jinzd-ai-cli 0.1.28 → 0.1.30

package/CLAUDE.md

@@ -343,6 +343,83 @@ const stdinLines = Array.isArray(rawStdin) ? rawStdin.map(String)
 - [x] **web_fetch DNS 解析时 SSRF 防护**（v0.1.25）：新增 `resolveAndCheck()` 函数，用 `dns.promises.lookup()` 预解析域名，检查结果 IP 是否为私有地址。初始 URL 和 redirect 目标均校验。
 - [ ] **`persistentCwd` 全局状态**：bash 工具的当前工作目录是模块级全局变量，多 session 并发时可能串扰。现阶段单 session REPL 无影响，GUI 多会话扩展时需重构为 per-session 状态。
 
+## 本轮开发完成记录（2026-03-01，v0.1.26 → v0.1.30）
+
+### 安全修复续篇：低危 L2–L7
+
+**L2**（`src/tools/builtin/web-fetch.ts`）：`htmlToText()` 函数新增 200 KB HTML 大小上限（`HTML_REGEX_LIMIT = 200_000`），超出则截断后再做正则替换，防止恶意大 HTML 导致正则性能崩溃。
+
+**L3**（`src/session/session-manager.ts`）：新增 `safeDate(value: unknown): Date` 辅助函数，对无效日期字符串返回 `new Date(0)` 而非 `Invalid Date`，防止 `listSessions()` 和 `searchMessages()` 的日期比较静默失败。
+
+**L4**（`src/tools/builtin/ask-user.ts` + `src/tools/builtin/google-search.ts`）：为模块级全局上下文对象（`askUserContext` / `googleSearchContext`）添加架构说明注释，提示 GUI 多会话扩展时需重构为 per-session 状态。
+
+**L5**（`src/config/schema.ts`）：为 `timeouts` 字段补充三层优先级文档注释：
+1. `modelParams[modelId].timeout` — 最高
+2. `timeouts[providerId]` — provider 级默认
+3. 内置 provider 硬编码默认值 — 最低兜底
+
+**L6/L7**：确认已安全（`call.arguments['path']` 已用 `String(... ?? '')` 兜底；主循环 `line` handler 已有 try/catch）。
+
+---
+
+### 新增功能 1：多模态图片输入（P0）
+
+**背景**：用户希望通过 `@image.png 描述这张图` 语法将图片发送给各 AI Provider。
+
+**类型层**（已有，无需改动）：`src/core/types.ts` 的 `ImageContentPart { type: 'image_url', image_url: { url: 'data:mime;base64,...' } }` 与 OpenAI 格式完全兼容。
+
+**Claude Provider**（`src/providers/claude.ts`）：
+- 新增 `contentToClaudeParts()` 私有方法：将内部 `image_url` 格式转换为 Anthropic SDK 期望的 `{ type: 'image', source: { type: 'base64', media_type, data } }` 格式
+- 应用到 `chat()`、`chatStream()`、`chatWithTools()` 的消息构建
+
+**Gemini Provider**（`src/providers/gemini.ts`）：
+- 移除错误的 `getContentText()` 调用（会丢弃图片）
+- 新增 `contentToGeminiParts()` 私有方法：转换为 Gemini SDK 格式 `{ inlineData: { mimeType, data } }`
+- 修复 `toGeminiHistory()`、`chat()`、`chatStream()`、`chatWithTools()` 的消息构建
+- `chatWithTools()` 中变量 `lastMessage: string` 重命名为 `lastMsgParts: Part[]`，历史嵌入改为 `{ role: 'user', parts: lastMsgParts }`
