Struct RequestHandlerBuilder 

pub struct RequestHandlerBuilder { /* private fields */ }

Fluent builder for RequestHandler.

Required

• executor: Any AgentExecutor implementation (passed as a concrete type; the builder erases it to Arc<dyn AgentExecutor> during build).

Optional (with defaults)

• task_store: defaults to InMemoryTaskStore.
• push_config_store: defaults to InMemoryPushConfigStore.
• push_sender: defaults to None.
• interceptors: defaults to an empty chain.
• agent_card: defaults to None.
• tenant_resolver: defaults to None (no tenant resolution).
• tenant_config: defaults to None (no per-tenant limits).
• max_concurrent_streams: defaults to DEFAULT_MAX_CONCURRENT_STREAMS (1024).
• executor_timeout: defaults to None (unbounded — recommended to set explicitly; see with_executor_timeout).

Implementations

impl RequestHandlerBuilder

pub fn new(executor: impl AgentExecutor) -> RequestHandlerBuilder

Creates a new builder with the given executor.

The executor is type-erased to Arc<dyn AgentExecutor>.

pub fn with_task_store( self, store: impl TaskStore + 'static, ) -> RequestHandlerBuilder

Sets a custom task store.

pub fn with_task_store_arc( self, store: Arc<dyn TaskStore>, ) -> RequestHandlerBuilder

Sets a custom task store from an existing Arc.

Use this when you want to share a store instance across multiple handlers or access it from background tasks.

pub fn with_task_store_config( self, config: TaskStoreConfig, ) -> RequestHandlerBuilder

Configures the default InMemoryTaskStore with custom TTL and capacity settings.

Panics

Panics in debug builds if a custom task store has already been set via with_task_store, since the config would be silently ignored.

pub fn with_push_config_store( self, store: impl PushConfigStore + 'static, ) -> RequestHandlerBuilder

Sets a custom push configuration store.

pub fn with_push_sender( self, sender: impl PushSender + 'static, ) -> RequestHandlerBuilder

Sets a push notification sender.

pub fn with_interceptor( self, interceptor: impl ServerInterceptor + 'static, ) -> RequestHandlerBuilder

Adds a server interceptor to the chain.

pub const fn with_executor_timeout( self, timeout: Duration, ) -> RequestHandlerBuilder

Sets a timeout for executor execution.

If the executor does not complete within this duration, the task is marked as failed with a timeout error.

There is deliberately no default: any fixed value would silently mark legitimately long-running agent tasks as failed. Production deployments should set a timeout matched to their workload so a hung executor cannot pin a task (and its stream resources) forever.

pub fn with_agent_card(self, card: AgentCard) -> RequestHandlerBuilder

Sets the agent card for discovery responses.

pub const fn with_event_queue_capacity( self, capacity: usize, ) -> RequestHandlerBuilder

Sets the event queue channel capacity for streaming.

Defaults to DEFAULT_QUEUE_CAPACITY (256 items). Higher values allow more events to be buffered before backpressure is applied. Lower values reduce memory footprint at the cost of earlier slow-consumer stalls.

pub const fn with_max_event_size( self, max_event_size: usize, ) -> RequestHandlerBuilder

Sets the maximum serialized event size in bytes.

Events exceeding this size are rejected to prevent OOM conditions. Defaults to 16 MiB.

pub const fn with_max_concurrent_streams( self, max: usize, ) -> RequestHandlerBuilder

Sets the maximum number of concurrent streaming event queues.

Limits memory usage from concurrent streams. When the limit is reached, new streaming requests will fail. Defaults to DEFAULT_MAX_CONCURRENT_STREAMS (1024); pass usize::MAX to effectively disable the ceiling.

pub const fn with_event_queue_write_timeout( self, timeout: Duration, ) -> RequestHandlerBuilder

👎Deprecated since 0.7.0:

has no effect: event queue writes never block; slow consumers receive an explicit lag error instead. Will be removed in 0.8.

Sets the write timeout for event queue sends.

Retained for API compatibility only. Event queue writes never block (the broadcast channel is non-blocking), so this value has no effect; a slow streaming consumer receives an explicit lag error on its reader instead of exerting backpressure on the executor.

pub const fn with_handler_limits( self, limits: HandlerLimits, ) -> RequestHandlerBuilder

Sets configurable limits for the handler (ID lengths, metadata size, etc.).

Defaults to HandlerLimits::default().

pub fn with_metrics( self, metrics: impl Metrics + 'static, ) -> RequestHandlerBuilder

Sets a metrics observer for handler activity.

Defaults to NoopMetrics which discards all events.

pub fn with_tenant_resolver( self, resolver: impl TenantResolver, ) -> RequestHandlerBuilder

Sets a tenant resolver for multi-tenant deployments.

The resolver extracts a tenant identifier from each incoming request’s CallContext. When combined with with_tenant_config, this enables per-tenant resource limits and configuration.

Defaults to None (single-tenant mode).

pub const fn require_resolved_tenant(self) -> RequestHandlerBuilder

Enables strict multi-tenancy: reject any request for which a configured with_tenant_resolver returns None (no tenant could be determined) instead of falling back to the shared default ("") partition.

Use this when every request must carry an identifiable tenant — it closes the gap where header-less or unauthenticated callers would otherwise all share one default bucket. Has no effect unless a resolver is configured.

Defaults to false (the resolver’s documented None → default-partition behavior is preserved).

pub const fn allow_unauthenticated_extended_card(self) -> RequestHandlerBuilder

Serves GetExtendedAgentCard even when no authenticating interceptor is registered.

Spec §13.3 requires the extended card endpoint to be authenticated, so by default a handler whose card declares capabilities.extendedAgentCard: true refuses to serve the card unless the interceptor chain contains at least one interceptor whose authenticates() returns true. Call this only when the extended card genuinely contains nothing sensitive (e.g. it is identical to the public card) or authentication is enforced upstream of this process (API gateway, service mesh, mTLS sidecar).

pub fn with_tenant_config( self, config: PerTenantConfig, ) -> RequestHandlerBuilder

Sets per-tenant configuration for multi-tenant deployments.

PerTenantConfig allows differentiated service levels (timeouts, capacity limits, rate limits) per tenant. Pair with with_tenant_resolver to extract the tenant identity from incoming requests.

Defaults to None (uniform limits for all callers).

pub fn build(self) -> Result<RequestHandler, ServerError>

Builds the RequestHandler.

Errors

Returns ServerError::InvalidParams if the configuration is invalid:

• Agent card with empty supported_interfaces
• Zero executor timeout (would cause immediate timeouts)

Trait Implementations

impl Debug for RequestHandlerBuilder

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

Formats the value using the given formatter.