Struct AgentSession 

pub struct AgentSession { /* private fields */ }

The product-level session. Cheap to clone — the harness is behind an Arc.

Implementations

impl AgentSession

pub fn new( harness: AgentHarness, catalog: Vec<Model>, scoped_models: Vec<ScopedModel>, cwd: impl Into<PathBuf>, ) -> Self

Wrap an existing harness. catalog is the resolved model catalog the session may switch within; scoped_models is the (possibly empty) settings-driven subset used by Ctrl+P / /model.

pub fn harness(&self) -> &AgentHarness

The wrapped harness, for the call sites (TUI drain task, extension bridge) that already speak the harness API.

pub fn harness_arc(&self) -> Arc<AgentHarness>

The Arc handle, for spawning work onto the runtime.

pub fn lane(&self) -> Arc<dyn AgentLane>

The main-lane runner.

pub fn cwd(&self) -> &Path

The working directory the session was opened in.

pub fn catalog(&self) -> &[Model]

The resolved model catalog.

pub fn scoped_models(&self) -> &[ScopedModel]

The active scoped-model set ([] → the full catalog is in play).

pub async fn prompt( &self, text: &str, images: Vec<ImageContent>, ) -> HarnessResult<RunResult>

Send a prompt with optional image attachments. Mirrors TS AgentSession.prompt.

pub async fn send_message( &self, message: AgentMessage, ) -> HarnessResult<RunResult>

Send an already-built user message. Mirrors TS AgentSession.sendUserMessage.

pub async fn steer(&self, message: AgentMessage) -> HarnessResult<()>

Queue a steering message for the in-flight run. Mirrors TS AgentSession.steer.

pub async fn follow_up(&self, message: AgentMessage) -> HarnessResult<()>

Queue a follow-up message for after the in-flight run. Mirrors TS AgentSession.followUp.

pub async fn abort(&self) -> HarnessResult<()>

Abort the in-flight run. Mirrors TS AgentSession.abort.

pub async fn model(&self) -> HarnessResult<Model>

The current model.

pub async fn set_model(&self, model: Model) -> HarnessResult<()>

Switch the model. Mirrors TS AgentSession.setModel.

pub async fn thinking_level(&self) -> HarnessResult<ThinkingLevel>

The current thinking level.

pub async fn set_thinking_level( &self, level: ThinkingLevel, ) -> HarnessResult<()>

Set the thinking level. Mirrors TS AgentSession.setThinkingLevel.

pub fn cycle_catalog(&self) -> Vec<Model>

The effective model cycle order: the scoped set when configured, else the full catalog. Mirrors TS AgentSession.getScopedModels.

pub async fn compact(&self, instructions: Option<&str>) -> HarnessResult<()>

Explicit compaction. Mirrors TS AgentSession.compact.

pub async fn retry(&self) -> HarnessResult<Option<RunResult>>

Retry the last turn. The harness owns the retry policy; this re-runs the last user message when one exists. Mirrors TS AgentSession.retry.

pub async fn tip_id(&self) -> HarnessResult<Option<String>>

The current leaf entry id.

pub async fn entries(&self) -> HarnessResult<Vec<Entry>>

All entries on the session, oldest first.

pub async fn snapshot(&self) -> HarnessResult<LaneSnapshot>

The lane read model. Mirrors the TS runtime lane snapshot.

pub async fn name(&self) -> HarnessResult<Option<String>>

Name/label facts. Mirrors TS AgentSession.setName / setLabel.

pub async fn set_name(&self, name: Option<&str>) -> HarnessResult<()>

Set (or clear) the session name.

pub async fn set_label( &self, target_id: &str, label: Option<&str>, ) -> HarnessResult<()>

Attach (or clear) a label on an entry.

pub async fn value(&self, key: &str) -> HarnessResult<Option<Value>>

Read a durable session value (see rpi_harness::session::values).

pub async fn set_value(&self, key: &str, value: Value) -> HarnessResult<()>

Persist a durable session value (latest-wins). Used for per-session display preferences so they survive resume.

pub async fn usage_stats(&self) -> HarnessResult<SessionUsageStats>

Rolled-up usage + context stats for the footer.

pub async fn total_usage(&self) -> HarnessResult<Usage>

The summed usage across every assistant message (for /context).

pub async fn last_assistant_text(&self) -> HarnessResult<Option<String>>

The last finalized assistant text (for /copy). Mirrors TS AgentSession.getLastAssistantText.

pub async fn export_to_html(&self, path: &Path) -> HarnessResult<()>

exportToHtml(). Mirrors TS AgentSession.exportToHtml.

pub async fn export_to_jsonl(&self, path: &Path) -> HarnessResult<()>

exportToJsonl(). Mirrors TS AgentSession.exportToJsonl.

pub async fn export_to_markdown(&self, path: &Path) -> HarnessResult<()>

exportToMarkdown() — the local /export default.

pub async fn export( &self, format: ExportFormat, path: &Path, ) -> HarnessResult<()>

Export to format, deriving the default filename from the session name/leaf when path is a directory.

pub async fn default_export_file_name(&self, format: ExportFormat) -> String

The filename /export would write for format.

pub async fn fork(&self, label: Option<&str>) -> HarnessResult<()>

Fork the session at the current leaf into a labelled branch. The navigation machinery creates the branch; the label tags its tip. Mirrors TS AgentSession.fork.

pub async fn result(&self, run_id: &str) -> HarnessResult<Option<RunResult>>

Read back the terminal outcome of a run by id. Mirrors TS AgentSession.getResult.

Trait Implementations

impl Clone for AgentSession

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.