+
+**OpenAI 兼容 Provider**：已就绪，格式一致，无需修改。
+
+**REPL 层**（`src/repl/repl.ts`）：
+- 新增常量 `MAX_IMAGE_BYTES = 10 * 1024 * 1024`（10 MB）
+- `parseAtReferences()` 图片读取前用 `statSync` 校验大小，超限时加入 `refs` 类型为 `'toolarge'`，不内联
+- `handleChat()` 新增 `toolarge` 分支：打印黄色警告 `⚠ Image too large (> 10 MB): <path>`
+- 修复 `getVisionModelHint()`：Claude/Gemini 返回 `null`（原生支持，无需警告）；DeepSeek 显示明确不支持提示；zhipu 推荐 `glm-4.6v`；kimi `moonshot-v1-*` 推荐对应 vision 版本
+
+---
+
+### 新增功能 2：Escape/Ctrl+C 中断 AI 流式输出（P0）
+
+**背景**：流式生成期间无法中断，必须等待 AI 返回完整内容。
+
+**架构设计**：两个流式生成入口——`handleChatSimple()` 和 `handleChatWithTools()` 的 tee streaming 分支（`save_last_response` 工具）——均支持中断。主路径（`chatWithTools()` → `renderResponse()`）为非流式，暂不支持。
+
+**`src/core/types.ts`**：`ChatRequest` 新增 `signal?: AbortSignal`，透传给 Provider API 调用。
+
+**Provider 层**：
+- `claude.ts`：`messages.stream()` 第二参数传入 `{ signal: request.signal }`（Anthropic SDK 支持）
+- `openai-compatible.ts`：流式 `create()` 第二参数传入 `{ signal: request.signal }`（OpenAI SDK 支持）
+- `gemini.ts`：生成器循环开头检查 `if (request.signal?.aborted) break`（兼容性保险）
+
+**`src/repl/renderer.ts`**：`renderStream()` 的 `options` 新增 `signal?: AbortSignal`：
+- `for await` 循环开头检查 `signal.aborted` → 标记 `interrupted = true` 并 `break`
+- 捕获 SDK 抛出的 `AbortError`（name 检测）→ 同样标记 `interrupted`
+- 中断时 `flushBuf()` 输出残留缓冲，打印灰色 `[interrupted]` 提示
+- 返回值不变（`{ content, usage, tokensShown }`），`content` 为已生成的部分内容
+
+**`src/repl/repl.ts`**：
+- 新增类属性 `streamAbortController: AbortController | null` + `_escHandler: ((d: Buffer) => void) | null`
+- 新增 `setupStreamInterrupt()` 方法：创建 `AbortController`，`process.stdin.resume()`（绕过 `rl.pause()` 暂停），注册原始字节监听器检测纯 ESC（`0x1b`，单字节，区别于 ESC 序列）和 Ctrl+C（`0x03`）
+- 新增 `teardownStreamInterrupt()` 方法：移除监听器，`process.stdin.pause()`，清空引用
+- SIGINT 处理器最顶部新增分支：`streamAbortController` 非空时 `abort()` 并 return，不退出程序
+- `handleChatSimple()` 流式分支和 tee streaming 分支均用 `setupStreamInterrupt()`/`teardownStreamInterrupt()` 包裹 + `signal` 传入
+
+**版本与收尾**
+- `src/core/constants.ts`：VERSION `0.1.26` → `0.1.30`（合并前几次 bump）
+- `package.json`：version `0.1.29` → `0.1.30`
+
+---
+
 ## 本轮开发完成记录（2026-03-01，v0.1.25 → v0.1.26）
 
