@nodetool-ai/runtime 0.8.3 → 0.8.4-a24.1

package/dist/providers/gemini-web-provider.js

@@ -4,19 +4,23 @@
  * No API key: requests run on the user's Google account and its Gemini plan.
  *
  * A turn is submitted through the page's own composer, so the site computes
- * its anti-abuse tokens itself. The page script records the response of the
- * page's own `StreamGenerate` request, and this module parses it. The tab
- * renders nothing while it is hidden, so the response is the only reliable
- * place to read the result.
+ * its anti-abuse tokens itself. The page streams the turn over a WebChannel
+ * of base64 protobuf frames, which this module does not parse. The composer's
+ * stop button shows while the turn runs. When it goes away, the provider reads
+ * the stored turn with the read call the page itself uses (`hNvQHb`). The tab
+ * renders no reply text while it is hidden, so the read is the only reliable
+ * place to get the result.
  *
  * The tab stays in the background. Chrome can freeze a background tab, and
  * a frozen Gemini runs a send click only when the tab is shown, which starts
  * a second turn later. The extension tab unfreezes the page before each
  * script. A click that still starts no request is dropped with a reload, and
- * the turn is retried once.
+ * the turn is retried once. Chrome also throttles the timers of a page hidden
+ * for a while to one wake-up a minute, so no page script waits on a timer: each
+ * one does a single step, and every wait polls from here ({@link pollTab}).
  *
  * A video finishes minutes after its turn ends. The provider polls the chat
- * with the read call the page itself uses (`hNvQHb`) until the video appears.
+ * with the same read until the video appears.
  *
  * Generated media need the user's Google cookies and send no CORS headers for
  * the page, so the browser itself downloads them ({@link ExtensionTab.download}).
@@ -33,19 +37,21 @@ import { PROVIDER_IDS } from "@nodetool-ai/protocol";
 import { fetchExternalMedia } from "../external-media-fetch.js";
 import { BaseProvider } from "./base-provider.js";
 import { composeDelta, composePrompt } from "./chatgpt-provider.js";
-import { createExtensionTab, stageBytesInTab } from "./extension-tab.js";
+import { createExtensionTab, pollTab, stageBytesInTab } from "./extension-tab.js";
 import { detectImageMime } from "./image-mime.js";
 const log = createLogger("nodetool.runtime.providers.gemini-web");
 const HOST = "gemini.google.com";
 const BASE_URL = `https://${HOST}`;
 /** The model id that keeps the mode the user picked last in Gemini. */
 export const GEMINI_WEB_AUTO_MODEL = "auto";
-/** The image model: Gemini's own image tool, run by the current mode. */
+/** The image model: Gemini's own image tool, run by the current mode. `gemini-web-image:<mode id>` picks a mode. */
 export const GEMINI_WEB_IMAGE_MODEL = "gemini-web-image";
 /** The video model: Gemini's video tool (Veo). */
 export const GEMINI_WEB_VIDEO_MODEL = "gemini-web-video";
 const COMPOSER = "rich-textarea .ql-editor";
 const SUBMIT = "gem-icon-button.send-button button";
+/** The send button turns into a stop button while a turn runs. */
+const STOP = "gem-icon-button.send-button.stop";
 const ATTACHMENT = "uploader-file-preview";
 const INPUT_AREA = "input-area-v2";
 const TOOLS_BUTTON = "gem-icon-button.menu-button button";
@@ -60,7 +66,21 @@ const VIDEO_POLL_INTERVAL_MS = 5_000;
 const READY_TIMEOUT_MS = 20_000;
 const APP_SETTLE_TIMEOUT_MS = 10_000;
 /** The send script waits in the page for uploads to finish before it clicks. */
-const SEND_TIMEOUT_MS = 90_000;
+/** How long the send button may stay disabled while attachments upload. */
+const SEND_READY_TIMEOUT_MS = 75_000;
+/** How long after the send click the page has to start its request. */
+const SEND_START_TIMEOUT_MS = 10_000;
+/** How long the chat read may lag behind the end of a turn. */
+const TURN_STORED_TIMEOUT_MS = 30_000;
+/** How long a background tab may take to render a composer menu button. */
+const MENU_TIMEOUT_MS = 10_000;
+/** How long an opened menu may take to show its items. */
+const MENU_ITEM_TIMEOUT_MS = 3_000;
+/** The first open of the mode picker after a page load can take a few seconds. */
+const MODE_LIST_TIMEOUT_MS = 6_000;
+/** A paste before the uploader is ready shows no preview. Paste again after this wait. */
+const PASTE_TIMEOUT_MS = 3_000;
+const PASTE_ATTEMPTS = 5;
 const CHAT_TIMEOUT_MS = 10 * 60_000;
 const IMAGE_TIMEOUT_MS = 10 * 60_000;
 const VIDEO_TIMEOUT_MS = 15 * 60_000;
@@ -71,20 +91,45 @@ let tab = null;
 /** Replace the tab. Tests use this to avoid a real browser. */
 export function setGeminiWebTab(next) {
     tab = next;
-    liveModels = null;
+    liveModes = null;
 }
 function currentTab() {
     return (tab ??= createExtensionTab(HOST));
 }
 /** Calls share one tab, so each waits for the previous one. */
+let activeSignal;
 let queue = Promise.resolve();
 function serialized(run) {
-    const next = queue.then(run, run);
+    const owned = async () => {
+        try {
+            return await run();
+        }
+        finally {
+            activeSignal = undefined;
+            await tab?.release?.();
+        }
+    };
+    const next = queue.then(owned, owned);
     queue = next.catch(() => undefined);
     return next;
 }
+/**
+ * The modes of the Gemini mode picker, listed before any call has read the
+ * live picker. An account can offer more, or fewer. Picking a mode the
+ * account lacks fails with the list of the modes it has.
