wtagent 0.1.0-alpha.3 → 0.1.0-alpha.4

package/README.md

@@ -35,14 +35,14 @@ WTAgent 把 GPT 网页聊天连接到本地工具，提供类似 Codex 的 CLI A
 需要 Node.js 20.17+ 和 Chrome/Chromium。
 
 ```bash
-npm install -g wtagent@alpha --registry=https://registry.npmjs.org/
+npm install -g wtagent --registry=https://registry.npmjs.org/
 wtagent login
 ```
 
 Windows PowerShell：
 
 ```powershell
-npm install -g wtagent@alpha --registry=https://registry.npmjs.org/
+npm install -g wtagent --registry=https://registry.npmjs.org/
 wtagent doctor
 wtagent login
 ```
@@ -54,6 +54,8 @@ cd my-project
 wtagent
 ```
 
+启动后先选择 ChatGPT 模式：`Pro` 会为新对话选择 Pro，`Current` 会保留网页当前设置且不执行模型切换。非交互调用默认使用 `Current`；需要 Pro 时可显式传入 `--mode Pro`。
+
 也可以直接附带任务：
 
 ```bash
@@ -100,14 +102,14 @@ Boundaries:
 Requires Node.js 20.17+ and Chrome/Chromium.
 
 ```bash
-npm install -g wtagent@alpha --registry=https://registry.npmjs.org/
+npm install -g wtagent --registry=https://registry.npmjs.org/
 wtagent login
 ```
 
 Windows PowerShell:
 
 ```powershell
-npm install -g wtagent@alpha --registry=https://registry.npmjs.org/
+npm install -g wtagent --registry=https://registry.npmjs.org/
 wtagent doctor
 wtagent login
 ```
@@ -119,6 +121,8 @@ cd my-project
 wtagent
 ```
 
+At startup, choose a ChatGPT mode: `Pro` selects Pro for the new conversation, while `Current` keeps the current web setting without changing it. Non-interactive runs default to `Current`; pass `--mode Pro` to opt in explicitly.
+
 Or provide the first task directly:
 
 ```bash

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wtagent",
-  "version": "0.1.0-alpha.3",
+  "version": "0.1.0-alpha.4",
   "description": "Turn your own web AI session into a local tool-using agent.",
   "repository": {
     "type": "git",

package/src/browser/chatgpt-web-adapter.js

@@ -5,6 +5,8 @@ import { discoverChromeExecutable } from "../platform/chrome-discovery.js";
 import { ensureDirectory } from "../platform/paths.js";
 import { BrowserAdapterError } from "../shared/errors.js";
 import {
+  chooseModeOption,
+  labelMatchesToken,
   runModeSelection,
   slugMatchesToken,
   normalizeToken,
@@ -236,6 +238,15 @@ export class ChatGPTWebAdapter {
     const page = this.page;
     return {
       alreadyOnMode: async (requested) => {
+        const token = normalizeToken(requested);
+        const switcher = await this.#findModelSwitcher();
+        const switcherLabel = switcher
+          ? await switcher.innerText().catch(() => "")
+          : "";
+        if (labelMatchesToken(switcherLabel, token)) {
+          return true;
+        }
+
         const messages = this.#assistantMessages();
         if (await messages.count() === 0) {
           return false;
@@ -243,17 +254,24 @@ export class ChatGPTWebAdapter {
         const slug = await messages.last()
           .getAttribute("data-message-model-slug")
           .catch(() => null);
-        return Boolean(slug) && slugMatchesToken(slug, normalizeToken(requested));
+        return Boolean(slug) && slugMatchesToken(slug, token);
       },
       hasSwitcher: async () => Boolean(await this.#findModelSwitcher()),
-      openMenu: async () => {
+      openMenu: async (requested) => {
         const switcher = await this.#findModelSwitcher();
         if (switcher) {
-          await switcher.click().catch(() => null);
+          const opened = await switcher.click({ timeout: 5_000 })
+            .then(() => true)
+            .catch(() => false);
+          if (!opened) {
+            await switcher.click({ force: true, timeout: 5_000 })
+              .catch(() => null);
+          }
           // Wait for the Radix menu portal to attach before reading options.
-          await page.locator('[role="menu"] [role="menuitem"], [role="menu"] [role="menuitemradio"]')
+          await page.locator('[role="menu"]:visible [role="menuitem"]:visible, [role="menu"]:visible [role="menuitemradio"]:visible')
             .first().waitFor({ state: "visible", timeout: 5_000 })
             .catch(() => null);
+          await this.#revealNestedModeOption(requested);
         }
       },
       readOptions: async () => this.#readModeOptions(),
@@ -273,6 +291,32 @@ export class ChatGPTWebAdapter {
           .catch(() => false);
         return ok;
       },
+      waitSelected: async (requested) => {
+        const token = normalizeToken(requested);
+        const deadline = Date.now() + 5_000;
+        while (Date.now() < deadline) {
+          const switcher = await this.#findModelSwitcher(500);
+          const label = switcher
+            ? await switcher.innerText().catch(() => "")
+            : "";
+          if (labelMatchesToken(label, token)) {
+            return true;
+          }
+
+          const selectedOption = (await this.#readModeOptions()).find(
+            (option) => option.selected
+              && (
+                slugMatchesToken(option.slug, token)
+                || labelMatchesToken(option.label, token)
+              ),
+          );
+          if (selectedOption) {
+            return true;
+          }
+          await page.waitForTimeout(100);
+        }
+        return false;
+      },
       closeMenu: async () => {
         await page.keyboard.press("Escape").catch(() => null);
         await page.waitForTimeout(150);
@@ -283,8 +327,36 @@ export class ChatGPTWebAdapter {
 