 ### 新增功能：Context 自动管理 + 测试报告 + 脚手架
@@ -871,8 +948,8 @@ const stdinLines = Array.isArray(rawStdin) ? rawStdin.map(String)
 ### ❌ 缺失功能路线图（新增）
 
 #### P0 — 核心竞争力缺口
-- [ ] **中断生成**：`Escape` 键立即停止 AI 流式输出（当前无法中断，必须等待完成）
-- [ ] **多模态输入（图片）**：支持粘贴/引用图片路径作为输入（当前纯文本）
+- [x] **中断生成**（v0.1.30）：`Escape` 键或 `Ctrl+C` 立即停止 AI 流式输出，不退出程序，显示 `[interrupted]` 后恢复提示符；已生成内容保留到 session
+- [x] **多模态输入（图片）**（v0.1.30）：`@image.png 描述这张图` 语法发送图片；Claude/Gemini/OpenAI 兼容格式自动转换；10MB 大小限制；`getVisionModelHint()` 正确识别 Claude/Gemini 原生支持视觉
 - [ ] **`/add-dir` 命令**：运行时动态添加目录到上下文
 - [ ] **并行工具调用**：AI 一次返回多个工具同时执行（当前为分组串行）
 

package/dist/{chunk-3EJWGNHV.js → chunk-IW6VVPO4.js}

@@ -8,7 +8,7 @@ import { platform } from "os";
 import chalk from "chalk";
 
 // src/core/constants.ts
-var VERSION = "0.1.26";
+var VERSION = "0.1.30";
 var APP_NAME = "ai-cli";
 var CONFIG_DIR_NAME = ".aicli";
 var CONFIG_FILE_NAME = "config.json";

package/dist/index.js

@@ -27,7 +27,7 @@ import {
   SUBAGENT_MAX_ROUNDS_LIMIT,
   VERSION,
   runTestsTool
-} from "./chunk-3EJWGNHV.js";
+} from "./chunk-IW6VVPO4.js";
 
 // src/index.ts
 import { program } from "commander";
@@ -341,11 +341,41 @@ var ClaudeProvider = class extends BaseProvider {
       baseURL: options?.baseUrl
     });
   }
+  /**
+   * 将内部 MessageContentPart[] 格式转换为 Anthropic SDK 期望的 ContentBlockParam[]。
+   *
+   * 内部格式（OpenAI 兼容）：{ type: 'image_url', image_url: { url: 'data:mime;base64,...' } }
+   * Anthropic 格式：{ type: 'image', source: { type: 'base64', media_type: ..., data: ... } }
+   *
+   * 文本 part 和纯字符串保持不变。
+   */
+  contentToClaudeParts(content) {
+    if (typeof content === "string") return content;
+    const blocks = [];
+    for (const part of content) {
+      if (part.type === "text") {
+        blocks.push({ type: "text", text: part.text });
+      } else if (part.type === "image_url") {
+        const match = part.image_url.url.match(/^data:([^;]+);base64,(.+)$/);
+        if (match) {
+          blocks.push({
+            type: "image",
+            source: {
+              type: "base64",
+              media_type: match[1],
+              data: match[2]
+            }
+          });
+        }
+      }
+    }
+    return blocks.length > 0 ? blocks : "";
+  }
   async chat(request) {
     try {
       const messages = request.messages.filter((m) => m.role !== "system").map((m) => ({
         role: m.role,
-        content: m.content
+        content: this.contentToClaudeParts(m.content)
       }));
       const response = await this.client.messages.create({
         model: request.model,
@@ -372,14 +402,14 @@ var ClaudeProvider = class extends BaseProvider {
     try {
       const messages = request.messages.filter((m) => m.role !== "system").map((m) => ({
         role: m.role,
-        content: m.content
+        content: this.contentToClaudeParts(m.content)
       }));
       const stream = this.client.messages.stream({
         model: request.model,
         messages,
         system: request.systemPrompt,
         max_tokens: request.maxTokens ?? 8192
-      });
+      }, { signal: request.signal });
       for await (const event of stream) {
         if (event.type === "content_block_delta" && event.delta.type === "text_delta") {
           yield { delta: event.delta.text, done: false };
@@ -412,7 +442,7 @@ var ClaudeProvider = class extends BaseProvider {
           required: Object.entries(t.parameters).filter(([, s]) => s.required).map(([k]) => k)
         }
       }));
-      const baseMessages = request.messages.filter((m) => m.role !== "system").map((m) => ({ role: m.role, content: m.content }));
+      const baseMessages = request.messages.filter((m) => m.role !== "system").map((m) => ({ role: m.role, content: this.contentToClaudeParts(m.content) }));
       const extraMessages = request._extraMessages ?? [];
       const allMessages = [...baseMessages, ...extraMessages];
       const response = await this.client.messages.create({
@@ -490,14 +520,6 @@ var ClaudeProvider = class extends BaseProvider {
 
