@nickyzj2023/ai 1.4.4 → 1.5.0

package/dist/cli.mjs

@@ -1,10 +1,11 @@
-import { a as get_time_default, c as runAgent, i as get_weather_default, o as defineModel, r as loadMCPTools } from "./src-CuZ-6Jye.mjs";
+import { a as get_time_default, i as get_weather_default, o as runAgent, r as loadMCPTools, s as defineModel } from "./src-C1Zeveeb.mjs";
 import { compactStr, extractErrorMessage, isObject } from "@nickyzj2023/utils";
 import readline from "node:readline";
 import { mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import { homedir } from "node:os";
 import { dirname, join } from "node:path";
 import { MarkdownRenderer } from "@wterm/markdown";
+import { execFile } from "node:child_process";
 //#region src/utils/config.ts
 /**
 * 获取全局配置文件路径（Windows/macOS/Linux通用）
@@ -197,10 +198,110 @@ async function runSetup() {
 	}
 }
 //#endregion
+//#region src/utils/clipboard.ts
+/** 剪贴板图片的原始字节上限：超过则拒绝，避免把超大图片塞入请求体 */
+const MAX_CLIPBOARD_IMAGE_BYTES = 5242880;
+/**
+* 执行命令并原样取回stdout（二进制安全）
+* @param maxBuffer 允许的最大输出字节数
+*/
+const run = (command, args, maxBuffer) => {
+	return new Promise((resolve, reject) => {
+		execFile(command, args, {
+			encoding: "buffer",
+			timeout: 5e3,
+			maxBuffer
+		}, (error, stdout) => {
+			if (error) reject(error);
+			else resolve(stdout);
+		});
+	});
+};
+/** 组装成OpenAI接口能接受的data URL */
+const toDataUrl = (buffer, mime = "image/png") => ({
+	url: `data:${mime};base64,${buffer.toString("base64")}`,
+	bytes: buffer.length
+});
+/** Windows：用PowerShell把剪贴板里的图片转成base64后读出 */
+const readWindows = async () => {
+	const script = [
+		"Add-Type -AssemblyName System.Windows.Forms",
+		"Add-Type -AssemblyName System.Drawing",
+		"$img = [System.Windows.Forms.Clipboard]::GetImage()",
+		"if ($img) {",
+		"$ms = New-Object System.IO.MemoryStream",
+		"$img.Save($ms, [System.Drawing.Imaging.ImageFormat]::Png)",
+		"[Convert]::ToBase64String($ms.ToArray())",
+		"}"
+	].join("; ");
+	const base64 = (await run("powershell.exe", [
+		"-NoProfile",
+		"-Command",
+		script
+	], 67108864)).toString("utf8").trim();
+	return base64 ? toDataUrl(Buffer.from(base64, "base64")) : null;
+};
+/** Linux：Wayland优先，其次X11，最后用python3的tkinter兜底；命令未安装或剪贴板里不是图片时跳过 */
+const readLinux = async () => {
+	for (const candidate of [{
+		command: "wl-paste",
+		args: [
+			"--no-newline",
+			"--type",
+			"image/png"
+		]
+	}, {
+		command: "xclip",
+		args: [
+			"-selection",
+			"clipboard",
+			"-t",
+			"image/png",
+			"-o"
+		]
+	}]) try {
+		const stdout = await run(candidate.command, candidate.args, MAX_CLIPBOARD_IMAGE_BYTES * 2);
+		if (stdout.length === 0) continue;
+		return candidate.base64 ? toDataUrl(Buffer.from(stdout.toString("ascii").trim(), "base64")) : toDataUrl(stdout);
+	} catch {}
+	return null;
+};
+/**
+* 读取系统剪贴板里的图片
+* @returns 剪贴板里没有图片、或当前环境拿不到时返回null
+*/
+const readClipboardImage = async () => {
+	try {
+		switch (process.platform) {
+			case "win32": return await readWindows();
+			default: return await readLinux();
+		}
+	} catch {
+		return null;
+	}
+};
+//#endregion
+//#region src/utils/internalHelper.ts
+const formatter = new Intl.NumberFormat("en-US", {
+	notation: "compact",
+	maximumFractionDigits: 1
+});
+/** 把数字转换成千分位 */
+const humanizeNumber = (number) => {
+	return formatter.format(number);
+};
+//#endregion
 //#region src/interfaces/tui.ts
 var TUI = class {
 	rl = null;
 	onPrompt = null;
+	/**
+	* 已粘贴、等待随下次请求一起发出的图片
+	* TODO: 改为pendingMedias
+	*/
+	pendingImages = [];
+	/** 剪贴板读取是异步的，回车时要先等它结束，避免漏掉刚粘贴的图片 */
+	clipboardTask = Promise.resolve();