   #modeOptionLocators() {
     return this.page.locator(
-      '[role="menu"] [role="menuitemradio"], [role="menu"] [role="menuitem"]',
+      '[role="menu"]:visible [role="menuitemradio"]:visible, '
+        + '[role="menu"]:visible [role="menuitem"]:visible',
+    );
+  }
+
+  async #revealNestedModeOption(requested) {
+    const present = (options) => {
+      const choice = chooseModeOption(options, requested);
+      return choice.status !== "unavailable";
+    };
+
+    let options = await this.#readModeOptions();
+    if (present(options)) {
+      return;
+    }
+
+    const submenuTriggers = this.page.locator(
+      '[role="menu"]:visible [role="menuitem"][aria-haspopup="menu"]:visible',
     );
+    const count = await submenuTriggers.count().catch(() => 0);
+    for (let index = 0; index < count; index += 1) {
+      const trigger = submenuTriggers.nth(index);
+      await trigger.dispatchEvent("pointermove", { pointerType: "mouse" })
+        .catch(() => null);
+      await this.page.waitForTimeout(250);
+      options = await this.#readModeOptions();
+      if (present(options)) {
+        return;
+      }
+    }
   }
 
   // Enumerates the open menu's options into { index, slug, label, disabled }.
@@ -296,13 +368,14 @@ export class ChatGPTWebAdapter {
     const options = [];
     for (let index = 0; index < count; index += 1) {
       const item = locator.nth(index);
-      const [testId, dataValue, dataTestValue, id, ariaDisabled, dataDisabled, dataState, label] =
+      const [testId, dataValue, dataTestValue, id, ariaDisabled, ariaChecked, dataDisabled, dataState, label] =
         await Promise.all([
           item.getAttribute("data-testid").catch(() => null),
           item.getAttribute("data-value").catch(() => null),
           item.getAttribute("data-test-value").catch(() => null),
           item.getAttribute("id").catch(() => null),
           item.getAttribute("aria-disabled").catch(() => null),
+          item.getAttribute("aria-checked").catch(() => null),
           item.getAttribute("data-disabled").catch(() => null),
           item.getAttribute("data-state").catch(() => null),
           item.innerText().catch(() => ""),
@@ -313,11 +386,13 @@ export class ChatGPTWebAdapter {
         || dataDisabled === ""
         || dataState === "disabled"
         || !await item.isEnabled().catch(() => true);
+      const selected = ariaChecked === "true" || dataState === "checked";
       options.push({
         index,
         slug,
         label: (label ?? "").trim(),
         disabled,
+        selected,
       });
     }
     return options;

package/src/browser/mode-selection.js

@@ -38,16 +38,17 @@ export function slugMatchesToken(slug, token) {
   return false;
 }
 
-function labelMatchesToken(label, token) {
+export function labelMatchesToken(label, token) {
   return String(label ?? "")
     .split(/\r?\n/)
     .some((line) => normalizeToken(line) === token);
 }
 
 // Given the enumerated menu options and the requested mode, decide what to do.
-// options: [{ index, slug, label, disabled }] in DOM order.
+// options: [{ index, slug, label, disabled, selected }] in DOM order.
 // Returns { status, targetIndex, selectedLabel, reason }.
 //   status "select"              -> requested mode is available; click it.
+//   status "already"             -> requested mode is already selected.
 //   status "fallback"            -> requested mode is limited/disabled; click
 //                                   the option immediately before it.
 //   status "unavailable"         -> requested mode is not present in the menu.
@@ -74,6 +75,14 @@ export function chooseModeOption(options, requested) {
   }
 