 // src/providers/gemini.ts
 import { GoogleGenerativeAI } from "@google/generative-ai";
-
-// src/core/types.ts
-function getContentText(content) {
-  if (typeof content === "string") return content;
-  return content.filter((p) => p.type === "text").map((p) => p.text).join("");
-}
-
-// src/providers/gemini.ts
 var GeminiProvider = class extends BaseProvider {
   client;
   /** 自定义 base URL（来自 config.customBaseUrls.gemini），用于配置代理或镜像站 */
@@ -555,10 +577,35 @@ var GeminiProvider = class extends BaseProvider {
   get reqOpts() {
     return this.baseUrl ? { baseUrl: this.baseUrl } : void 0;
   }
+  /**
+   * 将内部 MessageContentPart[] 格式转换为 Gemini SDK 期望的 Part[]。
+   *
+   * 内部格式（OpenAI 兼容）：{ type: 'image_url', image_url: { url: 'data:mime;base64,...' } }
+   * Gemini 格式：{ inlineData: { mimeType: 'image/png', data: 'base64...' } }
+   *
+   * 空字符串保留为 [{ text: '' }] 以避免 Gemini SDK 报空 parts 错误。
+   */
+  contentToGeminiParts(content) {
+    if (typeof content === "string") {
+      return [{ text: content }];
+    }
+    const parts = [];
+    for (const p of content) {
+      if (p.type === "text") {
+        if (p.text.trim()) parts.push({ text: p.text });
+      } else if (p.type === "image_url") {
+        const match = p.image_url.url.match(/^data:([^;]+);base64,(.+)$/);
+        if (match) {
+          parts.push({ inlineData: { mimeType: match[1], data: match[2] } });
+        }
+      }
+    }
+    return parts.length > 0 ? parts : [{ text: "" }];
+  }
   toGeminiHistory(messages) {
     return messages.filter((m) => m.role !== "system").map((m) => ({
       role: m.role === "assistant" ? "model" : "user",
-      parts: [{ text: getContentText(m.content) }]
+      parts: this.contentToGeminiParts(m.content)
     }));
   }
   async chat(request) {
@@ -571,9 +618,11 @@ var GeminiProvider = class extends BaseProvider {
         systemInstruction: request.systemPrompt
       }, this.reqOpts);
       const history = this.toGeminiHistory(request.messages.slice(0, -1));
-      const lastMessage = getContentText(request.messages[request.messages.length - 1].content);
+      const lastMsgParts = this.contentToGeminiParts(
+        request.messages[request.messages.length - 1].content
+      );
       const chat = genModel.startChat({ history });
-      const result = await chat.sendMessage(lastMessage);
+      const result = await chat.sendMessage(lastMsgParts);
       const meta = result.response.usageMetadata;
       return {
         content: result.response.text(),
@@ -595,10 +644,13 @@ var GeminiProvider = class extends BaseProvider {
         systemInstruction: request.systemPrompt
       }, this.reqOpts);
       const history = this.toGeminiHistory(request.messages.slice(0, -1));
-      const lastMessage = getContentText(request.messages[request.messages.length - 1].content);
+      const lastMsgParts = this.contentToGeminiParts(
+        request.messages[request.messages.length - 1].content
+      );
       const chat = genModel.startChat({ history });
-      const result = await chat.sendMessageStream(lastMessage);
+      const result = await chat.sendMessageStream(lastMsgParts);
       for await (const chunk of result.stream) {
+        if (request.signal?.aborted) break;
         yield { delta: chunk.text(), done: false };
       }
       const finalResponse = await result.response;
@@ -642,7 +694,7 @@ var GeminiProvider = class extends BaseProvider {
         }))
       }];
       const baseHistory = this.toGeminiHistory(request.messages.slice(0, -1));
-      const lastMessage = getContentText(
+      const lastMsgParts = this.contentToGeminiParts(
         request.messages[request.messages.length - 1].content
       );
       const extraHistory = request._extraMessages ?? [];
@@ -650,12 +702,12 @@ var GeminiProvider = class extends BaseProvider {
       let msgToSend;
       if (extraHistory.length === 0) {
         fullHistory = [...baseHistory];
-        msgToSend = lastMessage;
+        msgToSend = lastMsgParts;
       } else {
         const lastFnResponse = extraHistory[extraHistory.length - 1];
         fullHistory = [
           ...baseHistory,
-          { role: "user", parts: [{ text: lastMessage }] },
+          { role: "user", parts: lastMsgParts },
           ...extraHistory.slice(0, -1)
         ];
         msgToSend = lastFnResponse.parts;
@@ -806,7 +858,8 @@ var OpenAICompatibleProvider = class extends BaseProvider {
         stream_options: { include_usage: true },
         ...request.thinking ? { thinking: { type: "enabled" } } : {}
       }, {
-        timeout: request.timeout ?? this.defaultTimeout
+        timeout: request.timeout ?? this.defaultTimeout,
+        signal: request.signal
       });
       for await (const chunk of stream) {
         const choice = chunk.choices[0];
@@ -1327,6 +1380,12 @@ import { readFileSync as readFileSync2, writeFileSync as writeFileSync2, existsS
 import { join as join2 } from "path";
 import { v4 as uuidv4 } from "uuid";
 
