Trait RiskyExecutors 

pub trait RiskyExecutors {
    // Required methods
    async fn run(&self, args: RunArgs) -> Result<()>;
    async fn ps(&self, args: PsArgs) -> Result<()>;
    async fn msg(&self, args: MsgArgs) -> Result<()>;
    async fn cancel(&self, args: CancelArgs) -> Result<()>;
    async fn pause(&self, args: PauseArgs) -> Result<()>;
    async fn resume(&self, args: ResumeArgs) -> Result<()>;
    async fn respond(&self, args: RespondArgs) -> Result<()>;
    async fn doctor(&self, args: DoctorArgs) -> Result<()>;
    async fn setup(&self, args: SetupArgs) -> Result<()>;
    async fn dashboard(&self, args: DashboardArgs) -> Result<()>;
    async fn serve(&self, args: ServeArgs) -> Result<()>;
    async fn agent_client(&self, args: AgentClientArgs) -> Result<()>;
    async fn daemon(&self, args: DaemonArgs) -> Result<()>;
    async fn mcp(&self, args: McpArgs) -> Result<()>;
    async fn auth(&self, args: AuthArgs) -> Result<()>;
}

The subset of commands whose real execution performs I/O that a unit test must never trigger. dispatch() routes these through this trait so its routing logic stays unit-testable with a mock; the real implementations are supplied by the binary (main.rs’s RealExecutors).

async fn in a trait is fine here: dispatch takes &impl RiskyExecutors (static dispatch, no dyn), so no boxing or Send bound is required.

Required Methods

async fn run(&self, args: RunArgs) -> Result<()>

lev run - auto-starts the daemon (real process spawn) if needed and spawns the agent into the shared world over the control socket.

async fn ps(&self, args: PsArgs) -> Result<()>

lev ps - resolves the control-socket path and queries the daemon.

async fn msg(&self, args: MsgArgs) -> Result<()>

lev msg - resolves the control-socket path and sends a message.

async fn cancel(&self, args: CancelArgs) -> Result<()>

lev cancel - resolves the control-socket path and cancels a run.

async fn pause(&self, args: PauseArgs) -> Result<()>

lev pause - resolves the control-socket path and pauses a run.

async fn resume(&self, args: ResumeArgs) -> Result<()>

lev resume - resolves the control-socket path and resumes a run.

async fn respond(&self, args: RespondArgs) -> Result<()>

lev respond - resolves the control-socket path and answers/lists interactions.

async fn doctor(&self, args: DoctorArgs) -> Result<()>

lev doctor - makes real billed inference calls, and (unless --no-daemon) auto-starts the daemon and spawns a throwaway run.

async fn setup(&self, args: SetupArgs) -> Result<()>

lev setup - interactive (blocking stdin) or --non-interactive.

async fn dashboard(&self, args: DashboardArgs) -> Result<()>

lev dash - takes over the real terminal and blocks on real keyboard input.

async fn serve(&self, args: ServeArgs) -> Result<()>

lev serve - binds a real port and serves indefinitely.

async fn agent_client(&self, args: AgentClientArgs) -> Result<()>

lev agent-client - takes over real stdin/stdout to speak the Agent Client Protocol against the shared-world daemon.

async fn daemon(&self, args: DaemonArgs) -> Result<()>

lev daemon - binds the control socket and serves the shared world.

async fn mcp(&self, args: McpArgs) -> Result<()>

lev mcp - rewrites config, opens a browser for OAuth, touches the token store.

async fn auth(&self, args: AuthArgs) -> Result<()>

lev auth - reads the config file and may write the OS credential store.

Dyn Compatibility

This trait is not dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors