@steipete/oracle 0.9.0 → 0.10.0

package/dist/src/browser/actions/promptComposer.js

@@ -1,15 +1,15 @@
-import { INPUT_SELECTORS, PROMPT_PRIMARY_SELECTOR, PROMPT_FALLBACK_SELECTOR, SEND_BUTTON_SELECTORS, CONVERSATION_TURN_SELECTOR, STOP_BUTTON_SELECTOR, ASSISTANT_ROLE_SELECTOR, } from '../constants.js';
-import { delay } from '../utils.js';
-import { logDomFailure } from '../domDebug.js';
-import { buildClickDispatcher } from './domEvents.js';
-import { BrowserAutomationError } from '../../oracle/errors.js';
+import { INPUT_SELECTORS, PROMPT_PRIMARY_SELECTOR, PROMPT_FALLBACK_SELECTOR, SEND_BUTTON_SELECTORS, CONVERSATION_TURN_SELECTOR, STOP_BUTTON_SELECTOR, ASSISTANT_ROLE_SELECTOR, } from "../constants.js";
+import { delay } from "../utils.js";
+import { logDomFailure } from "../domDebug.js";
+import { buildClickDispatcher } from "./domEvents.js";
+import { BrowserAutomationError } from "../../oracle/errors.js";
 const ENTER_KEY_EVENT = {
-    key: 'Enter',
-    code: 'Enter',
+    key: "Enter",
+    code: "Enter",
     windowsVirtualKeyCode: 13,
     nativeVirtualKeyCode: 13,
 };
-const ENTER_KEY_TEXT = '\r';
+const ENTER_KEY_TEXT = "\r";
 export async function submitPrompt(deps, prompt, logger) {
     const { runtime, input } = deps;
     await waitForDomReady(runtime, logger, deps.inputTimeoutMs ?? undefined);
@@ -63,8 +63,8 @@ export async function submitPrompt(deps, prompt, logger) {
         awaitPromise: true,
     });
     if (!focusResult.result?.value?.focused) {
-        await logDomFailure(runtime, logger, 'focus-textarea');
-        throw new Error('Failed to focus prompt textarea');
+        await logDomFailure(runtime, logger, "focus-textarea");
+        throw new Error("Failed to focus prompt textarea");
     }
     await input.insertText({ text: prompt });
     // Some pages (notably ChatGPT when subscriptions/widgets load) need a brief settle
@@ -99,12 +99,12 @@ export async function submitPrompt(deps, prompt, logger) {
     })()`,
         returnByValue: true,
     });
-    const editorTextRaw = verification.result?.value?.editorText ?? '';
-    const fallbackValueRaw = verification.result?.value?.fallbackValue ?? '';
-    const activeValueRaw = verification.result?.value?.activeValue ?? '';
-    const editorTextTrimmed = editorTextRaw?.trim?.() ?? '';
-    const fallbackValueTrimmed = fallbackValueRaw?.trim?.() ?? '';
-    const activeValueTrimmed = activeValueRaw?.trim?.() ?? '';
+    const editorTextRaw = verification.result?.value?.editorText ?? "";
+    const fallbackValueRaw = verification.result?.value?.fallbackValue ?? "";
+    const activeValueRaw = verification.result?.value?.activeValue ?? "";
+    const editorTextTrimmed = editorTextRaw?.trim?.() ?? "";
+    const fallbackValueTrimmed = fallbackValueRaw?.trim?.() ?? "";
+    const activeValueTrimmed = activeValueRaw?.trim?.() ?? "";
     if (!editorTextTrimmed && !fallbackValueTrimmed && !activeValueTrimmed) {
         // Learned: occasionally Input.insertText doesn't land in the editor; force textContent/value + input events.
         await runtime.evaluate({
@@ -152,16 +152,16 @@ export async function submitPrompt(deps, prompt, logger) {
     })()`,
         returnByValue: true,
     });
-    const observedEditor = postVerification.result?.value?.editorText ?? '';
-    const observedFallback = postVerification.result?.value?.fallbackValue ?? '';
-    const observedActive = postVerification.result?.value?.activeValue ?? '';
+    const observedEditor = postVerification.result?.value?.editorText ?? "";
+    const observedFallback = postVerification.result?.value?.fallbackValue ?? "";
+    const observedActive = postVerification.result?.value?.activeValue ?? "";
     const observedLength = Math.max(observedEditor.length, observedFallback.length, observedActive.length);
     if (promptLength >= 50_000 && observedLength > 0 && observedLength < promptLength - 2_000) {
         // Learned: very large prompts can truncate silently; fail fast so we can fall back to file uploads.
