ExecutableTool

CORE · interface · 类型合同

One registered tool, typed by its input, output, host context and id literal.

导入

TypeScript

import type { ExecutableTool } from '@jangts/agent-core/types';

JavaScript

/** @typedef {import('@jangts/agent-core/types').ExecutableTool} ExecutableTool */

其他公开入口：@jangts/agent-core/types、@jangts/agent-core/structs。

定义

interface ExecutableTool<Input, Output, App, Id>

继承或实现： ModelTool<Input, ToolProposal, App>。

泛型参数

类型参数	约束	默认类型
Input	—	unknown
Output	—	unknown
App	—	unknown
Id	string	string

字段结构

字段	类型	必需	说明
id	Id	必需	只读。Stable id; also the name the model calls.
kind	string	必需	只读。Declared grouping used by registries and documentation; not an authorization boundary.
description	string	必需	只读。Model-visible explanation of when to call the tool.
inputSchema	SchemaAdapter<Input>	必需	只读。Schema applied to raw model arguments before authorize and execute.
authorize	(input: Input, context: OperationContext<App>) => MaybePromise<void>	必需	Host permission check. Runs before execution and may be asynchronous.
execute	(input: Input, context: OperationContext<App> & import("../debug.js").NodeReportingContext) => MaybePromise<Output>	必需	Executes the already-authorized input; must not repeat authorization decisions.
debugSteps	((result: Output, context: import("../debug.js").NodeReportingContext) => void) | undefined	可省略	Optional result hook; writes through context.createSubFlowLogger() are recorded immediately.
parseOutput	(value: unknown) => Output	必需	Validates untrusted output (for example a cached replay) before it is returned.
active	ToolCondition<App> | undefined	可省略	只读。Omitted means active. Authorization remains a separate execution boundary.
asIntent	AsIntent<App> | undefined	可省略	只读。Omitted means this tool is not advertised by withIntent().
resolve	(input: Input, context: Readonly<{ stageId: string; }>) => ToolProposal	必需	Converts validated arguments into the value handed back to the caller. Runs inside the model call, so it must return synchronously; throwing rejects the selection.

成员目录

成员	用途
authorize	Host permission check. Runs before execution and may be asynchronous.
execute	Executes the already-authorized input; must not repeat authorization decisions.
parseOutput	Validates untrusted output (for example a cached replay) before it is returned.
resolve	Converts validated arguments into the value handed back to the caller. Runs inside the model call, so it must return synchronously; throwing rejects the selection.

成员详情

authorize

Host permission check. Runs before execution and may be asynchronous.

ExecutableTool.authorize(input: Input, context: OperationContext<App>): MaybePromise<void>

Host permission check. Runs before execution and may be asynchronous.

参数	类型	传递	说明
input	Input	必需	—
context	OperationContext<App>	必需	—

context 的字段

字段	类型	必需	说明
app	unknown extends App ? App : Readonly<App>	必需	只读。见类型定义及调用场景。
logger	SubFlowLog | undefined	可省略	只读。Explicit optional recorder for this operation; never inferred from diagnostics.
signal	AbortSignal	必需	只读。Non-optional cancellation signal; never undefined once execution has started.
diagnostics	DiagnosticContext | undefined	可省略	只读。Diagnostic scope of the enclosing call, when recording is enabled.

返回： MaybePromise<void>。

等待后的值： void。

execute

Executes the already-authorized input; must not repeat authorization decisions.

ExecutableTool.execute(input: Input, context: OperationContext<App> & import("../debug.js").NodeReportingContext): MaybePromise<Output>

Executes the already-authorized input; must not repeat authorization decisions.

参数	类型	传递	说明
input	Input	必需	—
context	OperationContext<App> & NodeReportingContext	必需	—

context 的字段

字段	类型	必需	说明
app	unknown extends App ? App : Readonly<App>	必需	只读。见类型定义及调用场景。
logger	SubFlowLog | undefined	可省略	只读。Explicit optional recorder for this operation; never inferred from diagnostics.
signal	AbortSignal	必需	只读。Non-optional cancellation signal; never undefined once execution has started.
diagnostics	DiagnosticContext | undefined	可省略	只读。Diagnostic scope of the enclosing call, when recording is enabled.
report	(update: import("./execution/workflow.js").WorkflowNodeReport) => void	必需	Reports progress for the active node; Core supplies the execution identity.
createSubFlowLogger	() => SubFlowLogger	必需	Bound to this context and owned by runtime; repeated calls in one visit return the same instance.

返回： MaybePromise<Output>。

等待后的值： Awaited<Output>。

parseOutput

Validates untrusted output (for example a cached replay) before it is returned.

