ErasedTool

CORE · interface · Type contract

Type-erased tool shape the registry executes after parsing at the boundary.

Import

TypeScript

import type { ErasedTool } from '@jangts/agent-core/tools';

JavaScript

/** @typedef {import('@jangts/agent-core/tools').ErasedTool} ErasedTool */

Other public entry points:@jangts/agent-core, @jangts/agent-core/types, @jangts/agent-core/structs, @jangts/agent-core/tools.

Definition

interface ErasedTool

Field structure

Field	Type	Required	Description
id	string	Required	Read-only. See the type definition and usage context.
kind	string	Required	Read-only. See the type definition and usage context.
description	string | undefined	Optional	Read-only. Optional here because the erased view does not require model-facing prose.
debugSteps	((result: unknown, context: NodeReportingContext) => void) | undefined	Optional	Read-only. See the type definition and usage context.
parseInput	(value: unknown) => unknown	Required	See the type definition and usage context.
parseOutput	(value: unknown) => unknown	Required	See the type definition and usage context.
execute	(input: unknown, context: ToolExecutionContext) => unknown	Required	See the type definition and usage context.

Member index

Member	Purpose
parseInput	See the overload below for full inputs and outputs.
parseOutput	See the overload below for full inputs and outputs.
execute	See the overload below for full inputs and outputs.

Member details

parseInput

ErasedTool.parseInput(value: unknown): unknown

Parameter	Type	Passing	Description
value	unknown	Required	—

Returns: unknown.

parseOutput

ErasedTool.parseOutput(value: unknown): unknown

Parameter	Type	Passing	Description
value	unknown	Required	—

Returns: unknown.

execute

ErasedTool.execute(input: unknown, context: ToolExecutionContext): unknown

Parameter	Type	Passing	Description
input	unknown	Required	—
context	ToolExecutionContext	Required	—

Fields of context

Field	Type	Required	Description
logger	ScopedSubFlowLogger | undefined	Optional	Read-only. Delegate bound by Core to this actual invocation.
diagnostics	DiagnosticContext | undefined	Optional	Read-only. Diagnostic scope of the invocation, when recording is enabled.
mode	"sync" | "async"	Required	Read-only. Declared execution discipline; mirrors whether the registry accepted an async tool.
invocationId	string	Required	Read-only. Opaque invocation id shared by receipt, events and diagnostics.
signal	AbortSignal	Required	Read-only. Non-optional cancellation for this invocation.
report	(update: import("./execution/workflow.js").WorkflowNodeReport) => void	Required	Reports progress for the active node; Core supplies the execution identity.
createSubFlowLogger	() => SubFlowLogger	Required	Bound to this context and owned by runtime; repeated calls in one visit return the same instance.

Returns: unknown.

Examples

• Agent: run first, then bind tools and dependencies step by step: The Engine executes a bounded loop; after a Workflow is bound, each run only receives the data that changed in state.
• Reusing a base Workflow and Agent branches: add preserves the old Workflow and produces a new combined Workflow; fork inherits existing bindings and the branch origin.
• Bind once, multiple models per run, and switching across runs: Provide stable model configuration at construction; requests select by configuration ID instead of injecting a Provider repeatedly.
• Real tool invocation and in-request runtime: Core registers and executes real tools and returns receipts carrying an invocation; business authorization is owned by the implementation registered by the App.
• Decorators: complete handlers and existing host methods: Classes construct definitions automatically; methods bind the real host. Outputs still use original define, constructor and bind contracts.

A single ordinary step is enough to run

Beginner · Node.js

TypeScript

import { AgentCoreEngine } from '@jangts/agent-core/engine';

const agent = new AgentCoreEngine<string>();
const answer = await agent.run(() => ({ type: 'complete', value: '你好，Core' }));
console.log(answer);

JavaScript

import { AgentCoreEngine } from '@jangts/agent-core/engine';
const agent = new AgentCoreEngine();
const answer = await agent.run(() => ({ type: 'complete', value: '你好，Core' }));
console.log(answer);

Add an explicit Workflow and still call run

Configuration · Node.js

TypeScript

import { AgentCoreEngine, defineWorkflow, route } from '@jangts/agent-core/engine';

type State = { name: string; answer?: string };
const workflow = defineWorkflow<State, string>({
  id: 'greeting',
  nodes: [{
    id: 'greet', description: '根据本次姓名生成问候', startsRound: true,
    run: state => ({ answer: '你好，' + state.name }),
    routes: [route.deliver('answer')],
  }],
});

const agent = new AgentCoreEngine({ workflow });
console.log(await agent.run({ name: 'Ivan' }));
console.log(await agent.run({ name: '访客' }));

JavaScript

import { AgentCoreEngine, defineWorkflow, route } from '@jangts/agent-core/engine';
const workflow = defineWorkflow({
    id: 'greeting',
    nodes: [{
            id: 'greet', description: '根据本次姓名生成问候', startsRound: true,
            run: state => ({ answer: '你好，' + state.name }),
            routes: [route.deliver('answer')],
        }],
});
const agent = new AgentCoreEngine({ workflow });
console.log(await agent.run({ name: 'Ivan' }));
console.log(await agent.run({ name: '访客' }));

Related policy source

• Execution implementation of the bounded loop, commit, and cancellation
• Check order for tool input, execution, output, and late results

Related definitions

• NodeReportingContext
• ToolExecutionContext