Struct RunnerConfig 

#[non_exhaustive]
pub struct RunnerConfig {

    pub app_name: String,
    pub agent: Arc<dyn Agent>,
    pub session_service: Arc<dyn SessionService>,
    pub artifact_service: Option<Arc<dyn ArtifactService>>,
    pub memory_service: Option<Arc<dyn Memory>>,
    pub plugin_manager: Option<Arc<PluginManager>>,
    pub run_config: Option<RunConfig>,
    pub compaction_config: Option<EventsCompactionConfig>,
    pub context_cache_config: Option<ContextCacheConfig>,
    pub cache_capable: Option<Arc<dyn CacheCapable>>,
    pub request_context: Option<RequestContext>,
    pub cancellation_token: Option<CancellationToken>,
    pub intra_compaction_config: Option<IntraCompactionConfig>,
    pub intra_compaction_summarizer: Option<Arc<dyn BaseEventsSummarizer>>,
}

Configuration for constructing a Runner.

Use Runner::builder() for a compile-time-safe way to construct this.

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

app_name: String

Application name used for session scoping.

agent: Arc<dyn Agent>

The root agent to execute.

session_service: Arc<dyn SessionService>

Session persistence backend.

artifact_service: Option<Arc<dyn ArtifactService>>

Optional artifact storage service.

memory_service: Option<Arc<dyn Memory>>

Optional memory/RAG service.

plugin_manager: Option<Arc<PluginManager>>

Optional plugin manager for lifecycle hooks.

run_config: Option<RunConfig>

Optional run configuration (streaming mode, etc.) If not provided, uses default (SSE streaming)

compaction_config: Option<EventsCompactionConfig>

Optional context compaction configuration. When set, the runner will periodically summarize older events to reduce context size sent to the LLM.

context_cache_config: Option<ContextCacheConfig>

Optional context cache configuration for automatic prompt caching lifecycle. When set alongside cache_capable, the runner will automatically create and manage cached content resources for supported providers.

When cache_capable is set but this field is None, the runner automatically uses ContextCacheConfig::default() (4096 min tokens, 600s TTL, refresh every 3 invocations).

cache_capable: Option<Arc<dyn CacheCapable>>

Optional cache-capable model reference for automatic cache management. Set this to the same model used by the agent if it supports caching.

request_context: Option<RequestContext>

Optional request context from the server’s auth middleware bridge. When set, the runner passes it to InvocationContext so that user_scopes() and user_id() reflect the authenticated identity.

cancellation_token: Option<CancellationToken>

Optional cooperative cancellation token for externally managed runs.

intra_compaction_config: Option<IntraCompactionConfig>

Optional intra-invocation compaction configuration. When set, the runner estimates token count before each agent run and triggers mid-invocation summarization when the threshold is exceeded.

intra_compaction_summarizer: Option<Arc<dyn BaseEventsSummarizer>>

Optional summarizer for intra-invocation compaction. Required when intra_compaction_config is set.