 	/** content专用的markdown渲染器 */
 	md = null;
 	/** 是否允许输入 */
@@ -222,27 +323,57 @@ var TUI = class {
 			input: process.stdin,
 			output: process.stdout
 		});
+		process.stdin.on("keypress", this.onKeypress);
 		this.prompting();
 		this.rl.on("close", () => {
+			process.stdin.off("keypress", this.onKeypress);
 			this.rl?.close();
 			this.rl = null;
 		});
 	}
+	/** 监听用户的特定按键 */
+	onKeypress = (_str, key) => {
+		if (this.isBusy) return;
+		if (key?.ctrl && key?.name === "v") this.clipboardTask = this.clipboardTask.then(() => this.attachImage());
+	};
 	/** 监听用户输入 */
 	prompting() {
 		if (this.isBusy) return;
 		this.rl?.question("> ", async (answer) => {
-			const input = answer.trim();
-			if (!input) {
+			await this.clipboardTask;
+			const text = answer.trim();
+			const images = this.pendingImages;
+			this.pendingImages = [];
+			if (!text && images.length === 0) {
 				this.prompting();
 				return;
 			}
 			this.isBusy = true;
-			await this.onPrompt?.(input);
+			await this.onPrompt?.({
+				text,
+				images
+			});
 			this.isBusy = false;
 			this.prompting();
 		});
 	}
+	/** 读取剪贴板图片，推入待发送列表 */
+	async attachImage() {
+		const image = await readClipboardImage();
+		if (!image) {
+			this.printNotice("[剪贴板中没有图片]");
+			return;
+		}
+		if (image.bytes > 5242880) {
+			this.printNotice(`[图片过大（${humanizeNumber(image.bytes)}B），已忽略]`);
+			return;
+		}
+		this.pendingImages.push({
+			type: "image_url",
+			image_url: { url: image.url }
+		});
+		this.printNotice(`[已粘贴图片：共${this.pendingImages.length}张]`);
+	}
 	/**
 	* 所有print方法调用前：
 	* 状态切换时打印换行，并记录当前状态
@@ -294,27 +425,33 @@ var TUI = class {
 		this.preparePrint("tool_result");
 		process.stdout.write(this.colorize(`[工具结果：${name}] ${_result}\n`, "90"));
 	}
-	formatter = new Intl.NumberFormat("en-US", {
-		notation: "compact",
-		maximumFractionDigits: 1
-	});
-	format(number) {
-		return this.formatter.format(number);
-	}
 	/** 打印轮次结束原因、token消耗 */
 	printFinish(finishReason, usage) {
 		this.preparePrint("done");
-		process.stdout.write(this.colorize(`[本轮结束：${finishReason}] ${usage ? `输入${this.format(usage.prompt_tokens)}token，输出${this.format(usage.completion_tokens)}token，总共${this.format(usage.total_tokens)}token` : ""}${finishReason === "stop" ? "\n\n" : "\n"}`, "90"));
+		process.stdout.write(this.colorize(`[本轮结束：${finishReason}] ${usage ? `输入${humanizeNumber(usage.prompt_tokens)}，输出${humanizeNumber(usage.completion_tokens)}，总共${humanizeNumber(usage.total_tokens)}` : ""}${usage?.speed ? `，速度${usage.speed}/秒` : ""}${finishReason === "stop" ? "\n\n" : "\n"}`, "90"));
 	}
-	/**
-	* 打印一行黄色状态提示（如“等待MCP工具加载完成…”）
-	*/
-	printStatus(message) {
-		process.stdout.write(this.colorize(message, "93"));
+	/** 打印一行黄字提示，并重绘当前输入行 */
+	printNotice(message) {
+		process.stdout.write(`${this.colorize(message, "93")}\n`);
+		this.rl?.prompt(true);
 	}
 };
 //#endregion
 //#region src/cli.ts
+/**
+* 组装用户消息的content
+* @remarks 有图片时必须用ContentPart[]（OpenAI多模态协议），纯文本则保持字符串
+*/
+const buildUserContent = (text, images) => {
+	if (images.length === 0) return text;
+	const content = [];
+	if (text) content.push({
+		type: "text",
+		text
+	});
+	content.push(...images);
+	return content;
+};
 /** 启动交互对话：配置来自全局配置文件，环境变量可临时覆盖 */
 const startChat = async () => {
 	const config = loadConfig();
@@ -328,14 +465,17 @@ const startChat = async () => {
 		mcpReady = true;
 		return tools;
 	});
-	const messages = [];
+	const messages = [{
+		role: "system",
+		content: `你是${model.model}，当前时间${(/* @__PURE__ */ new Date()).toLocaleString()}`
+	}];
 	let tools = null;
