Struct FunctionTool 

pub struct FunctionTool { /* private fields */ }

A tool wrapping a user-provided async closure.

Implementations

impl FunctionTool

pub fn new( name: impl Into<String>, description: impl Into<String>, parameters: Option<Schema>, f: Arc<dyn for<'a> Fn(Value, &'a mut ToolContext) -> BoxFuture<'a, Result<Value>> + Send + Sync + 'static>, ) -> Self

Construct.

pub fn with_long_running(self, yes: bool) -> Self

Mark the tool as long-running.

pub fn require_confirmation(self, yes: bool) -> Self

Require explicit user confirmation before each call (human-in-the- loop). The agent pauses with an adk_request_confirmation request instead of dispatching; see crate::core::tool_confirmation.

pub fn with_confirmation_hint(self, hint: impl Into<String>) -> Self

Custom hint shown to the user when confirmation is requested.

pub fn from_async<F, Fut>( name: impl Into<String>, description: impl Into<String>, parameters: Option<Schema>, f: F, ) -> Self

where F: for<'a> Fn(Value, &'a mut ToolContext) -> Fut + Send + Sync + 'static, Fut: Future<Output = Result<Value>> + Send + 'static,

Wrap any async fn(Value, &mut ToolContext) -> Result<Value>.

Trait Implementations

impl Debug for FunctionTool

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

Formats the value using the given formatter.

impl DynTool for FunctionTool

fn name(&self) -> &str

The tool name (matches the model’s function-call name).

fn description(&self) -> &str

Human description.

fn is_long_running(&self) -> bool

Whether this tool yields a long-running operation.

fn requires_confirmation(&self, _args: &Value) -> bool

Whether this call requires explicit user confirmation before it can run (human-in-the-loop). When true, the agent pauses with an adk_request_confirmation request instead of dispatching the tool; see crate::core::tool_confirmation. args lets implementations decide per call (e.g. only confirm destructive operations).

fn confirmation_hint(&self, _args: &Value) -> String

Human-readable hint shown to the user when confirmation is requested. Defaults to a generic message naming the tool.

fn declaration(&self) -> Option<FunctionDeclaration>

JSON-Schema declaration of the tool’s parameters; None for tools (e.g. Gemini built-ins) that should not be advertised to the model.

fn run<'life0, 'life1, 'async_trait>( &'life0 self, args: Value, ctx: &'life1 mut ToolContext, ) -> Pin<Box<dyn Future<Output = Result<Value>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Execute the tool with JSON args.

fn auth_config(&self) -> Option<&AuthConfig>

Auth requirements for this tool. When Some, the runner resolves the credential via crate::auth::CredentialManager before calling Self::run, and injects the result into crate::core::ToolContext::auth_credential. When the underlying flow needs interactive consent the runner emits an adk_request_credential function-call event instead of dispatching the tool.

fn process_llm_request<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, req: &'life1 mut LlmRequest, _ctx: &'life2 mut ToolContext, ) -> Pin<Box<dyn Future<Output = Result<()>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Hook called before the request is sent. Default: append the tool’s declaration into req.config.tools.