-        await logDomFailure(runtime, logger, 'prompt-too-large');
-        throw new BrowserAutomationError('Prompt appears truncated in the composer (likely too large).', {
-            stage: 'submit-prompt',
-            code: 'prompt-too-large',
+        await logDomFailure(runtime, logger, "prompt-too-large");
+        throw new BrowserAutomationError("Prompt appears truncated in the composer (likely too large).", {
+            stage: "submit-prompt",
+            code: "prompt-too-large",
             promptLength,
             observedLength,
         });
@@ -169,19 +169,19 @@ export async function submitPrompt(deps, prompt, logger) {
     const clicked = await attemptSendButton(runtime, logger, deps?.attachmentNames);
     if (!clicked) {
         await input.dispatchKeyEvent({
-            type: 'keyDown',
+            type: "keyDown",
             ...ENTER_KEY_EVENT,
             text: ENTER_KEY_TEXT,
             unmodifiedText: ENTER_KEY_TEXT,
         });
         await input.dispatchKeyEvent({
-            type: 'keyUp',
+            type: "keyUp",
             ...ENTER_KEY_EVENT,
         });
-        logger('Submitted prompt via Enter key');
+        logger("Submitted prompt via Enter key");
     }
     else {
-        logger('Clicked send button');
+        logger("Clicked send button");
     }
     const commitTimeoutMs = Math.max(60_000, deps.inputTimeoutMs ?? 0);
     // Learned: the send button can succeed but the turn doesn't appear immediately; verify commit via turns/stop button.
@@ -231,8 +231,8 @@ export async function clearPromptComposer(Runtime, logger) {
         returnByValue: true,
     });
     if (!result.result?.value?.cleared) {
-        await logDomFailure(Runtime, logger, 'clear-composer');
-        throw new Error('Failed to clear prompt composer');
+        await logDomFailure(Runtime, logger, "clear-composer");
+        throw new Error("Failed to clear prompt composer");
     }
     await delay(250);
 }
@@ -265,24 +265,42 @@ function buildAttachmentReadyExpression(attachmentNames) {
       document.querySelector('form') ||
       document.body ||
       document;
-    const match = (node, name) => (node?.textContent || '').toLowerCase().includes(name);
+    const labelText = (node) =>
+      [
+        node?.textContent,
+        node?.getAttribute?.('aria-label'),
+        node?.getAttribute?.('title'),
+        node?.getAttribute?.('data-testid'),
+      ]
+        .filter(Boolean)
+        .join(' ')
+        .toLowerCase();
+    const match = (node, name) => labelText(node).includes(name);
 
     // Restrict to attachment affordances; never scan generic div/span nodes (prompt text can contain the file name).
     const attachmentSelectors = [
       '[data-testid*="chip"]',
       '[data-testid*="attachment"]',
       '[data-testid*="upload"]',
-      '[aria-label="Remove file"]',
-      'button[aria-label="Remove file"]',
+      '[data-testid*="file"]',
+      '[aria-label*="Remove file"]',
+      'button[aria-label*="Remove file"]',
+      '[aria-label*="remove file"]',
+      'button[aria-label*="remove file"]',
     ];
+    const attachmentRoots = Array.from(new Set([composer, document])).filter(Boolean);
 
     const chipsReady = names.every((name) =>
-      Array.from(composer.querySelectorAll(attachmentSelectors.join(','))).some((node) => match(node, name)),
+      attachmentRoots.some((root) =>
+        Array.from(root.querySelectorAll(attachmentSelectors.join(','))).some((node) => match(node, name)),
+      ),
     );
     const inputsReady = names.every((name) =>
-      Array.from(composer.querySelectorAll('input[type="file"]')).some((el) =>
-        Array.from((el instanceof HTMLInputElement ? el.files : []) || []).some((file) =>
-          file?.name?.toLowerCase?.().includes(name),
+      attachmentRoots.some((root) =>
+        Array.from(root.querySelectorAll('input[type="file"]')).some((el) =>
+          Array.from((el instanceof HTMLInputElement ? el.files : []) || []).some((file) =>
+            file?.name?.toLowerCase?.().includes(name),
+          ),
         ),
       ),
     );
@@ -297,28 +315,36 @@ async function attemptSendButton(Runtime, _logger, attachmentNames) {
     const script = `(() => {
     ${buildClickDispatcher()}
     const selectors = ${JSON.stringify(SEND_BUTTON_SELECTORS)};
-    let button = null;
+    const isVisible = (node) => {
+      if (!(node instanceof HTMLElement)) return false;
+      const rect = node.getBoundingClientRect();