+ */
+const KNOWN_MODES = [
+    { id: "8c46e95b1a07cecc", name: "3.5 Flash-Lite" },
+    { id: "56fdd199312815e2", name: "3.8 Flash" },
+    { id: "e6fa609c3fa255c0", name: "3.1 Pro" },
+    { id: "95a4c0abcba24364", name: "4 Argon" }
+];
 /** Modes read from the page during an earlier call, so listing never attaches. */
-let liveModels = null;
+let liveModes = null;
+/** The live modes when a call has read them, else {@link KNOWN_MODES}. */
+function listedModes() {
+    return liveModes?.modes ?? KNOWN_MODES;
+}
 function sleep(ms, signal) {
     return new Promise((resolve, reject) => {
         if (signal?.aborted)
@@ -95,7 +140,12 @@ function sleep(ms, signal) {
     });
 }
 async function evaluate(expression, timeoutMs) {
-    return (await currentTab().evaluate(expression, timeoutMs));
+    activeSignal?.throwIfAborted();
+    return (await currentTab().evaluate(expression, timeoutMs, activeSignal));
+}
+/** {@link pollTab} on the Gemini tab. */
+function poll(expression, timeoutMs) {
+    return pollTab(currentTab(), expression, timeoutMs, activeSignal);
 }
 /** The page path of a chat drops the `c_` prefix of its id. */
 function chatPath(conversationId) {
@@ -112,7 +162,7 @@ async function waitForComposer() {
     while (!ready && Date.now() < deadline) {
         ready = await evaluate(`/*nt:ready*/ !!document.querySelector(${JSON.stringify(COMPOSER)})`);
         if (!ready)
-            await sleep(250);
+            await sleep(250, activeSignal);
     }
     // `oPEP7c` holds the signed-in account. A signed-out page has a composer but no account.
     const signedIn = await evaluate(`/*nt:session*/ !!(window.WIZ_global_data && WIZ_global_data.oPEP7c)`);
@@ -124,7 +174,7 @@ async function waitForComposer() {
     // click before that starts no request. The mode picker is the last part.
     const settled = Date.now() + APP_SETTLE_TIMEOUT_MS;
     while (Date.now() < settled && !(await evaluate(`/*nt:settled*/ !!document.querySelector(${JSON.stringify(MODE_BUTTON)})`))) {
-        await sleep(250);
+        await sleep(250, activeSignal);
     }
 }
 /** Wait until the open chat shows its earlier turns. */
@@ -133,13 +183,13 @@ async function waitForHistory() {
     while (Date.now() < deadline) {
         if (await evaluate(`/*nt:history*/ document.querySelectorAll("user-query").length > 0`))
             return;
-        await sleep(250);
+        await sleep(250, activeSignal);
     }
     throw new ChatUnavailableError("The Gemini chat did not load its earlier turns");
 }
 /** Open a new chat with no tool selected. */
 async function openNewChat() {
-    await currentTab().navigate(`${BASE_URL}/app`);
+    await currentTab().navigate(`${BASE_URL}/app`, activeSignal);
     await waitForComposer();
     // A selected tool shows as a chip with a close icon.