ExecutableTool.parseOutput(value: unknown): Output

Validates untrusted output (for example a cached replay) before it is returned.

参数	类型	传递	说明
value	unknown	必需	—

返回： Output。

等待后的值： Awaited<Output>。

resolve

Converts validated arguments into the value handed back to the caller.

Runs inside the model call, so it must return synchronously; throwing rejects the selection.

ExecutableTool.resolve(input: Input, context: Readonly<{ stageId: string; }>): ToolProposal

Converts validated arguments into the value handed back to the caller.

Runs inside the model call, so it must return synchronously; throwing rejects the selection.

参数	类型	传递	说明
input	Input	必需	—
context	Readonly<{ stageId: string; }>	必需	—

context 的字段

字段	类型	必需	说明
stageId	string	必需	见类型定义及调用场景。

返回： ToolProposal。

返回值结构

字段	类型	必需	说明
toolId	string	必需	只读。见类型定义及调用场景。
input	unknown	必需	只读。见类型定义及调用场景。

使用示例

• 真实工具调用与请求内运行时：Core 注册并执行真实工具，返回带 invocation 的回执；业务授权由 App 注册的实现拥有。

最小本地只读工具

入门 · Node.js

TypeScript

import { ToolRegistry } from '@jangts/agent-core/tools';
const tools = new ToolRegistry().register({
  id: 'get_status', kind: 'read', description: '读取演示服务状态',
  parseInput(value: unknown) { if (!value || typeof value !== 'object' || Array.isArray(value)) throw new TypeError('需要对象'); return {}; },
  execute: () => '服务运行正常',
  parseOutput(value: unknown) { if (typeof value !== 'string') throw new TypeError('需要文本'); return value; },
}).seal();
const result = tools.invoke({ toolId: 'get_status', input: {} });
console.log(result.value, result.invocation.invocationId);

JavaScript

import { ToolRegistry } from '@jangts/agent-core/tools';
const tools = new ToolRegistry().register({
    id: 'get_status', kind: 'read', description: '读取演示服务状态',
    parseInput(value) { if (!value || typeof value !== 'object' || Array.isArray(value))
        throw new TypeError('需要对象'); return {}; },
    execute: () => '服务运行正常',
    parseOutput(value) { if (typeof value !== 'string')
        throw new TypeError('需要文本'); return value; },
}).seal();
const result = tools.invoke({ toolId: 'get_status', input: {} });
console.log(result.value, result.invocation.invocationId);

提前绑定，节点不再重复注入工具

配置 · Node.js

第一个示例严格校验外部输入；第二个仅展示绑定关系，真实业务应使用相应 Schema 与授权。

TypeScript

import { AgentCoreEngine, defineWorkflow, route } from '@jangts/agent-core/engine';
import { ToolRegistry } from '@jangts/agent-core/tools';
type State = { answer?: unknown };
const tools = new ToolRegistry().register({ id: 'get_status', kind: 'read',
  parseInput: () => ({}), execute: async () => '服务运行正常', parseOutput: value => value,
}).seal();
const workflow = defineWorkflow<State, unknown>({ id: 'status', nodes: [{
  id: 'read', description: '读取状态', startsRound: true,
  async run(_state, { tools }) { const { value } = await tools.invoke({ toolId: 'get_status', input: {} }); return { answer: value }; },
  routes: [route.deliver('answer')],
}] });
const agent = new AgentCoreEngine({ workflow }).bindToolRuntime(tools);
console.log(await agent.run({}));

JavaScript

import { AgentCoreEngine, defineWorkflow, route } from '@jangts/agent-core/engine';
import { ToolRegistry } from '@jangts/agent-core/tools';
const tools = new ToolRegistry().register({ id: 'get_status', kind: 'read',
    parseInput: () => ({}), execute: async () => '服务运行正常', parseOutput: value => value,
}).seal();
const workflow = defineWorkflow({ id: 'status', nodes: [{
            id: 'read', description: '读取状态', startsRound: true,
            async run(_state, { tools }) { const { value } = await tools.invoke({ toolId: 'get_status', input: {} }); return { answer: value }; },
            routes: [route.deliver('answer')],
        }] });
const agent = new AgentCoreEngine({ workflow }).bindToolRuntime(tools);
console.log(await agent.run({}));

相关策略源码

• 工具输入、执行、输出及迟到结果的检查顺序

相关定义

• IntentOptions
• MaybePromise
• ModelTool
• NodeReportingContext
• OperationContext
• SchemaAdapter
• ToolProposal