generators

KIT · constant · 运行时值

导入

TypeScript

import { generators } from '@jangts/agent-core/mock';

JavaScript

import { generators } from '@jangts/agent-core/mock';

其他公开入口：@jangts/agent-core/mock。

定义

成员目录

成员	用途
summary	Extractive summary, one paragraph, at most maxChars (default 240) characters.
bullets	Same selection as summary, rendered as - bullet lines (default budget 500).
keywords	Most frequent non-stopword terms, most frequent first.
title	First sentence of the source with leading Markdown heading marks removed.
reply	Short conversational answer that echoes the topic. Continuation prompts ("继续", "换个说法") reuse the previous user message so a demo conversation stays coherent without a model.
extract	Extract the first match of each configured pattern into a JSON object. A capture group is preferred over the whole match, so id:(\d+) yields just the digits. A field with no match is null rather than omitted, keeping the output shape stable.
integer	Uniform integer in an inclusive range.
float	Continuous value in [min, max), or exactly min when both bounds are equal.
boolean	Fair coin derived from the call PRNG.
pick	Uniform choice from a fixed pool. The pool is copied at construction and each result is copied again, so a consumer mutating a returned object cannot corrupt later selections.
id	Identifier that includes the call index plus a random suffix. The suffix is 32 random bits rendered in hex; it is a fixture value, not a UUID and not guaranteed unique or unguessable.
date	ISO-8601 timestamp read from the injected clock. With a fixed clock option the value is constant, which keeps snapshot-style tests stable.
sentence	One of three fixed Chinese sample sentences. The sentences are placeholders that read naturally in demos; selection still consumes the per-call PRNG so the value changes with the seed.
imageUrl	Inline SVG data URL, so no network or filesystem access is needed.
sequence	Play a fixed list one item per accepted call. Selection is driven by context.ruleIndex (1-based), so it advances with the rule's own hit count and is unaffected by other rules or by concurrent calls.
toolCall	Shorthand for a reply that only emits one tool call.

成员详情

summary

Extractive summary, one paragraph, at most maxChars (default 240) characters.

generators.summary(options?: TextOptions): Generator

参数	类型	传递	说明
options	TextOptions	可省略	—

options 的字段

字段	类型	必需	说明
source	TextSource | undefined	可省略	Explicit text or async provider; defaults to the request prompt.
maxChars	number | undefined	可省略	Maximum characters in the returned text (1-30 000); generator-specific default.
count	number | undefined	可省略	How many items to select (1-100), e.g. bullets, keywords or summary sentences.

返回： Generator。

bullets

Same selection as summary, rendered as - bullet lines (default budget 500).

generators.bullets(options?: TextOptions): Generator

参数	类型	传递	说明
options	TextOptions	可省略	—

options 的字段

字段	类型	必需	说明
source	TextSource | undefined	可省略	Explicit text or async provider; defaults to the request prompt.
maxChars	number | undefined	可省略	Maximum characters in the returned text (1-30 000); generator-specific default.
count	number | undefined	可省略	How many items to select (1-100), e.g. bullets, keywords or summary sentences.

返回： Generator。

keywords

Most frequent non-stopword terms, most frequent first.

generators.keywords(options?: TextOptions): Generator

参数	类型	传递	说明
options	TextOptions	可省略	—

options 的字段

字段	类型	必需	说明
source	TextSource | undefined	可省略	Explicit text or async provider; defaults to the request prompt.
maxChars	number | undefined	可省略	Maximum characters in the returned text (1-30 000); generator-specific default.
count	number | undefined	可省略	How many items to select (1-100), e.g. bullets, keywords or summary sentences.

返回： Generator。

title

First sentence of the source with leading Markdown heading marks removed.

generators.title(options?: TextOptions): Generator

参数	类型	传递	说明
options	TextOptions	可省略	—

options 的字段

字段	类型	必需	说明
source	TextSource | undefined	可省略	Explicit text or async provider; defaults to the request prompt.
maxChars	number | undefined	可省略	Maximum characters in the returned text (1-30 000); generator-specific default.
count	number | undefined	可省略	How many items to select (1-100), e.g. bullets, keywords or summary sentences.

返回： Generator。

reply

Short conversational answer that echoes the topic.