-	const tui = new TUI(async (input) => {
+	const tui = new TUI(async ({ text, images }) => {
 		messages.push({
 			role: "user",
-			content: input
+			content: buildUserContent(text, images)
 		});
-		if (!mcpReady) tui.printStatus("等待MCP工具加载完成…");
+		if (!mcpReady) tui.printNotice("MCP工具还未加载完成，请稍候……");
 		tools ??= [
 			get_weather_default,
 			get_time_default,

package/dist/index.d.mts

@@ -131,6 +131,8 @@ type Usage = {
   prompt_tokens: number;
   completion_tokens: number;
   total_tokens: number;
+  /** 每秒生成多少个token */
+  speed?: number;
   [key: string]: any;
 };
 type AgentEvent = LLMEvent | {
@@ -249,8 +251,15 @@ declare namespace Compact {
      */
     summarizeOptions?: Partial<Compact.SummarizeOptions>;
   };
-  type ReplacerOfToolResultContent = (content: Message["content"], options?: Record<string, any>) => Promise<string> | string;
-  type ReplacerOfMediaContent = (content: Message["content"], options?: Record<string, any>) => Promise<string> | string;
+  type ReplacerOfToolResultContent = (content: Message["content"], options?: {
+    model?: Model;
+    [key: string]: any;
+  }) => Promise<string> | string;
+  type ReplacerOfMediaContent = (content: Message["content"], options?: {
+    model?: Model;
+    modality?: Modality;
+    [key: string]: any;
+  }) => Promise<string> | string;
   type SummarizeOptions = {
     /** 指导大模型如何总结消息 */
     systemPrompt: string;

package/dist/index.mjs

@@ -1,2 +1,2 @@
-import { a as get_time_default, c as runAgent, i as get_weather_default, n as MCPRouter, o as defineModel, r as loadMCPTools, s as defineTool, t as compact } from "./src-CuZ-6Jye.mjs";
+import { a as get_time_default, c as defineTool, i as get_weather_default, n as MCPRouter, o as runAgent, r as loadMCPTools, s as defineModel, t as compact } from "./src-C1Zeveeb.mjs";
 export { MCPRouter, compact, defineModel, defineTool, get_time_default as getTime, get_weather_default as getWeather, loadMCPTools, runAgent };

package/dist/{src-CuZ-6Jye.mjs → src-C1Zeveeb.mjs}

@@ -91,64 +91,6 @@ async function* stream(model, messages, tools = []) {
 	};
 }
 //#endregion
-//#region src/agent.ts
-async function* runAgent(model, messages, tools) {
-	const toolMap = new Map(tools.map((tool) => [tool.function.name, tool]));
-	while (true) {
-		let content = "";
-		const toolCalls = [];
-		for await (const e of stream(model, messages, tools)) switch (e.type) {
-			case "reasoning_delta":
-				yield e;
-				break;
-			case "content_delta":
-				content += e.delta;
-				yield e;
-				break;
-			case "tool_call":
-				toolCalls.push({
-					id: e.id,
-					type: "function",
-					function: {
-						name: e.name,
-						arguments: e.args
-					}
-				});
-				yield e;
-				break;
-			case "done": yield e;
-		}
-		const message = {
-			role: "assistant",
-			content
-		};
-		if (toolCalls.length > 0) message.tool_calls = toolCalls;
-		messages.push(message);
-		if (toolCalls.length === 0) return;
-		for (const call of toolCalls) {
-			const { name, arguments: args } = call.function;
-			const tool = toolMap.get(name);
-			let result = "";
-			if (!tool) result = `不存在工具“${name}”`;
-			else {
-				const [error, response] = await to(tool.execute(JSON.parse(args)));
-				result = error ? `工具“${name}”执行出错：${error.message}` : JSON.stringify(response);
-			}
-			messages.push({
-				role: "tool",
-				tool_call_id: call.id,
-				content: result
-			});
-			yield {
-				type: "tool_result",
-				id: call.id,
-				name,
-				result
-			};
-		}
-	}
-}
-//#endregion
 //#region src/utils/helper.ts
 /**