     await evaluate(`/*nt:clear-tools*/ (async () => {
@@ -151,207 +201,217 @@ async function openNewChat() {
 }
 /** Turn on the tool whose menu item shows `icon`. */
 async function enableTool(icon) {
-    await evaluate(`/*nt:tool*/ (async () => {
-    const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
-    const icon = ${JSON.stringify(icon)};
-    let tools = null;
-    for (let i = 0; i < 40 && !tools; i++) {
-      tools = document.querySelector(${JSON.stringify(TOOLS_BUTTON)});
-      if (!tools) await sleep(250);
-    }
-    if (!tools) throw new Error("The Gemini composer has no tools menu");
-    tools.click();
-    let item = null;
-    for (let i = 0; i < 20 && !item; i++) {
-      await sleep(150);
-      item = [...document.querySelectorAll(${JSON.stringify(`${OVERLAY} [role=menuitemcheckbox]`)})].find((b) => b.querySelector('mat-icon[fonticon="' + icon + '"]'));
-    }
-    if (!item) throw new Error("The Gemini tools menu has no " + icon + " tool. The account may not offer it");
-    if (item.getAttribute("aria-checked") !== "true") item.click();
-    for (let i = 0; i < 20; i++) {
-      await sleep(150);
-      const chip = [...document.querySelectorAll(${JSON.stringify(`${INPUT_AREA} button`)})].find((b) => b.querySelector('mat-icon[fonticon="' + icon + '"]'));
-      if (chip) return true;
-    }
-    throw new Error("The Gemini " + icon + " tool did not turn on");
-  })()`);
+    const tools = await poll(`/*nt:tool-menu*/ !!document.querySelector(${JSON.stringify(TOOLS_BUTTON)})`, MENU_TIMEOUT_MS);
+    if (!tools)
+        throw new Error("The Gemini composer has no tools menu");
+    await evaluate(`/*nt:tool-open*/ (() => { document.querySelector(${JSON.stringify(TOOLS_BUTTON)}).click(); return true; })()`);
+    const item = await poll(`/*nt:tool-item*/ (() => {
+    const item = [...document.querySelectorAll(${JSON.stringify(`${OVERLAY} [role=menuitemcheckbox]`)})].find((b) => b.querySelector(${JSON.stringify(`mat-icon[fonticon="${icon}"]`)}));
+    return item ? (item.getAttribute("aria-checked") === "true" ? "on" : "off") : null;
+  })()`, MENU_ITEM_TIMEOUT_MS);
+    if (!item)
+        throw new Error(`The Gemini tools menu has no ${icon} tool. The account may not offer it`);
+    if (item === "off") {
+        await evaluate(`/*nt:tool-pick*/ (() => {
+      const item = [...document.querySelectorAll(${JSON.stringify(`${OVERLAY} [role=menuitemcheckbox]`)})].find((b) => b.querySelector(${JSON.stringify(`mat-icon[fonticon="${icon}"]`)}));
+      if (item) item.click();
+      return true;
+    })()`);
+    }
+    const chip = await poll(`/*nt:tool-chip*/ [...document.querySelectorAll(${JSON.stringify(`${INPUT_AREA} button`)})].some((b) => b.querySelector(${JSON.stringify(`mat-icon[fonticon="${icon}"]`)}))`, MENU_ITEM_TIMEOUT_MS);
+    if (!chip)
+        throw new Error(`The Gemini ${icon} tool did not turn on`);
 }
 /** Pick a video aspect ratio in the composer's aspect menu. */
 async function selectVideoAspect(aspectRatio) {
     const icon = aspectRatio === "9:16" ? "crop_9_16" : "crop_16_9";
-    await evaluate(`/*nt:aspect*/ (async () => {
-    const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
+    const state = await evaluate(`/*nt:aspect*/ (() => {
     const button = [...document.querySelectorAll(${JSON.stringify(`${INPUT_AREA} button`)})].find((b) => b.querySelector('mat-icon[fonticon^="crop_"]'));
     if (!button) throw new Error("The Gemini video composer has no aspect ratio menu");
-    if (button.querySelector('mat-icon[fonticon=${JSON.stringify(icon)}]')) return true;
+    if (button.querySelector(${JSON.stringify(`mat-icon[fonticon="${icon}"]`)})) return "selected";
     button.click();
-    let option = null;
-    for (let i = 0; i < 20 && !option; i++) {
-      await sleep(150);
-      option = [...document.querySelectorAll(${JSON.stringify(`${OVERLAY} [role=menuitemradio]`)})].find((b) => b.querySelector('mat-icon[fonticon=${JSON.stringify(icon)}]'));
-    }
-    if (!option) throw new Error("The Gemini aspect ratio menu has no ${aspectRatio} option");
+    return "opened";
+  })()`);
+    if (state === "selected")
+        return;
+    const picked = await poll(`/*nt:aspect-pick*/ (() => {
+    const option = [...document.querySelectorAll(${JSON.stringify(`${OVERLAY} [role=menuitemradio]`)})].find((b) => b.querySelector(${JSON.stringify(`mat-icon[fonticon="${icon}"]`)}));
+    if (!option) return false;
     option.click();
     return true;
-  })()`);
+  })()`, MENU_ITEM_TIMEOUT_MS);
+    if (!picked)
+        throw new Error(`The Gemini aspect ratio menu has no ${aspectRatio} option`);
 }
 /**
  * Open the mode picker, record its modes, and pick `modeId` unless it is
  * {@link GEMINI_WEB_AUTO_MODEL}. Modes carry a stable `data-mode-id`.
