Struct TeamRunner 

pub struct TeamRunner { /* private fields */ }

Binds an AgentTeam to concrete AgentExecutor sessions, enabling Lead → Worker → Reviewer automated workflows.

Implementations

impl TeamRunner

pub fn new(team: AgentTeam) -> Self

Create a new runner wrapping the given team.

pub fn with_agent( team: AgentTeam, agent: Arc<Agent>, workspace: &str, registry: Arc<AgentRegistry>, ) -> Self

Create a runner with a default agent context for simplified member addition.

Unlike TeamRunner::new, this constructor lets you call add_lead, add_worker, and add_reviewer without repeating the agent, workspace, and registry on every call.

pub fn bind_session( &mut self, member_id: &str, executor: Arc<dyn AgentExecutor>, ) -> Result<()>

Bind an executor to a team member.

Returns an error if member_id is not registered in the team.

pub fn bind_agent( &mut self, member_id: &str, agent: &Agent, workspace: &str, agent_name: &str, registry: &AgentRegistry, ) -> Result<()>

Bind a team member to a session created from a named agent definition.

Looks up agent_name in registry, applies the definition’s prompt, permissions, model, and max_steps to a new [AgentSession] via [Agent::session_for_agent], then binds it to member_id.

This is the primary integration point between the AgentTeam coordination layer and the markdown/YAML-defined subagent capability layer.

Errors

Returns an error if member_id is not in the team, agent_name is not in registry, or session creation fails.

pub fn add_lead( &mut self, agent_name: &str, opts: Option<TeamMemberOptions>, ) -> Result<()>

Add a Lead member and bind it to the named agent definition.

Requires a default agent context set via TeamRunner::with_agent. The member ID is fixed to "lead".

Use opts to override the model, prompt slots, or workspace (e.g. a git worktree).

Errors

Returns an error if no default context is set or agent_name is not found in the registry.

pub fn add_worker( &mut self, agent_name: &str, opts: Option<TeamMemberOptions>, ) -> Result<()>

Add a Worker member and bind it to the named agent definition.

Requires a default agent context set via TeamRunner::with_agent. Member IDs are auto-generated as "worker-1", "worker-2", etc. Multiple workers can be added; they run concurrently during execution.

Use opts to override the model, prompt slots, or workspace — set workspace to an isolated git worktree path to prevent filesystem conflicts between concurrent workers.

Errors

Returns an error if no default context is set or agent_name is not found in the registry.

pub fn add_reviewer( &mut self, agent_name: &str, opts: Option<TeamMemberOptions>, ) -> Result<()>

Add a Reviewer member and bind it to the named agent definition.

Requires a default agent context set via TeamRunner::with_agent. The member ID is fixed to "reviewer".

Use opts to override the model, prompt slots, or workspace.

Errors

Returns an error if no default context is set or agent_name is not found in the registry.

pub fn team_mut(&mut self) -> &mut AgentTeam

Get a mutable reference to the underlying team.

pub fn task_board(&self) -> Arc<TeamTaskBoard>

Access the shared task board.

pub async fn run_until_done(&self, goal: &str) -> Result<TeamRunResult>

Run the full Lead → Worker → Reviewer workflow until all tasks are done or max_rounds retry cycles are exceeded.

Steps:

1. Lead decomposes goal into tasks via JSON response.
2. Workers run concurrently until all tasks are in review or done.
3. Reviewer processes all InReview tasks (after workers finish).
4. If rejected tasks remain, workers retry them (back to step 2).

Running the reviewer after workers (rather than concurrently) prevents a race where the reviewer’s polling timeout fires before long-running agent calls have produced any InReview tasks.

Trait Implementations

impl Debug for TeamRunner

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

Formats the value using the given formatter.