Struct ToolSpec 

pub struct ToolSpec {
    pub name: String,
    pub description: String,
    pub parameters: Option<Value>,
    pub response: Option<Value>,
    pub returns: Option<Value>,
    pub set_state: BTreeMap<String, Value>,
    pub save_response_as: Option<String>,
    pub http: Option<HttpBinding>,
    pub mcp: Option<String>,
    pub background: bool,
    pub scheduling: Option<SchedulingSpec>,
}

A declared tool. Without an http binding it is a mock: it returns response (default {"ok": true}) and writes set_state — enough to model, validate, and demo a governed conversation before any real tool exists. With http it performs the request instead (and still applies set_state afterwards, so guards latch identically) — swap a mock for a binding without touching the flow.

Fields

name: String

Tool (function) name the model calls.

description: String

Description shown to the model.

parameters: Option<Value>

JSON Schema for the arguments (Gemini subset). None = no parameters.

response: Option<Value>

Canned JSON response (mock tools). Default {"ok": true}.

returns: Option<Value>

JSON Schema of what the tool returns. Code mode shows it in the function’s docstring, so a program reads the right keys; a mock tool without it shows the shape of its response.

set_state: BTreeMap<String, Value>

State keys written when the tool runs — how a tool latches guards.

save_response_as: Option<String>

Store the tool’s full response under this state key.

http: Option<HttpBinding>

Execute as an HTTP request instead of returning the canned response.

mcp: Option<String>

Execute by calling the tool of the same name on this MCP server: a command line (stdio) or an http(s):// URL. This is how a tool written in another language (a generated Python or Go tool server) implements a declared tool. The declaration here stays what the model sees, and set_state still applies.

background: bool

Run non-blocking: the model keeps speaking while the tool executes (behavior: NonBlocking on the wire; Google AI only, stripped on Vertex). Implied by scheduling.

scheduling: Option<SchedulingSpec>

How the async response is delivered (implies background).

Implementations

impl ToolSpec

pub fn returns_shape(&self) -> Option<String>

What the tool returns, as a program’s docstring shows it: from returns, or the shape of the canned response of a mock tool.

Trait Implementations

impl Clone for ToolSpec

fn clone(&self) -> ToolSpec

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ToolSpec

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for ToolSpec

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl JsonSchema for ToolSpec

fn schema_name() -> String

The name of the generated JSON Schema.

fn schema_id() -> Cow<'static, str>

Returns a string that uniquely identifies the schema produced by this type.

fn json_schema(generator: &mut SchemaGenerator) -> Schema

Generates a JSON Schema for this type.

fn is_referenceable() -> bool

Whether JSON Schemas generated for this type should be re-used where possible using the $ref keyword.

impl Serialize for ToolSpec

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.