  */
 async function selectMode(modeId) {
-    const modes = await evaluate(`/*nt:modes*/ (async () => {
-    const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
     // A background tab renders the mode picker seconds after the composer.
-    let toggle = null;
-    for (let i = 0; i < 40 && !toggle; i++) {
-      toggle = document.querySelector(${JSON.stringify(MODE_BUTTON)});
-      if (!toggle) await sleep(250);
-    }
-    if (!toggle) return [];
-    // The first open after a page load can take a few seconds.
-    if (toggle.getAttribute("aria-expanded") !== "true") toggle.click();
-    let items = [];
-    for (let i = 0; i < 40 && items.length === 0; i++) {
-      await sleep(150);
-      items = [...document.querySelectorAll("[data-mode-id]")];
+    const toggle = await poll(`/*nt:mode-menu*/ !!document.querySelector(${JSON.stringify(MODE_BUTTON)})`, MENU_TIMEOUT_MS);
+    let modes = [];
+    if (toggle) {
+        await evaluate(`/*nt:mode-open*/ (() => {
+      const toggle = document.querySelector(${JSON.stringify(MODE_BUTTON)});
+      if (toggle.getAttribute("aria-expanded") !== "true") toggle.click();
+      return true;
+    })()`);
+        // The first open after a page load can take a few seconds.
+        modes = (await poll(`/*nt:modes*/ (() => {
+      const items = [...document.querySelectorAll("[data-mode-id]")];
+      if (items.length === 0) return null;
+      const modes = items.map((item) => ({ id: item.getAttribute("data-mode-id"), name: item.innerText.split("\\n")[0].trim() }));
+      const wanted = ${JSON.stringify(modeId)};
+      const pick = items.find((item) => item.getAttribute("data-mode-id") === wanted);
+      const toggle = document.querySelector(${JSON.stringify(MODE_BUTTON)});
+      if (pick) pick.click();
+      else if (toggle && toggle.getAttribute("aria-expanded") === "true") toggle.click();
+      return modes;
+    })()`, MODE_LIST_TIMEOUT_MS)) ?? [];
+        await sleep(300);
     }
-    const modes = items.map((item) => ({ id: item.getAttribute("data-mode-id"), name: item.innerText.split("\\n")[0].trim() }));
-    const wanted = ${JSON.stringify(modeId)};
-    const pick = items.find((item) => item.getAttribute("data-mode-id") === wanted);
-    if (pick) pick.click();
-    else if (toggle.getAttribute("aria-expanded") === "true") toggle.click();
-    await sleep(300);
-    return modes;
-  })()`);
     if (modes.length > 0) {
-        liveModels = {
-            at: Date.now(),
-            models: modes.filter((m) => m.id).map((m) => ({ id: m.id, name: `Gemini ${m.name || m.id}`, provider: PROVIDER_IDS.GEMINI_WEB }))
-        };
+        liveModes = { at: Date.now(), modes: modes.filter((m) => m.id).map((m) => ({ id: m.id, name: m.name || m.id })) };
     }
     if (modeId !== GEMINI_WEB_AUTO_MODEL && !modes.some((m) => m.id === modeId)) {
         throw new Error(`Gemini has no mode ${modeId}. Its modes are: ${modes.map((m) => `${m.id} (${m.name})`).join(", ") || "none found"}`);
     }
 }
-/** Paste images into the composer, the way a clipboard paste would. */
+/**
+ * Paste images into the composer, the way a clipboard paste would. The page
+ * drops a paste that arrives before its uploader is ready, and the mode
+ * button can appear before that. So paste again until the previews show.
+ */
 async function attachImages(images) {
     if (images.length === 0)
         return;
     const files = [];
     for (const [index, image] of images.entries()) {
         const type = detectImageMime(image);
-        const id = await stageBytesInTab(currentTab(), image);
+        const id = await stageBytesInTab(currentTab(), image, activeSignal);
         files.push({ id, name: `image-${index + 1}.${type.split("/")[1] ?? "png"}`, type });
     }
-    await evaluate(`/*nt:attach*/ (() => {
+    const paste = `/*nt:attach*/ (() => {
     const files = ${JSON.stringify(files)};
     const composer = document.querySelector(${JSON.stringify(COMPOSER)});
     const transfer = new DataTransfer();
     for (const f of files) {
       const binary = atob(window.__ntStage[f.id]);
-      delete window.__ntStage[f.id];
       transfer.items.add(new File([Uint8Array.from(binary, (c) => c.charCodeAt(0))], f.name, { type: f.type }));
     }
     composer.focus();
     composer.dispatchEvent(new ClipboardEvent("paste", { clipboardData: transfer, bubbles: true, cancelable: true }));
     return true;
-  })()`);
+  })()`;
+    let attached = false;
+    for (let attempt = 1; !attached && attempt <= PASTE_ATTEMPTS; attempt++) {
+        await evaluate(paste);
+        attached = !!(await poll(`/*nt:attached*/ document.querySelectorAll(${JSON.stringify(ATTACHMENT)}).length >= ${files.length}`, PASTE_TIMEOUT_MS));
+    }
+    await evaluate(`/*nt:unstage*/ (() => { for (const id of ${JSON.stringify(files.map((f) => f.id))}) delete window.__ntStage[id]; return true; })()`);
+    if (!attached)
+        throw new Error("Gemini did not take the pasted images: no attachment preview appeared");
+}
+/** The send click started no request. The caller reloads the page and sends again. */
+class NotSentError extends Error {
 }
 /**
- * Record the page's `StreamGenerate` responses in `window.__ntGemini`, type
- * the text and click send once. The script waits for the attachment previews
- * and an enabled button first, then checks that the page started a request.
