Struct ToolCall 

pub struct ToolCall {
    pub tool_id: ToolName,
    pub params: Map<String, Value>,
    pub caller_id: Option<String>,
    pub context: Option<ExecutionContext>,
}

Structured tool invocation from LLM.

Produced by the agent loop when the LLM emits a structured tool call (as opposed to a legacy fenced code block). Dispatched to ToolExecutor::execute_tool_call.

Example

use zeph_tools::{ToolCall, ExecutionContext};
use zeph_common::ToolName;

let call = ToolCall {
    tool_id: ToolName::new("bash"),
    params: {
        let mut m = serde_json::Map::new();
        m.insert("command".to_owned(), serde_json::Value::String("echo hello".to_owned()));
        m
    },
    caller_id: Some("user-42".to_owned()),
    context: Some(ExecutionContext::new().with_name("repo")),
};
assert_eq!(call.tool_id, "bash");

Fields

tool_id: ToolName

The tool identifier, matching a value from ToolExecutor::tool_definitions.

params: Map<String, Value>

JSON parameters for the tool call, deserialized into the tool’s parameter struct.

caller_id: Option<String>

Opaque caller identifier propagated from the channel (user ID, session ID, etc.). None for system-initiated calls (scheduler, self-learning, internal).

context: Option<ExecutionContext>

Per-turn execution environment. None means use the executor default (process CWD and inherited env), which is identical to the behaviour before this field existed.

Trait Implementations

impl Clone for ToolCall

fn clone(&self) -> ToolCall

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ToolCall

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

Formats the value using the given formatter.