Continuation prompts ("继续", "换个说法") reuse the previous user message so a demo conversation stays coherent without a model.

generators.reply(options?: TextOptions & { template?: (input: { text: string; previous: string; }) => string; }): Generator

参数	类型	传递	说明
options	TextOptions & { template?: (input: { text: string; previous: string; }) => string; }	可省略	—

options 的字段

字段	类型	必需	说明
source	TextSource | undefined	可省略	Explicit text or async provider; defaults to the request prompt.
maxChars	number | undefined	可省略	Maximum characters in the returned text (1-30 000); generator-specific default.
count	number | undefined	可省略	How many items to select (1-100), e.g. bullets, keywords or summary sentences.
template	((input: { text: string; previous: string; }) => string) | undefined	可省略	见类型定义及调用场景。

返回： Generator。

extract

Extract the first match of each configured pattern into a JSON object.

A capture group is preferred over the whole match, so id:(\d+) yields just the digits. A field with no match is null rather than omitted, keeping the output shape stable.

generators.extract(options?: TextOptions & { fields?: Record<string, RegExp>; }): Generator

参数	类型	传递	说明
options	TextOptions & { fields?: Record<string, RegExp>; }	可省略	—

options 的字段

字段	类型	必需	说明
source	TextSource | undefined	可省略	Explicit text or async provider; defaults to the request prompt.
maxChars	number | undefined	可省略	Maximum characters in the returned text (1-30 000); generator-specific default.
count	number | undefined	可省略	How many items to select (1-100), e.g. bullets, keywords or summary sentences.
fields	Record<string, RegExp> | undefined	可省略	见类型定义及调用场景。

返回： Generator。

integer

Uniform integer in an inclusive range.

generators.integer(min?: number, max?: number): ValueGenerator<number>

参数	类型	传递	说明
min	number	可省略	—
max	number	可省略	—

返回： ValueGenerator<number>。

float

Continuous value in [min, max), or exactly min when both bounds are equal.

generators.float(min?: number, max?: number): ValueGenerator<number>

参数	类型	传递	说明
min	number	可省略	—
max	number	可省略	—

返回： ValueGenerator<number>。

boolean

Fair coin derived from the call PRNG.

generators.boolean(): ValueGenerator<boolean>

无调用参数。

返回： ValueGenerator<boolean>。

pick

Uniform choice from a fixed pool.

The pool is copied at construction and each result is copied again, so a consumer mutating a returned object cannot corrupt later selections.

generators.pick<T extends JsonValue>(items: readonly T[]): ValueGenerator<T>

类型参数	约束	默认类型
T	JsonValue	—

参数	类型	传递	说明
items	readonly T[]	必需	—

返回： ValueGenerator<T>。

id

Identifier that includes the call index plus a random suffix.

The suffix is 32 random bits rendered in hex; it is a fixture value, not a UUID and not guaranteed unique or unguessable.

generators.id(prefix?: string): ValueGenerator<string>

参数	类型	传递	说明
prefix	string	可省略	—

返回： ValueGenerator<string>。

date

ISO-8601 timestamp read from the injected clock.

With a fixed clock option the value is constant, which keeps snapshot-style tests stable.

generators.date(): ValueGenerator<string>

无调用参数。

返回： ValueGenerator<string>。

sentence

One of three fixed Chinese sample sentences.

The sentences are placeholders that read naturally in demos; selection still consumes the per-call PRNG so the value changes with the seed.

generators.sentence(): ValueGenerator<string>

无调用参数。

返回： ValueGenerator<string>。

imageUrl

Inline SVG data URL, so no network or filesystem access is needed.

generators.imageUrl({ width, height, }?: { width?: number; height?: number; }): ValueGenerator<string>

参数	类型	传递	说明
__0	{ width?: number; height?: number; }	可省略	—

__0 的字段

字段	类型	必需	说明
width	number | undefined	可省略	见类型定义及调用场景。
height	number | undefined	可省略	见类型定义及调用场景。

返回： ValueGenerator<string>。

sequence

Play a fixed list one item per accepted call.