+ * Type the text and click send once. The click waits for the attachment
+ * previews and an enabled button. Then the page must show the new turn: a
+ * new user query or the stop button.
  */
 async function submit(text, attachments) {
-    await evaluate(`/*nt:send*/ (async () => {
-    const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
-    if (!window.__ntGeminiHooked) {
-      window.__ntGeminiHooked = true;
-      const open = XMLHttpRequest.prototype.open;
-      const send = XMLHttpRequest.prototype.send;
-      XMLHttpRequest.prototype.open = function (method, url) {
-        this.__ntUrl = String(url);
-        return open.apply(this, arguments);
-      };
-      XMLHttpRequest.prototype.send = function () {
-        const state = window.__ntGemini;
-        if (state && /StreamGenerate/.test(this.__ntUrl || "")) {
-          state.started = true;
-          this.addEventListener("loadend", () => {
-            state.response = { status: this.status, body: String(this.responseText || "") };
-          });
-        }
-        return send.apply(this, arguments);
-      };
-    }
-    window.__ntGemini = { started: false, response: null };
+    await evaluate(`/*nt:send*/ (() => {
+    window.__ntGemini = { queries: document.querySelectorAll("user-query").length };
     const composer = document.querySelector(${JSON.stringify(COMPOSER)});
     composer.focus();
     document.execCommand("insertText", false, ${JSON.stringify(text)});
+    return true;
+  })()`);
     // The send button slides in after the first character. A click during
     // that animation starts no request.
-    await sleep(500);
-    const deadline = Date.now() + ${SEND_TIMEOUT_MS - 15_000};
-    for (;;) {
-      if (Date.now() > deadline) throw new Error("Gemini did not accept the message: the send button stayed disabled or the upload did not finish");
-      const button = document.querySelector(${JSON.stringify(SUBMIT)});
-      const previews = document.querySelectorAll(${JSON.stringify(ATTACHMENT)});
-      const uploading = [...previews].some((p) => p.querySelector("mat-progress-spinner, [role=progressbar]"));
-      if (button && !button.disabled && button.getAttribute("aria-disabled") !== "true" && previews.length >= ${attachments} && !uploading) {
-        button.click();
-        break;
-      }
-      await sleep(250);
-    }
-    const sendDeadline = Date.now() + 10000;
-    while (Date.now() < sendDeadline) {
-      if (window.__ntGemini.started) return true;
-      await sleep(250);
-    }
-    throw new Error("NT_NOT_SENT");
-  })()`, SEND_TIMEOUT_MS);
+    await sleep(500, activeSignal);
+    const ready = await poll(`/*nt:send-ready*/ (() => {
+    const button = document.querySelector(${JSON.stringify(SUBMIT)});
+    const previews = document.querySelectorAll(${JSON.stringify(ATTACHMENT)});
+    const uploading = [...previews].some((p) => p.querySelector("mat-progress-spinner, [role=progressbar]"));
+    return !!button && !button.disabled && button.getAttribute("aria-disabled") !== "true" && previews.length >= ${attachments} && !uploading;
+  })()`, SEND_READY_TIMEOUT_MS);
+    if (!ready)
+        throw new Error("Gemini did not accept the message: the send button stayed disabled or the upload did not finish");
+    await evaluate(`/*nt:click*/ (() => { document.querySelector(${JSON.stringify(SUBMIT)}).click(); return true; })()`);
+    const started = await poll(`/*nt:sent*/ !!window.__ntGemini && (document.querySelectorAll("user-query").length > window.__ntGemini.queries || !!document.querySelector(${JSON.stringify(STOP)}))`, SEND_START_TIMEOUT_MS);
+    if (!started)
+        throw new NotSentError("Gemini did not start a request after the send click");
 }
-/** Poll until the page's `StreamGenerate` request ends and return its response. */
-async function waitForResponse(timeoutMs, signal) {
+/** The chat id in the page path, or null on a new chat that has none yet. */
+async function currentChatId() {
+    const pathname = await evaluate(`/*nt:location*/ location.pathname`);
+    const id = /^\/app\/([0-9a-f]+)$/.exec(pathname)?.[1];
+    return id ? `c_${id}` : null;
+}
+/**
+ * Wait until the turn after `priorTurnId` ends, then read it from the chat.
