Struct ToolExecutor 

pub struct ToolExecutor { /* private fields */ }

Tool executor with workspace sandboxing

This is the main entry point for tool execution. It wraps the ToolRegistry.

Implementations

impl ToolExecutor

pub fn new(workspace: String) -> Self

pub fn new_with_artifact_limits( workspace: String, artifact_limits: ArtifactStoreLimits, ) -> Self

pub fn new_with_workspace_services( workspace: String, workspace_services: Arc<WorkspaceServices>, ) -> Self

pub fn new_with_workspace_services_and_artifact_limits( workspace: String, workspace_services: Arc<WorkspaceServices>, artifact_limits: ArtifactStoreLimits, ) -> Self

pub fn workspace(&self) -> &PathBuf

pub fn registry(&self) -> &Arc<ToolRegistry>

pub fn get_artifact(&self, artifact_uri: &str) -> Option<ToolArtifact>

Get a stored tool artifact by URI.

pub fn artifact_store(&self) -> ArtifactStore

Return a clone of the executor’s artifact store handle.

pub fn set_trace_sink(&self, sink: Arc<dyn TraceSink>)

Replace the sink used for compact execution trace events.

pub fn trace_sink(&self) -> Arc<dyn TraceSink>

Return the currently configured execution trace sink.

pub fn command_env(&self) -> Option<Arc<HashMap<String, String>>>

pub fn register_dynamic_tool(&self, tool: Arc<dyn Tool>)

pub fn unregister_dynamic_tool(&self, name: &str)

pub fn unregister_tools_by_prefix(&self, prefix: &str)

Unregister all dynamic tools whose names start with the given prefix.

pub fn register_program_catalog(&self, catalog: ProgramCatalog)

Replace the model-visible program tool with a custom PTC catalog.

pub async fn execute(&self, name: &str, args: &Value) -> Result<ToolResult>

Execute directly against this low-level executor.

This API intentionally does not install agent/session permission, HITL, hook, budget, queue, timeout, cancellation, or sanitization policy. Session hosts should use crate::AgentSession::tool (or its typed helpers), and agent runtimes must dispatch through their scoped tool invocation gateway.

pub async fn execute_with_context( &self, name: &str, args: &Value, ctx: &ToolContext, ) -> Result<ToolResult>

Execute directly with a caller-owned context.

Like Self::execute, this is an ungoverned standalone boundary. A ToolContext supplies capabilities to the tool but is not itself a substitute for the agent/session invocation gateway.

pub fn definitions(&self) -> Vec<ToolDefinition>