 * 辅助定义一个POST /chat/completions支持的model参数
@@ -186,22 +128,22 @@ const defineTool = (name, description, properties, execute) => {
 		execute
 	};
 };
+const segmenter = new Intl.Segmenter([], { granularity: "word" });
+const estimateTextTokens = (text) => {
+	let words = 0;
+	let others = 0;
+	for (const seg of segmenter.segment(text)) if (seg.isWordLike) words++;
+	else others++;
+	return Math.ceil(words * 1.5 + others / 4);
+};
 /**
 * 根据上下文里的中/英文/多模态消息，估算出可能消耗的token
 * - 单词 ≈ 1.5token
-* - 标点/空白等非词字符每 4 个 ≈ 1token
+* - 标点/空白等非词字符每4个 ≈ 1token
 * - 图片/音频/视频/文件 ≈ 4096token（不好估算，取个较大的值）
 */
 const estimateTokens = (messages) => {
 	if (!messages?.length) return 0;
-	const segmenter = new Intl.Segmenter([], { granularity: "word" });
-	const estimateTextTokens = (text) => {
-		let words = 0;
-		let others = 0;
-		for (const seg of segmenter.segment(text)) if (seg.isWordLike) words++;
-		else others++;
-		return Math.ceil(words * 1.5 + others / 4);
-	};
 	return messages.reduce((acc, message) => {
 		const { content, tool_calls, ...metadata } = message;
 		if (typeof content === "string") acc += estimateTextTokens(content);
@@ -213,6 +155,85 @@ const estimateTokens = (messages) => {
 	}, 0);
 };
 //#endregion
+//#region src/agent.ts
+async function* runAgent(model, messages, tools) {
+	const toolMap = new Map(tools.map((tool) => [tool.function.name, tool]));
+	while (true) {
+		let content = "";
+		let reasoning = "";
+		const toolCalls = [];
+		const promptMessage = messages.slice(-1);
+		let startTime = 0;
+		for await (const e of stream(model, messages, tools)) {
+			startTime ||= Date.now();
+			switch (e.type) {
+				case "reasoning_delta":
+					reasoning += e.delta;
+					yield e;
+					break;
+				case "content_delta":
+					content += e.delta;
+					yield e;
+					break;
+				case "tool_call":
+					toolCalls.push({
+						id: e.id,
+						type: "function",
+						function: {
+							name: e.name,
+							arguments: e.args
+						}
+					});
+					yield e;
+					break;
+				case "error":
+					yield e;
+					break;
+				case "done":
+					if (!e.usage) {
+						const promptTokens = estimateTokens(promptMessage);
+						const completionTokens = estimateTextTokens(content + reasoning);
+						e.usage = {
+							prompt_tokens: promptTokens,
+							completion_tokens: completionTokens,
+							total_tokens: estimateTokens(messages) + promptTokens + completionTokens
+						};
+					}
+					e.usage.speed = Math.floor(e.usage.completion_tokens / ((Date.now() - startTime) / 1e3));
+					yield e;
+			}
+		}
+		const message = {
+			role: "assistant",
+			content
+		};
+		if (toolCalls.length > 0) message.tool_calls = toolCalls;
+		messages.push(message);
+		if (toolCalls.length === 0) return;
+		for (const call of toolCalls) {
+			const { name, arguments: args } = call.function;
+			const tool = toolMap.get(name);
+			let result = "";
+			if (!tool) result = `不存在工具“${name}”`;
+			else {
+				const [error, response] = await to(tool.execute(JSON.parse(args)));
+				result = error ? `工具“${name}”执行出错：${error.message}` : JSON.stringify(response);
+			}
+			messages.push({
+				role: "tool",
+				tool_call_id: call.id,
+				content: result
+			});
+			yield {
+				type: "tool_result",
+				id: call.id,
+				name,
+				result
+			};
+		}
+	}
+}
+//#endregion
 //#region src/tools/get-time.ts
 var get_time_default = defineTool("get_time", "查询指定时区的当前时间", { timezone: {
 	type: "string",
@@ -296,12 +317,11 @@ const isToolCalls = (message) => {
 * 校验多模态消息
 */
 const isMediaMessage = (message) => {
-	const MEDIA_TYPES = [
