Struct LlmMessage 

pub struct LlmMessage {
    pub role: LlmMessageRole,
    pub content: LlmMessageContent,
    pub tool_calls: Option<Vec<ToolCall>>,
    pub tool_call_id: Option<String>,
    pub phase: Option<ExecutionPhase>,
    pub thinking: Option<String>,
    pub thinking_signature: Option<String>,
}

Message format for LLM calls (provider-agnostic)

Fields

role: LlmMessageRole

content: LlmMessageContent

tool_calls: Option<Vec<ToolCall>>

tool_call_id: Option<String>

phase: Option<ExecutionPhase>

Execution phase for assistant messages. Helps models distinguish between intermediate working commentary (Commentary) and completed answers (FinalAnswer) in multi-step tool-calling flows. Only set on assistant messages. Must be preserved when replaying conversation history.

thinking: Option<String>

Thinking content from extended thinking models (Anthropic Claude) Must be included in subsequent API calls when thinking is enabled

thinking_signature: Option<String>

Cryptographic signature for thinking content (Anthropic Claude) Required when sending thinking back in subsequent API calls

Implementations

impl LlmMessage

pub fn text(role: LlmMessageRole, content: impl Into<String>) -> LlmMessage

Create a message with text content

pub fn parts(role: LlmMessageRole, parts: Vec<LlmContentPart>) -> LlmMessage

Create a message with content parts (text, images, audio)

pub fn content_as_text(&self) -> String

Get content as plain text string (for simple cases)

pub fn prepend_text_prefix(&mut self, prefix: &str)

Prepend a prefix to the first text content.

Used by ReasonAtom to inject external actor identity (e.g. "[Alice] ") into user messages from external channels.

Trait Implementations

impl Clone for LlmMessage

fn clone(&self) -> LlmMessage

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LlmMessage

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

Formats the value using the given formatter.