Struct LlmInferenceHttpRequestStartRequest 

pub struct LlmInferenceHttpRequestStartRequest {
    pub agent_id: Option<String>,
    pub headers: HashMap<String, Vec<String>>,
    pub interaction_type: Option<String>,
    pub method: String,
    pub parent_agent_id: Option<String>,
    pub request_id: RequestId,
    pub session_id: Option<SessionId>,
    pub transport: Option<LlmInferenceHttpRequestStartTransport>,
    pub url: String,
}

The head of an outbound model-layer HTTP request.

Fields

agent_id: Option<String>

Stable per-agent-instance id attributing this request to a specific agent trajectory. Present when the request originates from an agent turn; absent for requests issued outside any agent context (e.g. some SDK callers). A request with an agentId but no parentAgentId is a root-agent request; one carrying both is a subagent request. Sourced from the runtime’s per-request agent context and surfaced on the envelope independently of transport, so it is available for both first-party (CAPI) and BYOK/custom-provider requests; on the CAPI transport the runtime derives the upstream X-Agent-Task-Id header from this same context. Consumers routing each provider call to a training trajectory should key on this rather than on lifecycle events, since it is available on the request path before sampling.

headers: HashMap<String, Vec<String>>

interaction_type: Option<String>

Coarse classification of the interaction that produced this request. Open string for forward-compatibility; known values include conversation-agent, conversation-subagent, conversation-sampling, conversation-background, conversation-compaction, and conversation-user. Absent when the runtime did not classify the request. Comes from the runtime’s per-request agent context independently of transport; on the CAPI transport the runtime derives the upstream X-Interaction-Type header from this same context.

method: String

HTTP method, e.g. GET, POST.

parent_agent_id: Option<String>

Id of the parent agent that spawned the agent issuing this request. Present only for subagent requests; absent for root-agent requests and non-agent requests. Combined with agentId, this lets consumers attribute a call to a child trajectory versus the root. Like agentId, it comes from the runtime’s per-request agent context independently of transport; on the CAPI transport the runtime derives the upstream X-Parent-Agent-Id header from this same context.

request_id: RequestId

Opaque runtime-minted id, unique per in-flight request. The SDK uses this to correlate httpRequestChunk frames and to address its httpResponseStart / httpResponseChunk replies back to the runtime.

session_id: Option<SessionId>

Id of the runtime session that triggered this request, when one is in scope. Absent for requests issued outside any session (e.g. startup model-catalog or capability resolution). This is a payload field — not a dispatch key — because the client-global API is registered process-wide rather than per session.

transport: Option<LlmInferenceHttpRequestStartTransport>

Transport the runtime would otherwise use for this request. http (the default when absent) covers plain HTTP and SSE responses; websocket indicates a full-duplex message channel where each body chunk maps to one WebSocket message and the binary flag distinguishes text from binary frames. The SDK consumer uses this to decide whether to service the request with an HTTP client or a WebSocket client. It is the one piece of request metadata the consumer cannot reliably infer from the URL or headers alone.

url: String

Absolute request URL.

Trait Implementations

impl Clone for LlmInferenceHttpRequestStartRequest

fn clone(&self) -> LlmInferenceHttpRequestStartRequest

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LlmInferenceHttpRequestStartRequest

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

Formats the value using the given formatter.

impl Default for LlmInferenceHttpRequestStartRequest

fn default() -> LlmInferenceHttpRequestStartRequest

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for LlmInferenceHttpRequestStartRequest

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for LlmInferenceHttpRequestStartRequest

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

where __S: Serializer,

Serialize this value into the given Serde serializer.