Selection is driven by context.ruleIndex (1-based), so it advances with the rule's own hit count and is unaffected by other rules or by concurrent calls.

generators.sequence(items: readonly Generator[]): Generator

参数	类型	传递	说明
items	readonly Generator[]	必需	—

返回： Generator。

toolCall

Shorthand for a reply that only emits one tool call.

generators.toolCall(name: string, args: JsonValue, options?: OutputOptions): Generator

参数	类型	传递	说明
name	string	必需	—
args	JsonValue	必需	—
options	OutputOptions	可省略	—

options 的字段

字段	类型	必需	说明
finishReason	"length" | "tool_calls" | "stop" | "content_filter" | undefined	可省略	Finish reason sent in the last content frame; defaults to tool_calls with tools, else stop.
usage	Usage | undefined	可省略	Usage object included in non-stream replies and in the terminal usage chunk.
delayMs	number | undefined	可省略	Artificial latency before the reply or first frame, in milliseconds; default 0.
chunkSize	number | undefined	可省略	Maximum Unicode code points per content/tool-call delta; default 24; never splits a code point.
intervalMs	number | undefined	可省略	Delay between SSE frames in milliseconds; default 0 (emit as fast as the consumer pulls).
wireChunkSize	number | undefined	可省略	Optional transport-level byte chunks, independent of content/codepoint chunks.
interruptAfter	number | undefined	可省略	Error the stream after this many data frames, without a success terminator.

返回： Generator。

使用示例

• MockGPT：响应描述与场景生成器：固定结果用于严格回归，输入相关生成器用于探索接线，显式响应描述用于故障和流式场景。

编码已校验的模型结果

扩展 · Node.js

适配器负责请求校验、首包前延迟和生命周期；编码器不执行规则或修改 MockGPT 调用记录。

TypeScript

import { encodeModelResponse } from '@jangts/agent-core/mock';
const controller = new AbortController();
const response = encodeModelResponse(
  { model: 'demo', messages: [{ role: 'user', content: '状态' }], stream: false },
  { content: '服务在线。' }, 'mock-1', 1700000000, controller.signal,
  (error, cancelled) => console.log({ error, cancelled }),
);
console.log(await response.json());

JavaScript

import { encodeModelResponse } from '@jangts/agent-core/mock';
const controller = new AbortController();
const response = encodeModelResponse({ model: 'demo', messages: [{ role: 'user', content: '状态' }], stream: false }, { content: '服务在线。' }, 'mock-1', 1700000000, controller.signal, (error, cancelled) => console.log({ error, cancelled }));
console.log(await response.json());

从输入生成可检查的摘要

入门 · Node.js

TypeScript

import { MockGPT, generators } from '@jangts/agent-core/mock';
const mock = new MockGPT({ rules: [{ generator: generators.summary({ maxChars: 120, count: 2 }) }] });
console.log(await mock.generate({ model: 'demo', messages: [{ role: 'user', content: '服务刚完成维护。健康检查通过。用户可以查询状态。' }] }));
console.log(mock.records);

JavaScript

import { MockGPT, generators } from '@jangts/agent-core/mock';
const mock = new MockGPT({ rules: [{ generator: generators.summary({ maxChars: 120, count: 2 }) }] });
console.log(await mock.generate({ model: 'demo', messages: [{ role: 'user', content: '服务刚完成维护。健康检查通过。用户可以查询状态。' }] }));
console.log(mock.records);

字段生成器复用同一随机源

配置 · Node.js

TypeScript

import { MockGPT, generators, generatorNames } from '@jangts/agent-core/mock';
const id = generators.id('user');
const age = generators.integer(18, 60);
const mock = new MockGPT({ seed: 42, clock: () => 1700000000000, rules: [{
  generator: context => ({ id: id(context), age: age(context) }),
}] });
console.log(await mock.generate({ model: 'demo', messages: [{ role: 'user', content: '生成资料' }] }));
console.log(generatorNames);

JavaScript

import { MockGPT, generators, generatorNames } from '@jangts/agent-core/mock';
const id = generators.id('user');
const age = generators.integer(18, 60);
const mock = new MockGPT({ seed: 42, clock: () => 1700000000000, rules: [{
            generator: context => ({ id: id(context), age: age(context) }),
        }] });
console.log(await mock.generate({ model: 'demo', messages: [{ role: 'user', content: '生成资料' }] }));
console.log(generatorNames);

相关定义

• Generator
• JsonValue
• OutputOptions
• ResponseDescription
• TextOptions