-		"image_url",
-		"input_audio",
-		"video_url"
-	];
-	return message && Array.isArray(message.content) && message.content.some((part) => MEDIA_TYPES.includes(part.type));
+	if (message && Array.isArray(message.content)) for (const part of message.content) {
+		if (part.type === "image_url") return "image";
+		if (part.type === "input_audio") return "audio";
+		if (part.type === "video_url") return "video";
+	}
 };
 /**
 * 查找assistant(tool_calls) + tool配对组范围
@@ -330,7 +350,8 @@ const defaultReplacerOfToolResultContent = async (content, options) => {
 		content: "请用一句话简述上面这条消息（直击要点，不要添加任何前言后语）"
 	}];
 	let simplifiedContent = "";
-	for await (const e of runAgent(model, messages, [])) if (e.type === "content_delta") simplifiedContent += e.delta;
+	if (!model) simplifiedContent = "工具结果已被消费";
+	else for await (const e of runAgent(model, messages, [])) if (e.type === "content_delta") simplifiedContent += e.delta;
 	else if (e.type === "error") throw new Error(e.message);
 	return simplifiedContent;
 };
@@ -352,7 +373,7 @@ const compactToolResults = async (compressible, options) => {
 };
 /** 默认的压缩多模态消息策略：让大模型精简消息内容 */
 const defaultReplacerOfMediaContent = async (content, options) => {
-	const { model } = options ?? {};
+	const { model, modality = "image" } = options ?? {};
 	const messages = [{
 		role: "user",
 		content
@@ -361,7 +382,8 @@ const defaultReplacerOfMediaContent = async (content, options) => {
 		content: "请用一句话简述上面这条消息（直击要点，不要添加任何前言后语）"
 	}];
 	let simplifiedContent = "";
-	for await (const e of runAgent(model, messages, [])) if (e.type === "content_delta") simplifiedContent += e.delta;
+	if (!model?.modalities?.includes(modality)) simplifiedContent = "媒体资源已被消费";
+	else for await (const e of runAgent(model, messages, [])) if (e.type === "content_delta") simplifiedContent += e.delta;
 	else if (e.type === "error") throw new Error(e.message);
 	return simplifiedContent;
 };
@@ -373,10 +395,16 @@ const compactMediaMessages = async (compressible, options) => {
 	if (!replacer && !model) return 0;
 	const _replacer = replacer || defaultReplacerOfMediaContent;
 	let count = 0;
-	for (const message of compressible) if (isMediaMessage(message)) {
-		const compacted = await _replacer(message.content, { model });
-		message.content = createXMLText("media", compacted);
-		count++;
+	for (const message of compressible) {
+		const mediaModality = isMediaMessage(message);
+		if (mediaModality) {
+			const compacted = await _replacer(message.content, {
+				model,
+				modality: mediaModality
+			});
+			message.content = createXMLText("media", compacted);
+			count++;
+		}
 	}
 	if (count > 0) logger(`压缩了${count}条多模态消息`);
 	return count;
@@ -484,4 +512,4 @@ const compact = Object.assign(async (messages, model, options) => {
 	discardMessagesUntil
 });
 //#endregion
-export { get_time_default as a, runAgent as c, get_weather_default as i, MCPRouter as n, defineModel as o, loadMCPTools as r, defineTool as s, compact as t };
+export { get_time_default as a, defineTool as c, get_weather_default as i, MCPRouter as n, runAgent as o, loadMCPTools as r, defineModel as s, compact as t };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nickyzj2023/ai",
-  "version": "1.4.4",
+  "version": "1.5.0",
   "description": "我的“pi”，参考了pi-from-scratch",
   "type": "module",
   "main": "./dist/index.mjs",
@@ -27,7 +27,7 @@
   },
   "devDependencies": {
     "@biomejs/biome": "^2.5.12",
-    "@types/node": "^26.4.1",
+    "@types/node": "^26.5.0",
     "tsdown": "^0.22.14",
     "typescript": "^7.0.2"
   },