Struct WorkflowSpawnInfo 

pub struct WorkflowSpawnInfo {

    pub agent_id: String,
    pub goal: String,
    pub role: String,
    pub isolation: String,
    pub context_inheritance: String,
    pub model_hint: Option<String>,
    pub trust: String,
    pub output_schema: Option<Value>,
    pub reducer: Option<String>,
    pub input_agent_ids: Vec<String>,
    pub judge_match: Option<JudgeMatch>,
    pub loop_max_iters: Option<usize>,
    pub classify_labels: Vec<String>,
    pub token_budget: Option<u64>,
}

Enough to run one spawned workflow node, carried to the SDK in the WorkflowBatchSpawned observation. Role/isolation/inheritance are canonical snake_case strings (serde names) so the host SDK can rebuild an agent run spec — the kernel generates these specs internally, so this is how the goal reaches the SDK that actually executes the node.

Fields

agent_id: String

goal: String

role: String

isolation: String

context_inheritance: String

model_hint: Option<String>

trust: String

W3 trust level ("trusted" | "quarantined") — the SDK runs quarantined nodes without privileges and crosses their output back only as a structured summary.

output_schema: Option<Value>

G3 structured output: the JSON Schema the node’s output must conform to, carried verbatim from WorkflowNode::output_schema. The SDK instructs the agent with it and validates + retries on its result. None when the node declared no schema. Additive ABI.

reducer: Option<String>

G2 deterministic compute: present only for a NodeKind::Reduce node — the name of the SDK-registered pure function the SDK runs (over input_agent_ids’ outputs) instead of an LLM agent. None for every ordinary node. Additive ABI.

input_agent_ids: Vec<String>

G2: the dependency agent ids whose outputs a NodeKind::Reduce node consumes (its depends_on, resolved to stable agent ids). Empty for non-reduce nodes. Additive ABI.

judge_match: Option<JudgeMatch>

Present only for a tournament judge spawn (A#2): the two entrant agent ids whose outputs this judge must compare. The SDK looks up those entrants’ produced candidates, runs the judge, and reports the winner in the result’s tournament_winner. None for every ordinary (entrant / spawn / loop / classify) node. Additive ABI: omitted on the wire when None.

loop_max_iters: Option<usize>

Present only for a NodeKind::Loop iteration spawn (A#2 v2): the loop’s max_iters. It both marks the spawn as a loop iteration — so the SDK knows to solicit and report a loop_continue stop signal from the agent — and gives the cap for the agent’s prompt. None for every non-loop node. Mirrors how reducer / judge_match distinguish reduce / judge spawns. Additive ABI: omitted on the wire when None.

classify_labels: Vec<String>

Present only for a NodeKind::Classify spawn (A#2): the branch labels the classifier must choose among. Non-empty marks the spawn as a classifier — the SDK instructs the agent to pick exactly one label and reports it in the result’s classify_branch. Empty for every non-classify node. Additive ABI: omitted on the wire when empty.

token_budget: Option<u64>

M4/G5: the node’s per-node cumulative token cap, if set. The SDK sets the child run’s max_total_tokens to this so the node self-terminates at the cap. Additive ABI: omitted when None.

Trait Implementations

impl Clone for WorkflowSpawnInfo

fn clone(&self) -> WorkflowSpawnInfo

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for WorkflowSpawnInfo

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for WorkflowSpawnInfo

fn deserialize<__D>( __deserializer: __D, ) -> Result<WorkflowSpawnInfo, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for WorkflowSpawnInfo

impl PartialEq for WorkflowSpawnInfo

fn eq(&self, other: &WorkflowSpawnInfo) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for WorkflowSpawnInfo

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

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for WorkflowSpawnInfo