Enum ServerEvent 

pub enum ServerEvent {
    AgentStatus {
        agent_id: String,
        run_id: String,
        status: String,
        stage: String,
        iteration: usize,
        tool_calls: usize,
        accepts_messages: bool,
        wait_reason: Option<WaitReason>,
    },
    ContextUpdate {
        agent_id: String,
        run_id: String,
        total_tokens: usize,
        max_tokens: usize,
    },
    Log {
        agent_id: String,
        run_id: String,
        line: String,
    },
    InteractionNeeded {
        agent_id: String,
        run_id: String,
        request: Value,
    },
    AgentSpawned {
        agent_id: String,
        run_id: String,
        parent_id: Option<String>,
        blueprint: String,
    },
    AgentCompleted {
        agent_id: String,
        run_id: String,
        status: String,
        result: Option<String>,
        final_output: Option<FinalOutputResp>,
    },
    Tokens {
        agent_id: String,
        run_id: String,
        prompt_tokens: usize,
        completion_tokens: usize,
        cached_tokens: usize,
        cache_write_tokens: usize,
    },
    World {
        event: Value,
    },
}

Events broadcast to WebSocket subscribers.

Variants

AgentStatus

Where a run stands, re-sent whenever any of it changes.

Fields

agent_id: String

The agent’s live id in the world.

run_id: String

The durable run id, and what a per-run subscription filters on.

status: String

The run’s status, as RunStatus renders it.

stage: String

The stage it is in, by name.

iteration: usize

Inference turns taken in that stage, reset on entering a new one.

tool_calls: usize

Tool calls made across the whole run.

accepts_messages: bool

Whether a client may send this run a mid-run message. False for a stage that declared accepts_messages = false, and for a run that has finished.

wait_reason: Option<WaitReason>

Why the run is parked, when it is.

Omitted for a run that is moving. Without it a subscriber sees a run turn waiting or paused and has to fetch the run to learn whether somebody is needed, which is the guess this vocabulary exists to remove.

ContextUpdate

How full the run’s context window is, after a turn changed it.

Fields

agent_id: String

The agent’s live id in the world.

run_id: String

The durable run id.

total_tokens: usize

Tokens held across every region.

max_tokens: usize

The whole window’s budget.

Log

One log line, as also written to the stage’s log files.

Fields

agent_id: String

The agent’s live id in the world.

run_id: String

The durable run id.

line: String

The line, without a trailing newline. Long lines are truncated for the broadcast; the on-disk stage log keeps the whole thing.

InteractionNeeded

The run is blocked on a person, and this is what it is asking.

Fields

agent_id: String

The agent’s live id in the world.

run_id: String

The durable run id.

request: Value

The prompt, forwarded as the runtime serialized it, so a new kind of request needs no server release.

AgentSpawned

A run started, including one spawned as a child of another.

Fields

agent_id: String

The agent’s live id in the world.

run_id: String

The durable run id.

parent_id: Option<String>

The parent’s agent id when this is a sub-agent, None at the root.

blueprint: String

The blueprint it was spawned from.

AgentCompleted

A run reached a terminal status.

Fields

agent_id: String

The agent’s live id in the world.

run_id: String

The durable run id.

status: String

The terminal status, as RunStatus renders it.

result: Option<String>

The run’s error, if it failed. Named result since before a run could produce one; kept for the consumers that read it.

final_output: Option<FinalOutputResp>

What the run handed back. This is the answer.

Tokens

Running token totals for the whole run, after an inference landed.

Fields

agent_id: String

The agent’s live id in the world.

run_id: String

The durable run id.

prompt_tokens: usize

Input tokens billed so far.

completion_tokens: usize

Output tokens billed so far.

cached_tokens: usize

Input tokens served from the provider’s prompt cache, counted within prompt_tokens rather than on top of it.

cache_write_tokens: usize

Tokens written into the provider’s prompt cache.

World

A world event with no dedicated WebSocket translation (stage transitions, tool call start/finish, and whatever the runtime adds next), forwarded verbatim. event is the runtime’s own serde-tagged WorldEvent JSON, so clients get new event kinds without a server release.

Fields

event: Value

The runtime’s own serde-tagged event JSON, forwarded as-is.

Implementations

impl ServerEvent

pub fn run_id(&self) -> &str

The run id this event belongs to, for per-run subscription filtering. World events read it from the wrapped JSON (every runtime event carries one; an absent field filters as the empty string).

Trait Implementations

impl Clone for ServerEvent

fn clone(&self) -> ServerEvent

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ServerEvent

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

Formats the value using the given formatter.

impl Serialize for ServerEvent

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

where __S: Serializer,

Serialize this value into the given Serde serializer.