   const pro = options[proIndex];
+  if (pro.selected) {
+    return {
+      status: "already",
+      targetIndex: null,
+      selectedLabel: pro.label || requested,
+      reason: `Already using ${requested}.`,
+    };
+  }
   if (!pro.disabled) {
     return {
       status: "select",
@@ -112,9 +121,10 @@ export function chooseModeOption(options, requested) {
 // port: {
 //   alreadyOnMode(requested) -> bool,
 //   hasSwitcher()            -> bool,
-//   openMenu()               -> void,
-//   readOptions()            -> [{index, slug, label, disabled}],
+//   openMenu(requested)      -> void,
+//   readOptions()            -> [{index, slug, label, disabled, selected}],
 //   clickOption(index)       -> bool (click landed),
+//   waitSelected(requested)  -> bool (picker reflects the requested mode),
 //   waitClosed()             -> bool (menu closed after click),
 //   closeMenu()              -> void,
 //   writeDiagnostics(label)  -> void,
@@ -146,7 +156,7 @@ export async function runModeSelection(port, requested, { maxAttempts = 2 } = {}
 
   let lastReason = "";
   for (let attempt = 1; attempt <= maxAttempts; attempt += 1) {
-    await port.openMenu();
+    await port.openMenu(requested);
     const options = await port.readOptions();
 
     // The Radix menu populates asynchronously; an empty read is a race, retry.
@@ -158,13 +168,25 @@ export async function runModeSelection(port, requested, { maxAttempts = 2 } = {}
 
     const choice = chooseModeOption(options, requested);
 
+    if (choice.status === "already") {
+      await port.closeMenu();
+      return { ...choice, requested, attempts: attempt };
+    }
+
     if (choice.status === "select" || choice.status === "fallback") {
       const clicked = await port.clickOption(choice.targetIndex);
-      const closed = clicked && await port.waitClosed();
-      if (closed) {
+      const expectedMode = choice.status === "fallback"
+        ? choice.selectedLabel
+        : requested;
+      const selected = clicked && (
+        typeof port.waitSelected === "function"
+          ? await port.waitSelected(expectedMode)
+          : await port.waitClosed()
+      );
+      if (selected) {
         return { ...choice, requested, attempts: attempt };
       }
-      lastReason = "The option click did not register.";
+      lastReason = "The option click did not change the selected mode.";
       await port.closeMenu();
       continue;
     }

package/src/cli/main.js

@@ -28,10 +28,16 @@ import { extractAtMentions } from "./at-files.js";
 import {
   classifyChatInput,
   promptForText,
+  promptForSelect,
   readChatMessage,
   ShellChatInput,
 } from "./prompt-input.js";
 import { createRenderer } from "./render-events.js";
+import {
+  CHATGPT_MODE_CHOICES,
+  modeFromPromptChoice,
+  normalizeConfiguredMode,
+} from "./mode-choice.js";
 
 function resolveRuntimePaths(options) {
   const appDataDir = path.resolve(options.home ?? getAppDataDir());
@@ -401,13 +407,34 @@ async function runAgent(taskParts, options) {
   const interactive = !options.once && process.stdin.isTTY && process.stdout.isTTY;
   const chatInput = interactive ? new ShellChatInput() : null;
 
+  if (interactive) {
+    printChatBanner(projectRoot);
+  }
+
+  let requestedMode;
+  if (options.mode != null) {
+    requestedMode = normalizeConfiguredMode(options.mode);
+  } else if (interactive) {
+    const modeChoice = await promptForSelect({
+      message: "ChatGPT mode",
+      choices: CHATGPT_MODE_CHOICES,
+    });
+    if (modeChoice == null) {
+      chatInput?.close();
+      console.log("");
+      return null;
+    }
+    requestedMode = modeFromPromptChoice(modeChoice);
+  } else {
+    // Non-interactive callers cannot answer a picker. Preserve the current web
+    // setting unless they explicitly opt into `--mode Pro`.
+    requestedMode = null;
+  }
+
   // In interactive mode an initial task is optional: the user can just start
   // typing at the prompt. In one-shot mode a task is required.
   let task = taskParts.join(" ").trim();
   if (!task) {
-    if (interactive) {
-      printChatBanner(projectRoot, options.mode);
-    }
     const initialMessage = interactive
       ? await readChatMessage(() => chatInput.read())
       : await promptForText({
@@ -421,7 +448,6 @@ async function runAgent(taskParts, options) {
     }
     task = initialMessage.trim();
   } else if (interactive) {
-    printChatBanner(projectRoot, options.mode);
     chatInput.remember(task);
   }
 
@@ -431,7 +457,7 @@ async function runAgent(taskParts, options) {
     sessionsDir: paths.sessionsDir,
     task,
     projectRoot,
-    mode: options.mode,
+    mode: requestedMode,
   });
 