+ * The stop button shows while the turn runs. The read finds the turn only
+ * once Gemini has stored it, a moment after the button goes away.
+ */
+async function waitForReply(priorTurnId, timeoutMs, signal) {
     const started = Date.now();
+    let endedAt = null;
     let failures = 0;
     for (;;) {
         await sleep(POLL_INTERVAL_MS, signal);
         try {
-            const state = await evaluate(`/*nt:result*/ (window.__ntGemini ?? null)`);
+            const running = await evaluate(`/*nt:running*/ !!document.querySelector(${JSON.stringify(STOP)})`);
+            const conversationId = running ? null : await currentChatId();
+            if (conversationId) {
+                const reply = parseChatTurn(await readChat(conversationId));
+                if (reply.turnId && reply.turnId !== priorTurnId)
+                    return reply;
+            }
             failures = 0;
-            if (!state)
-                throw new Error("The Gemini tab navigated away during the turn");
-            if (state.response)
-                return state.response;
+            if (running)
+                endedAt = null;
+            else
+                endedAt ??= Date.now();
         }
         catch (error) {
-            if (error instanceof Error && error.message.startsWith("The Gemini tab navigated"))
-                throw error;
             log.warn("Gemini poll failed", { failures: failures + 1, error: error instanceof Error ? error.message : String(error) });
             if (++failures > MAX_POLL_FAILURES)
                 throw error;
         }
+        if (endedAt !== null && Date.now() - endedAt > TURN_STORED_TIMEOUT_MS) {
+            const conversationId = await currentChatId().catch(() => null);
+            throw new Error(`Gemini ended the turn, but its chat shows no reply.${chatLink(conversationId)}`);
+        }
         if (Date.now() - started > timeoutMs)
             throw new Error(`Gemini did not finish within ${timeoutMs / 1000}s`);
     }
 }
 /**
- * Prepare the page with `setup`, paste the images, send the text, and parse
+ * Prepare the page with `setup`, paste the images, send the text, and read
  * the reply. When the click starts no request, `setup` runs again, which
  * reloads the page and drops the pending click, and the send is retried once.
  */
 async function sendAndRead(setup, text, images, timeoutMs, signal) {
+    activeSignal = signal;
+    let priorTurnId;
     for (let attempt = 1;; attempt++) {
+        activeSignal?.throwIfAborted();
         await setup();
+        activeSignal?.throwIfAborted();
+        // A continued chat already has turns. The reply is the next one.
+        const conversationId = await currentChatId();
+        priorTurnId = conversationId ? parseChatTurn(await readChat(conversationId)).turnId : null;
         await attachImages(images);
         try {
+            activeSignal?.throwIfAborted();
             await submit(text, images.length);
             break;
         }
         catch (error) {
-            if (!(error instanceof Error && error.message.includes("NT_NOT_SENT")))
+            if (!(error instanceof NotSentError))
                 throw error;
             if (attempt > 1)
                 throw new Error("Gemini did not send the message from the background tab. Show the Gemini tab once and try again");
@@ -359,10 +419,7 @@ async function sendAndRead(setup, text, images, timeoutMs, signal) {
         }
     }
     log.info("Gemini turn submitted", { images: images.length });
-    const response = await waitForResponse(timeoutMs, signal);
-    if (response.status !== 200)
-        throw new Error(`Gemini request failed: HTTP ${response.status}`);
-    return parseStreamGenerate(response.body);
+    return waitForReply(priorTurnId, timeoutMs, signal);
 }
 /** Read a chat the way the page does, with the `hNvQHb` call. Returns the raw response body. */
 async function readChat(conversationId) {
@@ -388,7 +445,7 @@ async function readChat(conversationId) {
     return await response.text();
   })()`);
 }
-/** Each `wrb.fr` entry's payload in a batchexecute-style body, one JSON frame per line. */
+/** Each `wrb.fr` entry's payload in a batchexecute body, one JSON frame per line. */
 function payloads(body) {
     const found = [];
     for (const line of body.split("\n")) {
@@ -423,6 +480,7 @@ function contains(value, match) {
         return Object.values(value).some((item) => contains(item, match));
     return false;
 }
+const VIDEO_PLACEHOLDER = /https?:\/\/googleusercontent\.com\/generated_video_content\//;
 /** A candidate's reply text, without the placeholders Gemini puts where media go. */
 function candidateText(candidate) {
     const parts = Array.isArray(candidate[1]) ? candidate[1] : [];
@@ -433,32 +491,41 @@ function candidateText(candidate) {
         .trim();
 }
 /**
- * Parse a `StreamGenerate` response body. A payload holds the ids at `[1]`
- * and the reply candidates at `[4]`. Later frames repeat the candidate with
- * more content, so the last one wins. Generated images sit under the
- * candidate's `[12]` as `gg-dl` URLs. A started video tool shows as a
- * `video_generation` entry in a status payload.