+// src/core/types.ts
+function getContentText(content) {
+  if (typeof content === "string") return content;
+  return content.filter((p) => p.type === "text").map((p) => p.text).join("");
+}
+
 // src/session/session.ts
 var Session = class _Session {
   id;
@@ -1793,19 +1852,36 @@ var Renderer = class {
       if (out) process.stdout.write(out);
       buf = "";
     };
-    for await (const chunk of stream) {
-      if (chunk.usage) {
-        usage = chunk.usage;
+    let interrupted = false;
+    try {
+      for await (const chunk of stream) {
+        if (options?.signal?.aborted) {
+          interrupted = true;
+          break;
+        }
+        if (chunk.usage) {
+          usage = chunk.usage;
+        }
+        if (chunk.done) {
+          flushBuf();
+          break;
+        }
+        if (!chunk.delta) continue;
+        fullContent += chunk.delta;
+        buf += chunk.delta;
+        if (fileStream) fileStream.write(chunk.delta);
+        flushBuf();
       }
-      if (chunk.done) {
+    } catch (err) {
+      if (err?.name === "AbortError") {
+        interrupted = true;
         flushBuf();
-        break;
+      } else {
+        throw err;
       }
-      if (!chunk.delta) continue;
-      fullContent += chunk.delta;
-      buf += chunk.delta;
-      if (fileStream) fileStream.write(chunk.delta);
-      flushBuf();
+    }
+    if (interrupted) {
+      process.stdout.write(chalk.dim(" [interrupted]\n"));
     }
     let tokensShown = false;
     if (options?.showTokens) {
@@ -3199,7 +3275,7 @@ ${text}
       description: "Run project tests and show structured report",
       usage: "/test [command|filter]",
       async execute(args, _ctx) {
-        const { executeTests } = await import("./run-tests-MKMB3TXE.js");
+        const { executeTests } = await import("./run-tests-VVR5SMST.js");
         const argStr = args.join(" ").trim();
         let testArgs = {};
         if (argStr) {
@@ -6931,6 +7007,7 @@ var IMAGE_MIME = {
   ".gif": "image/gif",
   ".webp": "image/webp"
 };
+var MAX_IMAGE_BYTES = 10 * 1024 * 1024;
 function parseAtReferences(input2, cwd) {
   const atPattern = /@(?:"([^"]+)"|'([^']+)'|(\S+))/g;
   const refs = [];
@@ -6947,6 +7024,11 @@ function parseAtReferences(input2, cwd) {
       continue;
     }
     if (mime) {
+      const fileSize = statSync6(absPath).size;
+      if (fileSize > MAX_IMAGE_BYTES) {
+        refs.push({ path: rawPath, type: "toolarge" });
+        continue;
+      }
       const data = readFileSync13(absPath).toString("base64");
       imageParts.push({
         type: "image_url",
@@ -7039,6 +7121,10 @@ var Repl = class {
   /** 技能管理器 */
   skillManager = null;
   customCommandManager = null;
+  /** 流式生成中断控制器（ESC/Ctrl+C 中断时使用） */
+  streamAbortController = null;
+  /** ESC 键监听器引用（用于 removeListener 时取消注册） */
+  _escHandler = null;
   /**
    * 交互式列表选择器进行中标志。
    * 与 toolExecutor.confirming 类似：主循环 line handler 在此为 true 时忽略 line 事件，
@@ -7453,6 +7539,10 @@ ${response.content.trim()}
       }
     }
     this.rl.on("SIGINT", () => {
+      if (this.streamAbortController) {
+        this.streamAbortController.abort();
+        return;
+      }
       if (this.toolExecutor.confirming) {
         this.toolExecutor.cancelConfirm();
         return;
@@ -7516,6 +7606,9 @@ ${response.content.trim()}
     for (const ref of refs) {
       if (ref.type === "notfound") {
         process.stdout.write(chalk10.yellow(`  \u26A0 File not found: ${ref.path}
+`));
+      } else if (ref.type === "toolarge") {
