English | 中文

cersei_rs

• cersei_rs
  • Installation
  • API Usage
    • Constants
    • logChat (Default Export)
    • logSession
    • chat
    • session
  • License
  • About

N-API wrapper for Cersei, providing agent creation, streaming chat session, and logging functionality in Node.js.

Installation

npm install cersei_rs

API Usage

Constants

Import constants from "cersei_rs/MSG" to identify event types in the stream:

• MSG_ERR (1): Error occurred.
• MSG_TOOL (2): Tool execution started.
• MSG_TXT (3): Streaming text content.
• MSG_THINK (4): Streaming thinking process.

logChat (Default Export)

Wrapper function that automatically logs agent execution (thinking process, tool calls, and output text) to stdout and returns a Promise resolving to the final accumulated text response.

import logChat from "cersei_rs";

const agent = logChat(baseUrl, apiKey, model);
const response = await agent("Write a programmer joke in tmp.md", "./gen");
console.log("\nResponse:", response);

logSession

Wrapper function that automatically logs agent execution for a session maintaining chat history, returning a function that yields a Promise resolving to the final accumulated text response.

import logSession from "cersei_rs/logSession";

const history = [
  { role: "user", content: "Remember my name is Cersei" },
  { role: "assistant", content: "Got it, I'll remember that your name is Cersei." },
];

const chatSession = logSession(baseUrl, apiKey, model, "./gen", history);
const response = await chatSession("What is my name?");
console.log("\nResponse:", response);

chat

Creates a stateless agent function that returns an async generator streaming events.

import chat from "cersei_rs/chat";
import { MSG_TXT, MSG_TOOL } from "cersei_rs/MSG";

const agent = chat(baseUrl, apiKey, model);

// Run task and stream events
const prompt = "Write a programmer joke in tmp.md";
const workingDir = "./gen";

for await (const [type, content, args] of agent(prompt, workingDir)) {
  switch (type) {
    case MSG_TXT:
      process.stdout.write(content);
      break;
    case MSG_TOOL:
      console.log(`\n[Tool]: ${content} ${args || ""}`);
      break;
  }
}

session

Creates an agent session that maintains chat history, returning a function that yields streaming events.

import session from "cersei_rs/session";
import { MSG_TXT } from "cersei_rs/MSG";

const history = [
  { role: "user", content: "Remember my name is Cersei" },
  { role: "assistant", content: "Got it, I'll remember that your name is Cersei." },
];

const chatSession = session(baseUrl, apiKey, model, "./gen", history);

for await (const [type, content] of chatSession("What is my name?")) {
  if (type === MSG_TXT) {
    process.stdout.write(content);
  }
}

License

MulanPSL-2.0

About

This library is developed by WebC.site.

WebC.site: A new paradigm of web development for AI

cersei_rs

• cersei_rs
  • 安装
  • API 使用说明
    • 常量定义
    • logChat (默认导出)
    • logSession
    • chat
    • session
  • 开源协议
  • 关于

Cersei 的 N-API 包装器，为 Node.js 环境提供智能体创建、流式对话会话，以及自动日志记录功能。

安装

npm install cersei_rs

API 使用说明

常量定义

从 "cersei_rs/MSG" 导入常量以识别流式输出中的事件类型：

• MSG_ERR (1): 异常错误事件。
• MSG_TOOL (2): 工具执行启动事件。
• MSG_TXT (3): 文本内容流式输出事件。
• MSG_THINK (4): 思考过程流式输出事件。

logChat (默认导出)

包装函数，能自动向 stdout 打印智能体执行日志（包含思考过程、工具调用和文本输出），并返回 Promise 解析为最终累积的文本响应。

import logChat from "cersei_rs";

const agent = logChat(baseUrl, apiKey, model);
const response = await agent("在 tmp.md 中写一个程序员的笑话", "./gen");
console.log("\n响应:", response);

logSession

包装函数，能自动向 stdout 打印维护对话历史的智能体执行日志，并返回 Promise 解析为最终累积的文本响应。

import logSession from "cersei_rs/logSession";

const history = [
  { role: "user", content: "记住我的名字叫 Cersei" },
  { role: "assistant", content: "好的，我已经记住了，您的名字是 Cersei。" },
];

const chatSession = logSession(baseUrl, apiKey, model, "./gen", history);
const response = await chatSession("我的名字叫什么？");
console.log("\n响应:", response);

chat

创建一个无状态的智能体执行函数，返回一个流式事件的异步生成器。

import chat from "cersei_rs/chat";
import { MSG_TXT, MSG_TOOL } from "cersei_rs/MSG";

const agent = chat(baseUrl, apiKey, model);

// 运行任务并以流式获取事件
const prompt = "在 tmp.md 中写一个程序员的笑话";
const workingDir = "./gen";

for await (const [type, content, args] of agent(prompt, workingDir)) {
  switch (type) {
    case MSG_TXT:
      process.stdout.write(content);
      break;
    case MSG_TOOL:
      console.log(`\n[工具]: ${content} ${args || ""}`);
      break;
  }
}

session

创建一个在会话上下文中维护对话历史的智能体函数，返回一个流式事件的异步生成器。

import session from "cersei_rs/session";
import { MSG_TXT } from "cersei_rs/MSG";

const history = [
  { role: "user", content: "记住我的名字叫 Cersei" },
  { role: "assistant", content: "好的，我已经记住了，您的名字是 Cersei。" },
];

const chatSession = session(baseUrl, apiKey, model, "./gen", history);

for await (const [type, content] of chatSession("我的名字叫什么？")) {
  if (type === MSG_TXT) {
    process.stdout.write(content);
  }
}

开源协议

MulanPSL-2.0

关于

本库由 WebC.site 开发。

WebC.site : 面向人工智能的网站开发新范式