+ * The newest turn in an `hNvQHb` chat read and its first candidate. The
+ * payload lists the turns at `[0]`, newest first. A turn holds its ids at
+ * `[0]` and its candidates at `[3][0]`. A read of a chat that Gemini has not
+ * stored yet carries an error entry and no payload.
  */
-export function parseStreamGenerate(body) {
-    let conversationId = null;
-    let candidate = null;
-    let videoPending = false;
-    for (const payload of payloads(body)) {
-        const ids = payload[1];
-        if (Array.isArray(ids) && typeof ids[0] === "string")
-            conversationId = ids[0];
-        const candidates = payload[4];
-        if (Array.isArray(candidates) && Array.isArray(candidates[0]))
-            candidate = candidates[0];
-        if (contains(payload, (text) => text === "video_generation"))
-            videoPending = true;
-    }
+function newestTurn(body) {
+    const payload = payloads(body)[0];
+    const turn = Array.isArray(payload?.[0]) ? payload[0][0] : null;
+    if (!Array.isArray(turn))
+        return null;
+    const candidates = Array.isArray(turn[3]) ? turn[3][0] : null;
+    const candidate = Array.isArray(candidates) && Array.isArray(candidates[0]) ? candidates[0] : null;
+    return { turn, candidate };
+}
+/**
+ * Parse the newest turn of an `hNvQHb` chat read. Generated images sit under
+ * the candidate's `[12]` as `lh3.googleusercontent.com/gg/` URLs. A started
+ * video shows as a video placeholder in the text or a `video_generation` entry.
+ */
+export function parseChatTurn(body) {
+    const found = newestTurn(body);
+    if (!found)
+        return { conversationId: null, turnId: null, text: "", images: [], videoPending: false };
+    const ids = Array.isArray(found.turn[0]) ? found.turn[0] : [];
+    const conversationId = typeof ids[0] === "string" ? ids[0] : null;
+    const turnId = typeof ids[1] === "string" ? ids[1] : null;
+    const candidate = found.candidate;
+    const videoPending = contains(found.turn, (text) => text === "video_generation") ||
+        (candidate !== null && contains(candidate[1], (text) => VIDEO_PLACEHOLDER.test(text)));
     if (!candidate)
-        return { conversationId, text: "", images: [], videoPending };
+        return { conversationId, turnId, text: "", images: [], videoPending };
     const images = new Set();
     const walk = (value) => {
         if (typeof value === "string") {
-            if (/^https:\/\/lh3\.googleusercontent\.com\/gg-dl\//.test(value))
+            if (/^https:\/\/lh3\.googleusercontent\.com\/gg(-dl)?\//.test(value))
                 images.add(value);
         }
         else if (Array.isArray(value)) {
@@ -466,23 +533,19 @@ export function parseStreamGenerate(body) {
         }
     };
     walk(candidate[12]);
-    return { conversationId, text: candidateText(candidate), images: [...images], videoPending };
+    return { conversationId, turnId, text: candidateText(candidate), images: [...images], videoPending };
 }
 /**
- * Find the generated video in an `hNvQHb` chat read. The payload lists the
- * chat's turns at `[0]`, newest first, and a turn's candidates at `[3][0]`.
- * A finished video is an array that holds `"video/mp4"` and, at `[7]`, its
+ * Find the generated video in the newest turn of an `hNvQHb` chat read. A
+ * finished video is an array that holds `"video/mp4"` and, at `[7]`, its
  * URLs: thumbnails on lh3 and the file itself on
  * `contribution.usercontent.google.com`.
  */
 export function findGeneratedVideo(body) {
-    const payload = payloads(body)[0];
-    const turn = Array.isArray(payload?.[0]) ? payload[0][0] : null;
-    if (!Array.isArray(turn))
+    const found = newestTurn(body);
+    if (!found)
         return { url: null, text: "" };
-    const candidates = Array.isArray(turn[3]) ? turn[3][0] : null;
-    const candidate = Array.isArray(candidates) && Array.isArray(candidates[0]) ? candidates[0] : null;
-    const text = candidate ? candidateText(candidate) : "";
+    const text = found.candidate ? candidateText(found.candidate) : "";
     let entry = null;
     const walk = (value) => {
         if (entry || !Array.isArray(value))
@@ -492,7 +555,7 @@ export function findGeneratedVideo(body) {
         else
             value.forEach(walk);
     };
-    walk(turn);
+    walk(found.turn);
     const urls = entry && Array.isArray(entry[7]) ? entry[7].filter((u) => typeof u === "string") : [];
     const url = urls.find((u) => u.startsWith("https://contribution.usercontent.google.com/download")) ?? null;
     return { url, text };
@@ -522,7 +585,7 @@ async function runChatTurnHoldingTab(turn) {
     const setup = async () => {
         if (turn.conversationId) {
             const path = chatPath(turn.conversationId);
-            await currentTab().navigate(`${BASE_URL}${path}`);
+            await currentTab().navigate(`${BASE_URL}${path}`, activeSignal);
             await waitForComposer();
             // A deleted or foreign chat redirects to a new one.
