation>

For assistant messages, the tool invocations requested by the model.

created_at: Option<u64>

Creation timestamp in milliseconds since Unix epoch (not serialized).

thinking_content: Option<String>

Optional thinking/reasoning content from the model.

Implementations

impl ModelMessage

pub fn system(content: impl Into<String>) -> Self

Creates a system-role message.

pub fn developer(content: impl Into<String>) -> Self

Creates a developer-role message (context block injected separately from the base system prompt for provider cache efficiency).

pub fn user(content: impl Into<String>) -> Self

Creates a user-role message.

pub fn user_multimodal( content: impl Into<String>, parts: Vec<ContentPart>, ) -> Self

Creates a user-role message with text and optional image attachments.

pub fn assistant(content: impl Into<String>) -> Self

Creates an assistant-role message without thinking content.

pub fn assistant_with_thinking( content: impl Into<String>, thinking: Option<String>, ) -> Self

Creates an assistant-role message with optional thinking content.

pub fn tool_result( tool_call_id: impl Into<String>, tool_name: impl Into<String>, content: impl Into<String>, ) -> Self

Creates a tool-result message responding to a specific tool call.

pub fn tool_result_with_parts( tool_call_id: impl Into<String>, tool_name: impl Into<String>, content: impl Into<String>, content_parts: Vec<ContentPart>, ) -> Self

Creates a tool-result message with optional multimodal content parts (e.g. images from view_image for vision-capable models).

pub fn assistant_tool_call(invocation: ToolInvocation) -> Self

Creates an assistant message that requests a single tool invocation.

pub fn assistant_tool_call_with_context( invocation: ToolInvocation, content: impl Into<String>, thinking: Option<String>, ) -> Self

Creates an assistant message that requests a tool invocation with accompanying text content and optional thinking.

pub fn assistant_tool_calls_with_context( invocations: Vec<ToolInvocation>, content: impl Into<String>, thinking: Option<String>, ) -> Self

Trait Implementations

impl Clone for ModelMessage

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelMessage

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for ModelMessage

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for ModelMessage

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.