Struct TextRunner 

pub struct TextRunner { /* private fields */ }

Runs TextAgents with full service wiring (session, memory, artifacts, plugins).

Auto-wires in-memory service implementations by default; override with builder methods for custom persistence.

Implementations

impl TextRunner

pub fn new(agent: Arc<dyn TextAgent>, app_name: impl Into<String>) -> Self

Create a new runner with in-memory defaults for all services.

pub fn session_service(self, svc: Arc<dyn SessionService>) -> Self

Override the session service.

pub fn memory_service(self, svc: Arc<dyn MemoryService>) -> Self

Override the memory service.

pub fn artifact_service(self, svc: Arc<dyn ArtifactService>) -> Self

Override the artifact service.

pub fn plugin(self, p: impl Plugin + 'static) -> Self

Add a plugin.

pub async fn run( &self, prompt: &str, user_id: &str, session_id: Option<&SessionId>, ) -> Result<String, AgentError>

Run with session management. Creates or resumes a session.

1. Creates a new session or loads an existing one
2. Sets "input" in state from prompt
3. Runs the agent
4. Persists the result as an event in the session
5. Returns the agent’s text output

This is a thin convenience wrapper over run_stream: it drains the event stream and returns the final response text (or the surfaced error). Both methods share the exact same execution path.

pub fn run_stream<'a>( &'a self, prompt: &'a str, user_id: &'a str, session_id: Option<&'a SessionId>, ) -> BoxStream<'a, RunEvent>

Run the agent as an event stream, mirroring ADK-Python’s Runner.run_async() which yields a Stream<Event>.

Yields RunEvent items as they happen, backed by the same execution path as run:

1. a user event carrying the prompt (also persisted to the session),
2. zero or more agent events surfacing state deltas produced during the run (one per changed key, so eval harnesses and UIs can observe mutations), and
3. a final response event authored by the root agent (also persisted).

Setup failures (session create/load, event persistence) and agent failures are surfaced as a terminal RunEvent::Error rather than a Result, so the item type stays a plain event. After an error the stream ends.

run(prompt, user, session) is equivalent to draining this stream and returning the last response event’s content.

pub async fn run_ephemeral(&self, prompt: &str) -> Result<String, AgentError>

Run without persistence (one-shot, ephemeral).

pub fn session_service_ref(&self) -> &dyn SessionService

Access the session service.

pub fn app_name(&self) -> &str

Access the app name.