+      if (rect.width <= 0 || rect.height <= 0) return false;
+      const style = window.getComputedStyle(node);
+      return style.display !== 'none' && style.visibility !== 'hidden';
+    };
+    const isEnabled = (node) => {
+      const ariaDisabled = node.getAttribute('aria-disabled');
+      const dataDisabled = node.getAttribute('data-disabled');
+      const style = window.getComputedStyle(node);
+      return !(
+        node.hasAttribute('disabled') ||
+        ariaDisabled === 'true' ||
+        dataDisabled === 'true' ||
+        style.pointerEvents === 'none' ||
+        style.display === 'none'
+      );
+    };
+    const candidates = [];
     for (const selector of selectors) {
-      button = document.querySelector(selector);
-      if (button) break;
+      candidates.push(...Array.from(document.querySelectorAll(selector)));
     }
+    const button = candidates.find((node) => isVisible(node) && isEnabled(node)) || null;
     if (!button) return 'missing';
-    const ariaDisabled = button.getAttribute('aria-disabled');
-    const dataDisabled = button.getAttribute('data-disabled');
-    const style = window.getComputedStyle(button);
-    const disabled =
-      button.hasAttribute('disabled') ||
-      ariaDisabled === 'true' ||
-      dataDisabled === 'true' ||
-      style.pointerEvents === 'none' ||
-      style.display === 'none';
-    // Learned: some send buttons render but are inert; only click when truly enabled.
-    if (disabled) return 'disabled';
     // Use unified pointer/mouse sequence to satisfy React handlers.
     dispatchClickSequence(button);
     return 'clicked';
   })()`;
-    const deadline = Date.now() + 8_000;
+    const deadline = Date.now() + 20_000;
     while (Date.now() < deadline) {
         const needAttachment = Array.isArray(attachmentNames) && attachmentNames.length > 0;
         if (needAttachment) {
@@ -332,14 +358,21 @@ async function attemptSendButton(Runtime, _logger, attachmentNames) {
             }
         }
         const { result } = await Runtime.evaluate({ expression: script, returnByValue: true });
-        if (result.value === 'clicked') {
+        if (result.value === "clicked") {
             return true;
         }
-        if (result.value === 'missing') {
+        if (result.value === "missing") {
             break;
         }
         await delay(100);
     }
+    if (Array.isArray(attachmentNames) && attachmentNames.length > 0) {
+        throw new BrowserAutomationError("Attachments never reached a clickable send button before timeout.", {
+            stage: "submit-prompt",
+            code: "attachment-send-not-ready",
+            attachmentNames,
+        });
+    }
     return false;
 }
 async function verifyPromptCommitted(Runtime, prompt, timeoutMs, logger, baselineTurns) {
@@ -351,7 +384,7 @@ async function verifyPromptCommitted(Runtime, prompt, timeoutMs, logger, baselin
     const stopSelectorLiteral = JSON.stringify(STOP_BUTTON_SELECTOR);
     const assistantSelectorLiteral = JSON.stringify(ASSISTANT_ROLE_SELECTOR);
     const turnSelectorLiteral = JSON.stringify(CONVERSATION_TURN_SELECTOR);
-    let baseline = typeof baselineTurns === 'number' && Number.isFinite(baselineTurns) && baselineTurns >= 0
+    let baseline = typeof baselineTurns === "number" && Number.isFinite(baselineTurns) && baselineTurns >= 0
         ? Math.floor(baselineTurns)
         : null;
     if (baseline === null) {
@@ -360,7 +393,7 @@ async function verifyPromptCommitted(Runtime, prompt, timeoutMs, logger, baselin
                 expression: `document.querySelectorAll(${turnSelectorLiteral}).length`,
                 returnByValue: true,
             });
-            const raw = typeof result?.value === 'number' ? result.value : Number(result?.value);
+            const raw = typeof result?.value === "number" ? result.value : Number(result?.value);
             if (Number.isFinite(raw)) {
                 baseline = Math.max(0, Math.floor(raw));
             }
@@ -450,15 +483,15 @@ async function verifyPromptCommitted(Runtime, prompt, timeoutMs, logger, baselin
         const info = result.value;
         const turnsCount = result.value?.turnsCount;
         const matchesPrompt = Boolean(info?.lastMatched || info?.userMatched || info?.prefixMatched);
-        const baselineUnknown = typeof info?.baseline === 'number' ? info.baseline < 0 : baselineLiteral < 0;
+        const baselineUnknown = typeof info?.baseline === "number" ? info.baseline < 0 : baselineLiteral < 0;
