Struct DashboardAgent 

pub struct DashboardAgent {

    pub id: String,
    pub blueprint_name: String,
    pub stage: String,
    pub stage_index: usize,
    pub num_stages: usize,
    pub status: AgentDisplayStatus,
    pub tokens_in: usize,
    pub tokens_out: usize,
    pub cached_tokens: usize,
    pub iteration: usize,
    pub waiting_prompt: Option<String>,
    pub wait_reason: Option<WaitReason>,
    pub pending_request: Option<InteractionRequest>,
    pub last_answered_request_id: Option<String>,
    pub context_snapshot: Option<Arc<ContextSnapshot>>,
    pub stages: Vec<StageRecord>,
    pub workdir: String,
    pub task: String,
    pub title: Option<String>,
    pub model: Option<String>,
    pub parent_id: Option<String>,
    pub depth: usize,
    pub started_at: i64,
    pub last_progress_at: Option<i64>,
    pub active_until: Option<i64>,
    pub waiting_secs: u64,
    pub accepts_messages: bool,
    pub taint_summary: Vec<(String, String)>,
    /* private fields */
}

An agent displayed in the dashboard.

Fields

id: String

The run id, which is also what every action against this row quotes.

blueprint_name: String

The blueprint’s name, as the manifest declares it.

stage: String

The stage the run is in, by name.

stage_index: usize

That stage’s position in the blueprint’s list.

num_stages: usize

How many stages the blueprint has, so the pair renders as “3 of 7”.

status: AgentDisplayStatus

What the row shows, including states no other status enum has.

tokens_in: usize

Cumulative prompt (input) tokens for background runs.

tokens_out: usize

Cumulative completion (output) tokens for background runs.

cached_tokens: usize

Cumulative tokens read from provider cache.

iteration: usize

Inference turns taken in the current stage.

waiting_prompt: Option<String>

The question a waiting run is asking, in one line, for the list row.

wait_reason: Option<WaitReason>

Why a waiting run is parked, when meta.json says.

WAITING on its own reads as “go and answer it”, which is wrong for a parent whose fan-out workers are still churning. None on a run that is not parked, and on one written by a build from before the field existed, which is why the row falls back to the bare status rather than assuming.

pending_request: Option<InteractionRequest>

Full structured interaction request (populated for WaitingInput agents)

last_answered_request_id: Option<String>

The request_id we most recently submitted a response for, used to suppress re-showing the same prompt before the worker has consumed the response.

context_snapshot: Option<Arc<ContextSnapshot>>

Live context window snapshot from context.json (background workers only) Shared, not owned: the live snapshot comes out of the sync tick’s stat-gated cache, and cloning a full context window per tick was the churn that cache exists to remove.

stages: Vec<StageRecord>

Per-stage records from stages.json

workdir: String

Working directory the agent ran in

task: String

Original task prompt

title: Option<String>

Auto-generated short title (None until the worker generates it).

model: Option<String>

Original model override

parent_id: Option<String>

Parent agent ID (if this is a sub-agent)

depth: usize

Depth in the sub-agent tree (0 = root)

started_at: i64

Unix timestamp when the run started (for elapsed display)

last_progress_at: Option<i64>

Unix timestamp of the run’s last recorded progress (None before the first progress mark). Drives the recent-activity sort.

active_until: Option<i64>

Frozen wall-clock time (Unix seconds) when the agent entered a waiting state. Used to prevent the elapsed timer from incrementing while waiting for input.

waiting_secs: u64

Total seconds spent waiting for user input across all completed waits. Subtracted from elapsed to show only actual running time.

accepts_messages: bool

Whether the current stage accepts mid-run user messages

taint_summary: Vec<(String, String)>

Per-region taint levels (region_name, taint_level_string). Empty when taint tracking is disabled or not yet populated.

Trait Implementations

impl Clone for DashboardAgent

fn clone(&self) -> DashboardAgent

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DashboardAgent

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

Formats the value using the given formatter.