+        process.stdout.write(chalk10.yellow(`  \u26A0 Image too large (> 10 MB): ${ref.path}
 `));
       } else if (ref.type === "image") {
         process.stdout.write(chalk10.dim(`  \u{1F4CE} Image: ${ref.path}
@@ -7571,6 +7664,8 @@ ${response.content.trim()}
   getVisionModelHint() {
     const model = this.currentModel;
     const provider = this.currentProvider;
+    if (provider === "claude") return null;
+    if (provider === "gemini") return null;
     if (model.includes("vision") || model.includes("vl") || model.endsWith(".6v") || // glm-4.6v
     model.endsWith("-v")) return null;
     if (provider === "kimi") {
@@ -7583,6 +7678,9 @@ ${response.content.trim()}
     if (provider === "zhipu") {
       return `model "${model}" does not support images. Use /model glm-4.6v`;
     }
+    if (provider === "deepseek") {
+      return `model "${model}" does not support vision. DeepSeek standard models are text-only.`;
+    }
     return `model "${model}" may not support image input`;
   }
   /**
@@ -7782,36 +7880,71 @@ ${response.content.trim()}
   shouldShowTokens() {
     return this.config.get("ui").showTokenCount;
   }
+  /**
+   * 流式生成开始前调用：创建 AbortController，监听 ESC 键（0x1b）和 Ctrl+C（0x03）原始字节。
+   * rl.pause() 会暂停 stdin，这里临时恢复以接收原始键盘字节。
+   */
+  setupStreamInterrupt() {
+    const ac = new AbortController();
+    this.streamAbortController = ac;
+    process.stdin.resume();
+    const handler = (data) => {
+      if (data[0] === 27 && data.length === 1 || data[0] === 3) {
+        ac.abort();
+      }
+    };
+    this._escHandler = handler;
+    process.stdin.on("data", handler);
+    return ac;
+  }
+  /**
+   * 流式生成结束后调用（finally 块中）：清理 ESC 监听器，还原 stdin 暂停状态。
+   */
+  teardownStreamInterrupt() {
+    if (this._escHandler) {
+      process.stdin.removeListener("data", this._escHandler);
+      this._escHandler = null;
+    }
+    process.stdin.pause();
+    this.streamAbortController = null;
+  }
   async handleChatSimple(provider, messages) {
     const session = this.sessions.current;
     const useStreaming = this.config.get("ui").streaming;
     const modelParams = this.getModelParams();
     if (useStreaming) {
-      const stream = provider.chatStream({
-        messages,
-        model: this.currentModel,
-        systemPrompt: this.buildCurrentSystemPrompt(),
-        stream: true,
-        temperature: modelParams.temperature,
-        maxTokens: modelParams.maxTokens,
-        timeout: modelParams.timeout,
-        thinking: modelParams.thinking
-      });
-      const showTokens = this.shouldShowTokens();
-      const { content, usage, tokensShown } = await this.renderer.renderStream(stream, {
-        showTokens,
-        sessionTotal: showTokens ? { ...this.sessionTokenUsage } : void 0
-      });
-      lastResponseStore.content = content;
-      session.addMessage({ role: "assistant", content, timestamp: /* @__PURE__ */ new Date() });
-      this.events.emit("message.after", { content });
-      if (usage) {
-        this.sessionTokenUsage.inputTokens += usage.inputTokens;
-        this.sessionTokenUsage.outputTokens += usage.outputTokens;
-        session.addTokenUsage(usage);
-        if (showTokens && !tokensShown) {
-          this.renderer.renderUsage(usage, this.sessionTokenUsage);
+      const ac = this.setupStreamInterrupt();
+      try {
+        const stream = provider.chatStream({
+          messages,
+          model: this.currentModel,
+          systemPrompt: this.buildCurrentSystemPrompt(),