   // Resolve @file attachments in the opening task, if any. The task itself is
@@ -454,7 +480,7 @@ async function runAgent(taskParts, options) {
   return await executeSession({ session, options, files, chatInput });
 }
 
-function printChatBanner(projectRoot, mode) {
+function printChatBanner(projectRoot) {
   const CYAN = "\x1b[36m";
   const DIM = "\x1b[2m";
   const RESET = "\x1b[0m";
@@ -585,12 +611,15 @@ async function runExport(sessionId, options) {
 const program = new Command()
   .name("wtagent")
   .description("Turn your web AI session into a local tool-using agent.")
-  .version("0.1.0-alpha.3")
+  .version("0.1.0-alpha.4")
   .option("--home <path>", "Application data directory")
   .option("--profile-dir <path>", "Dedicated Chrome profile directory")
   .option("--chrome-path <path>", "Chrome/Chromium executable")
   .option("-C, --project <path>", "Project directory", process.cwd())
-  .option("--mode <name>", "ChatGPT mode to select", "Pro")
+  .option(
+    "--mode <name>",
+    'ChatGPT mode: "Pro" selects Pro; "Current" keeps the web setting',
+  )
   .option(
     "--once",
     "Run a single request and exit instead of a conversation",

package/src/cli/mode-choice.js

@@ -0,0 +1,25 @@
+export const CHATGPT_MODE_CHOICES = Object.freeze([
+  Object.freeze({
+    name: "Pro — select Pro for this conversation",
+    value: "pro",
+  }),
+  Object.freeze({
+    name: "Current — keep the current ChatGPT setting",
+    value: "current",
+  }),
+]);
+
+export function normalizeConfiguredMode(value) {
+  const configured = String(value ?? "").trim();
+  if (/^current$/i.test(configured)) {
+    return null;
+  }
+  if (/^pro$/i.test(configured)) {
+    return "Pro";
+  }
+  throw new Error('ChatGPT mode must be either "Pro" or "Current".');
+}
+
+export function modeFromPromptChoice(choice) {
+  return choice === "pro" ? "Pro" : null;
+}

package/src/cli/prompt-input.js

@@ -1,4 +1,4 @@
-import { input } from "@inquirer/prompts";
+import { input, select } from "@inquirer/prompts";
 import { createInterface } from "node:readline/promises";
 
 const CLEAN_EXIT_ERRORS = new Set([
@@ -112,11 +112,11 @@ export class ShellChatInput {
   }
 }
 
-export async function promptForText(config, {
-  prompt = input,
+async function promptWithCleanExit(config, {
+  prompt,
   inputStream = process.stdin,
   outputStream = process.stdout,
-} = {}) {
+}) {
   if (inputStream.readableEnded) {
     return null;
   }
@@ -150,3 +150,17 @@ export async function promptForText(config, {
     inputStream.removeListener("close", onEof);
   }
 }
+
+export async function promptForText(config, options = {}) {
+  return await promptWithCleanExit(config, {
+    ...options,
+    prompt: options.prompt ?? input,
+  });
+}
+
+export async function promptForSelect(config, options = {}) {
+  return await promptWithCleanExit(config, {
+    ...options,
+    prompt: options.prompt ?? select,
+  });
+}

package/src/runtime/agent-runtime.js

@@ -239,7 +239,7 @@ export class AgentRuntime {
     let activeMode = resume
       ? (this.session.state.activeMode ?? null)
       : null;
-    if (!resume) {
+    if (!resume && mode) {
       const modeResult = await this.adapter.selectMode(mode);
       if (modeResult) {
         await this.emit("conversation.mode_selected", {
@@ -339,7 +339,7 @@ export class AgentRuntime {
     let protocolErrors = 0;
     const baseTurn = resume ? Number(this.session.state.turn || 0) : 0;
 
-    for (let step = 1; step <= this.limits.maxTurns; step += 1) {
+    for (let step = 1; ; step += 1) {
       const turnNumber = baseTurn + step;
       await this.session.update({ turn: turnNumber, phase: "waiting_model" });
       const raw = await this.adapter.waitForTurnComplete({
@@ -699,12 +699,5 @@ export class AgentRuntime {
       });
     }
 
-    const message = `Maximum turn limit reached (${this.limits.maxTurns}).`;
-    await this.session.update({
-      phase: "interrupted",
-      lastError: message,
-    });
-    await this.emit("run.interrupted", { message });
-    throw new Error(message);
   }
 }

package/src/shared/limits.js

@@ -1,5 +1,4 @@
 export const DEFAULT_LIMITS = Object.freeze({
-  maxTurns: 36,
   maxProtocolErrors: 3,
   modelTurnTimeoutMs: 10 * 60_000,
   modelStableWindowMs: 1_500,