             const pathname = await evaluate(`/*nt:location*/ location.pathname`);
@@ -534,7 +597,7 @@ async function runChatTurnHoldingTab(turn) {
         else {
             await openNewChat();
         }
-        if (turn.model !== GEMINI_WEB_AUTO_MODEL || !liveModels || Date.now() - liveModels.at > MODELS_TTL_MS) {
+        if (turn.model !== GEMINI_WEB_AUTO_MODEL || !liveModes || Date.now() - liveModes.at > MODELS_TTL_MS) {
             await selectMode(turn.model);
         }
     };
@@ -556,21 +619,26 @@ export class GeminiWebProvider extends BaseProvider {
     }
     /**
      * Listing never attaches to the tab: it returns `auto` plus the modes read
-     * during an earlier call. A mode id from the Gemini mode picker also works
-     * as a model id.
+     * during an earlier call, or the known public modes before any call. A mode
+     * id from the Gemini mode picker also works as a model id.
      */
     async getAvailableLanguageModels() {
         const auto = { id: GEMINI_WEB_AUTO_MODEL, name: "Gemini (current mode)", provider: PROVIDER_IDS.GEMINI_WEB };
-        return [auto, ...(liveModels?.models ?? [])];
+        return [auto, ...listedModes().map((m) => ({ id: m.id, name: `Gemini ${m.name}`, provider: PROVIDER_IDS.GEMINI_WEB }))];
     }
+    /** One image model for the current mode, and one for each listed mode. */
     async getAvailableImageModels() {
-        return [{
-                id: GEMINI_WEB_IMAGE_MODEL,
-                name: "Gemini Images (web)",
-                provider: PROVIDER_IDS.GEMINI_WEB,
-                supportedTasks: ["text_to_image", "image_to_image"],
-                aspectRatios: ["1:1", "16:9", "9:16", "4:3", "3:4"]
-            }];
+        const image = (id, name) => ({
+            id,
+            name,
+            provider: PROVIDER_IDS.GEMINI_WEB,
+            supportedTasks: ["text_to_image", "image_to_image"],
+            aspectRatios: ["1:1", "16:9", "9:16", "4:3", "3:4"]
+        });
+        return [
+            image(GEMINI_WEB_IMAGE_MODEL, "Gemini Images (web)"),
+            ...listedModes().map((m) => image(`${GEMINI_WEB_IMAGE_MODEL}:${m.id}`, `Gemini Images (${m.name})`))
+        ];
     }
     async getAvailableVideoModels() {
         return [{
@@ -626,7 +694,7 @@ export class GeminiWebProvider extends BaseProvider {
         }
         if (!result) {
             // The caller may have sent only the delta; a new chat needs the whole conversation.
-            const messages = canResume && args.loadFullHistory ? await args.loadFullHistory() : args.messages;
+            const messages = args.loadFullHistory ? await args.loadFullHistory() : args.messages;
             const images = await this.lastUserImages(messages, args.signal);
             result = await serialized(() => runChatTurnHoldingTab({ model: args.model, text: composePrompt(messages), images, signal: args.signal }));
         }
@@ -665,6 +733,13 @@ export class GeminiWebProvider extends BaseProvider {
         return this.generateVideo(params.prompt ?? "", image, params);
     }
     async generateImage(prompt, images, params) {
+        // `gemini-web-image:<mode id>` runs the image tool in that mode.
+        const modeId = params.model?.id.startsWith(`${GEMINI_WEB_IMAGE_MODEL}:`) ? params.model.id.slice(GEMINI_WEB_IMAGE_MODEL.length + 1) : null;
+        const setup = async () => {
+            await openNewChat();
+            if (modeId)
+                await selectMode(modeId);
+        };
         const text = [
             images.length > 0
                 ? "Edit the attached image(s) as described and reply with one image. Do not ask questions."
@@ -676,7 +751,7 @@ export class GeminiWebProvider extends BaseProvider {
         // The download runs in the same turn on the tab, so no other call can
         // navigate the tab away in between.
         return serialized(async () => {
-            const reply = await sendAndRead(openNewChat, text, images, IMAGE_TIMEOUT_MS, params.signal);
+            const reply = await sendAndRead(setup, text, images, IMAGE_TIMEOUT_MS, params.signal);
             const url = reply.images.at(-1);
             if (!url) {
                 throw new Error(`Gemini finished without an image${reply.text ? `: ${reply.text.slice(0, 300)}` : ""}.${chatLink(reply.conversationId)}`);