+          stream: true,
+          temperature: modelParams.temperature,
+          maxTokens: modelParams.maxTokens,
+          timeout: modelParams.timeout,
+          thinking: modelParams.thinking,
+          signal: ac.signal
+        });
+        const showTokens = this.shouldShowTokens();
+        const { content, usage, tokensShown } = await this.renderer.renderStream(stream, {
+          showTokens,
+          sessionTotal: showTokens ? { ...this.sessionTokenUsage } : void 0,
+          signal: ac.signal
+        });
+        lastResponseStore.content = content;
+        session.addMessage({ role: "assistant", content, timestamp: /* @__PURE__ */ new Date() });
+        this.events.emit("message.after", { content });
+        if (usage) {
+          this.sessionTokenUsage.inputTokens += usage.inputTokens;
+          this.sessionTokenUsage.outputTokens += usage.outputTokens;
+          session.addTokenUsage(usage);
+          if (showTokens && !tokensShown) {
+            this.renderer.renderUsage(usage, this.sessionTokenUsage);
+          }
         }
+      } finally {
+        this.teardownStreamInterrupt();
       }
     } else {
       const spinner = this.renderer.showSpinner("Thinking...");
@@ -7913,46 +8046,52 @@ ${response.content.trim()}
           const saveToFile = String(saveLastResponseCall.arguments["path"] ?? "");
           if (!saveToFile) {
           } else {
-            const genStream = provider.chatStream({
-              messages: apiMessages,
-              model: this.currentModel,
-              systemPrompt,
-              stream: true,
-              temperature: modelParams.temperature,
-              maxTokens: modelParams.maxTokens,
-              timeout: modelParams.timeout,
-              thinking: modelParams.thinking,
-              ...extraMessages.length > 0 ? { _extraMessages: extraMessages } : {}
-            });
-            const teeShowTokens = this.shouldShowTokens();
-            const { content: genContent, usage: genUsage, tokensShown: teeTokShown } = await this.renderer.renderStream(
-              genStream,
-              { saveToFile, showTokens: teeShowTokens, sessionTotal: teeShowTokens ? { ...this.sessionTokenUsage } : void 0 }
-            );
-            lastResponseStore.content = genContent;
-            if (genUsage) {
-              roundUsage.inputTokens += genUsage.inputTokens;
-              roundUsage.outputTokens += genUsage.outputTokens;
-            }
-            session.addMessage({ role: "assistant", content: genContent, timestamp: /* @__PURE__ */ new Date() });
-            this.events.emit("message.after", { content: genContent });
-            const lines = genContent.split("\n").length;
-            const bytes = Buffer.byteLength(genContent, "utf-8");
-            const syntheticResults = result.toolCalls.map((tc) => ({
-              callId: tc.id,
-              content: tc.name === "save_last_response" ? `File saved: ${saveToFile} (${lines} lines, ${bytes} bytes)` : `[skipped: file already saved by tee streaming]`,
-              isError: false
-            }));
-            const reasoningContent2 = "reasoningContent" in result ? result.reasoningContent : void 0;
-            const newMsgs2 = provider.buildToolResultMessages(result.toolCalls, syntheticResults, reasoningContent2);
-            extraMessages.push(...newMsgs2);
-            if (roundUsage.inputTokens > 0 || roundUsage.outputTokens > 0) {
-              this.sessionTokenUsage.inputTokens += roundUsage.inputTokens;
