Struct ActionDefinition 

pub struct ActionDefinition {
    pub name: String,
    pub description: String,
    pub parameters: Vec<Parameter>,
    pub example: Value,
    pub log_template: Option<LogTemplate>,
}

Definition of an action for prompt generation

This describes an action to the LLM, including its name, description, parameters, and an example.

Fields

name: String

Action type name (e.g., “send_tcp_data”, “close_connection”)

description: String

Human-readable description of what this action does

parameters: Vec<Parameter>

List of parameters this action accepts

example: Value

JSON example showing how to use this action

log_template: Option<LogTemplate>

Log template for action execution logging Defines protocol-specific log formats at INFO/DEBUG/TRACE levels

Implementations

impl ActionDefinition

pub fn is_tool(&self) -> bool

Check if this is a tool action (returns information and triggers LLM re-invocation)

Delegates to tools::is_tool_name — the single source of truth shared with ToolAction::is_tool_action, which decides at runtime whether the model is re-invoked. Classifying here independently is what previously rendered read_documentation under the “Available Actions” heading, whose boilerplate tells the model it will not be invoked again — the opposite of what happens.

pub fn to_prompt_text(&self) -> String

Convert to prompt text format for LLM

pub fn with_log_template(self, template: LogTemplate) -> Self

Add a log template to this action definition

The log template defines protocol-specific log formats at INFO/DEBUG/TRACE levels. This enables standardized, centralized logging without per-action logging code.

Example

ActionDefinition {
    name: "send_http_response".to_string(),
    // ...other fields...
    log_template: None,
}
.with_log_template(
    LogTemplate::new()
        .with_info("HTTP {status}, {output_bytes}B")
        .with_debug("Response: status={status}, body_len={output_bytes}")
)

pub fn category(&self) -> ToolCategory

Derive the tool category from the action name

pub fn mcp_visible(&self) -> bool

Whether this action should be exposed as an MCP tool in STDIO mode

pub fn to_tool_schema(&self) -> Value

Convert to OpenAI/Ollama native tool calling schema

Returns a JSON value in the format:

{
  "type": "function",
  "function": {
    "name": "action_name",
    "description": "What this action does",
    "parameters": {
      "type": "object",
      "properties": { ... },
      "required": [ ... ]
    }
  }
}

pub fn to_mcp_tool_schema(&self) -> Value

Convert to MCP tool schema for rmcp

Returns a JSON value in the format:

{
  "name": "action_name",
  "description": "What this action does",
  "inputSchema": {
    "type": "object",
    "properties": { ... },
    "required": [ ... ]
  }
}

Trait Implementations

impl Clone for ActionDefinition

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ActionDefinition

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

Formats the value using the given formatter.