Struct DebugMessage 

#[non_exhaustive]
pub struct DebugMessage {
    pub text: Option<String>,
    pub thinking: Option<String>,
    pub thinking_redacted: bool,
    pub tool_calls: Vec<DebugToolCall>,
    pub tool_results: Vec<DebugToolResult>,
    pub stop_reason: Option<String>,
    pub input_tokens: Option<u64>,
    pub output_tokens: Option<u64>,
}

A single assistant turn captured during a verbose invocation.

Each DebugMessage represents one assistant response, which may contain free-form text, tool calls, or both.

Examples

use ironflow_core::prelude::*;

let provider = ClaudeCodeProvider::new();
let result = Agent::new()
    .prompt("List files in src/")
    .verbose()
    .run(&provider)
    .await?;

if let Some(messages) = result.debug_messages() {
    for msg in messages {
        println!("{msg}");
    }
}

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

text: Option<String>

Free-form text produced by the assistant in this turn, if any.

thinking: Option<String>

Extended thinking blocks produced by the model in this turn.

Available only when the model emits thinking content blocks (Opus 4.7 adaptive thinking, Claude 3.7+ extended thinking, etc.). The blocks are joined in arrival order.

thinking_redacted: bool

true when the model emitted a thinking content block but the text was redacted (only a signature is provided).

Opus 4.7 adaptive thinking and the display: "omitted" setting both produce signature-only thinking blocks: the model proves it reasoned without exposing the chain of thought. The UI should still show a badge so the user knows thinking happened.

tool_calls: Vec<DebugToolCall>

Tool calls made by the assistant in this turn.

tool_results: Vec<DebugToolResult>

Tool results received from the user/runtime for the preceding tool calls.

In the Claude stream-json format, tool results come as "type":"user" messages whose content is a list of tool_result blocks. We attach them to the turn that emitted the matching tool_use so the timeline stays compact.

stop_reason: Option<String>

The model’s stop reason for this turn (e.g. "end_turn", "tool_use").

input_tokens: Option<u64>

Input tokens consumed by this turn, if reported.

output_tokens: Option<u64>

Output tokens generated by this turn, if reported.

Trait Implementations

impl Clone for DebugMessage

fn clone(&self) -> DebugMessage

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DebugMessage

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for DebugMessage

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Display for DebugMessage

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

Formats the value using the given formatter.

impl Serialize for DebugMessage

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

where __S: Serializer,

Serialize this value into the given Serde serializer.