-              this.sessionTokenUsage.outputTokens += roundUsage.outputTokens;
-              session.addTokenUsage(roundUsage);
-              if (teeShowTokens && !teeTokShown) {
-                this.renderer.renderUsage(roundUsage, this.sessionTokenUsage);
+            const teeAc = this.setupStreamInterrupt();
+            try {
+              const genStream = provider.chatStream({
+                messages: apiMessages,
+                model: this.currentModel,
+                systemPrompt,
+                stream: true,
+                temperature: modelParams.temperature,
+                maxTokens: modelParams.maxTokens,
+                timeout: modelParams.timeout,
+                thinking: modelParams.thinking,
+                signal: teeAc.signal,
+                ...extraMessages.length > 0 ? { _extraMessages: extraMessages } : {}
+              });
+              const teeShowTokens = this.shouldShowTokens();
+              const { content: genContent, usage: genUsage, tokensShown: teeTokShown } = await this.renderer.renderStream(
+                genStream,
+                { saveToFile, showTokens: teeShowTokens, sessionTotal: teeShowTokens ? { ...this.sessionTokenUsage } : void 0, signal: teeAc.signal }
+              );
+              lastResponseStore.content = genContent;
+              if (genUsage) {
+                roundUsage.inputTokens += genUsage.inputTokens;
+                roundUsage.outputTokens += genUsage.outputTokens;
+              }
+              session.addMessage({ role: "assistant", content: genContent, timestamp: /* @__PURE__ */ new Date() });
+              this.events.emit("message.after", { content: genContent });
+              const lines = genContent.split("\n").length;
+              const bytes = Buffer.byteLength(genContent, "utf-8");
+              const syntheticResults = result.toolCalls.map((tc) => ({
+                callId: tc.id,
+                content: tc.name === "save_last_response" ? `File saved: ${saveToFile} (${lines} lines, ${bytes} bytes)` : `[skipped: file already saved by tee streaming]`,
+                isError: false
+              }));
+              const reasoningContent2 = "reasoningContent" in result ? result.reasoningContent : void 0;
+              const newMsgs2 = provider.buildToolResultMessages(result.toolCalls, syntheticResults, reasoningContent2);
+              extraMessages.push(...newMsgs2);
+              if (roundUsage.inputTokens > 0 || roundUsage.outputTokens > 0) {
+                this.sessionTokenUsage.inputTokens += roundUsage.inputTokens;
+                this.sessionTokenUsage.outputTokens += roundUsage.outputTokens;
+                session.addTokenUsage(roundUsage);
+                if (teeShowTokens && !teeTokShown) {
+                  this.renderer.renderUsage(roundUsage, this.sessionTokenUsage);
+                }
               }
+            } finally {
+              this.teardownStreamInterrupt();
             }
             return;
           }

package/dist/{run-tests-MKMB3TXE.js → run-tests-VVR5SMST.js}

@@ -2,7 +2,7 @@
 import {
   executeTests,
   runTestsTool
-} from "./chunk-3EJWGNHV.js";
+} from "./chunk-IW6VVPO4.js";
 export {
   executeTests,
   runTestsTool

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jinzd-ai-cli",
-  "version": "0.1.28",
+  "version": "0.1.30",
   "description": "Cross-platform REPL-style AI CLI with multi-provider support",
   "type": "module",
   "main": "./dist/index.js",