         if (matchesPrompt && (baselineUnknown || info?.hasNewTurn)) {
-            return typeof turnsCount === 'number' && Number.isFinite(turnsCount) ? turnsCount : null;
+            return typeof turnsCount === "number" && Number.isFinite(turnsCount) ? turnsCount : null;
         }
         const fallbackCommit = info?.composerCleared &&
             Boolean(info?.hasNewTurn) &&
             ((info?.stopVisible ?? false) || info?.assistantVisible || info?.inConversation);
         if (fallbackCommit) {
-            return typeof turnsCount === 'number' && Number.isFinite(turnsCount) ? turnsCount : null;
+            return typeof turnsCount === "number" && Number.isFinite(turnsCount) ? turnsCount : null;
         }
         await delay(100);
     }
@@ -466,20 +499,23 @@ async function verifyPromptCommitted(Runtime, prompt, timeoutMs, logger, baselin
         logger(`Prompt commit check failed; latest state: ${await Runtime.evaluate({
             expression: script,
             returnByValue: true,
-        }).then((res) => JSON.stringify(res?.result?.value)).catch(() => 'unavailable')}`);
-        await logDomFailure(Runtime, logger, 'prompt-commit');
+        })
+            .then((res) => JSON.stringify(res?.result?.value))
+            .catch(() => "unavailable")}`);
+        await logDomFailure(Runtime, logger, "prompt-commit");
     }
     if (prompt.trim().length >= 50_000) {
-        throw new BrowserAutomationError('Prompt did not appear in conversation before timeout (likely too large).', {
-            stage: 'submit-prompt',
-            code: 'prompt-too-large',
+        throw new BrowserAutomationError("Prompt did not appear in conversation before timeout (likely too large).", {
+            stage: "submit-prompt",
+            code: "prompt-too-large",
             promptLength: prompt.trim().length,
             timeoutMs,
         });
     }
-    throw new Error('Prompt did not appear in conversation before timeout (send may have failed)');
+    throw new Error("Prompt did not appear in conversation before timeout (send may have failed)");
 }
 // biome-ignore lint/style/useNamingConvention: test-only export used in vitest suite
 export const __test__ = {
+    attemptSendButton,
     verifyPromptCommitted,
 };

package/dist/src/browser/actions/remoteFileTransfer.js

@@ -1,9 +1,9 @@
-import path from 'node:path';
-import { FILE_INPUT_SELECTORS } from '../constants.js';
-import { waitForAttachmentVisible } from './attachments.js';
-import { delay } from '../utils.js';
-import { logDomFailure } from '../domDebug.js';
-import { transferAttachmentViaDataTransfer } from './attachmentDataTransfer.js';
+import path from "node:path";
+import { FILE_INPUT_SELECTORS } from "../constants.js";
+import { waitForAttachmentVisible } from "./attachments.js";
+import { delay } from "../utils.js";
+import { logDomFailure } from "../domDebug.js";
+import { transferAttachmentViaDataTransfer } from "./attachmentDataTransfer.js";
 /**
  * Upload file to remote Chrome by transferring content via CDP
  * Used when browser is on a different machine than CLI
@@ -11,7 +11,7 @@ import { transferAttachmentViaDataTransfer } from './attachmentDataTransfer.js';
 export async function uploadAttachmentViaDataTransfer(deps, attachment, logger) {
     const { runtime, dom } = deps;
     if (!dom) {
-        throw new Error('DOM domain unavailable while uploading attachments.');
+        throw new Error("DOM domain unavailable while uploading attachments.");
     }
     logger(`Transferring ${path.basename(attachment.path)} to remote browser...`);
     // Find file input element
@@ -25,13 +25,13 @@ export async function uploadAttachmentViaDataTransfer(deps, attachment, logger)
         }
     }
     if (!fileInputSelector) {
-        await logDomFailure(runtime, logger, 'file-input');
-        throw new Error('Unable to locate ChatGPT file attachment input.');
+        await logDomFailure(runtime, logger, "file-input");
+        throw new Error("Unable to locate ChatGPT file attachment input.");
     }
     const transferResult = await transferAttachmentViaDataTransfer(runtime, attachment, fileInputSelector);
     logger(`File transferred: ${transferResult.fileName} (${transferResult.size} bytes)`);
     // Give ChatGPT a moment to process the file
     await delay(500);
     await waitForAttachmentVisible(runtime, transferResult.fileName, 10_000, logger);
-    logger('Attachment queued');
+    logger("Attachment queued");
 }