Fix attachment uploads: accept-aware input selection, upload-settle wait, scoped stop detection, reply-baseline wait

chatgpt-controller.mjs

@@ -210,9 +210,24 @@ export class ChatGPTController {
 
   async waitForPromptVisible({ timeoutMs = 10 * 60_000, pollMs = 500 } = {}) {
     const start = Date.now();
+    let evalFailures = 0;
     while (Date.now() - start < timeoutMs) {
       this.#throwIfStopRequested();
-      const st = await this.detectChallenge().catch(() => null);
+      // A dead target (window closed, session detached) makes every evaluate
+      // throw; swallowing that forever looked like a silent 10-minute hang.
+      // Surface it fast so callers can recreate the tab.
+      let st = null;
+      try {
+        st = await this.detectChallenge();
+        evalFailures = 0;
+      } catch (error) {
+        evalFailures++;
+        if (evalFailures >= 5) {
+          const err = new Error('tab_target_lost');
+          err.data = { evalFailures, lastError: String(error?.message || error) };
+          throw err;
+        }
+      }
       if (st?.blocked) await this.#enterBlockedState(st);
       if (st?.promptVisible) return st;
 
@@ -285,7 +300,8 @@ export class ChatGPTController {
         const style = window.getComputedStyle(n);
         return r.width > 0 && r.height > 0 && style.visibility !== 'hidden' && style.display !== 'none';
       };
-      const stop = Array.from(document.querySelectorAll(${stopSel})).find(visible);
+      const inScope = (n) => !n.closest('nav, aside, [role="navigation"], [data-testid*="history" i]');
+      const stop = Array.from(document.querySelectorAll(${stopSel})).filter(inScope).find(visible);
       if (!stop) return false;
       try {
         stop.click();
@@ -307,7 +323,25 @@ export class ChatGPTController {
   }
 
   async #typeHuman(text) {
-    for (const ch of String(text)) {
+    const value = String(text);
+    // Human-paced typing is fine for short prompts, but at 12-45ms/char a
+    // large prompt takes half an hour — during which a stray Enter can send a
+    // PARTIAL message and the eventual clickSend hits already_generating.
+    // Large prompts are inserted in bulk (like a paste), with a short typed
+    // tail to keep the composer's input events realistic.
+    if (value.length > 1500) {
+      const head = value.slice(0, -20);
+      const tail = value.slice(-20);
+      await this.page.insertText(head);
+      await sleep(jitter(80, 200));
+      for (const ch of tail) {
+        this.#throwIfStopRequested();
+        await this.page.insertText(ch);
+        await sleep(jitter(12, 45));
+      }
+      return;
+    }
+    for (const ch of value) {
       this.#throwIfStopRequested();
       await this.page.insertText(ch);
       await sleep(jitter(12, 45));
@@ -430,7 +464,8 @@ export class ChatGPTController {
           const style = window.getComputedStyle(n);
           return r.width > 0 && r.height > 0 && style.visibility !== 'hidden' && style.display !== 'none';
         };
-        const stopVisible = Array.from(document.querySelectorAll(${stopSel})).some(visible);
+        const inScope = (n) => !n.closest('nav, aside, [role="navigation"], [data-testid*="history" i]');
+        const stopVisible = Array.from(document.querySelectorAll(${stopSel})).filter(inScope).some(visible);
         const send = Array.from(document.querySelectorAll(${sendSel})).find(visible);
         const sendDisabled = !!send && !!send.disabled;
 
@@ -469,7 +504,11 @@ export class ChatGPTController {
     const sendSel = JSON.stringify(this.selectors.sendButton);
     const stopSel = JSON.stringify(this.selectors.stopButton);
     const res = await this.#eval(`(() => {
-      const stop = Array.from(document.querySelectorAll(${stopSel})).find((n) => {
+      // Scope stop detection to the content area: sidebar history rows can
+      // match loose selectors (e.g. aria-label*="cancel" vs a conversation
+      // titled "…Cancellations") and must never count as "generating".
+      const inScope = (n) => !n.closest('nav, aside, [role="navigation"], [data-testid*="history" i]');
+      const stop = Array.from(document.querySelectorAll(${stopSel})).filter(inScope).find((n) => {
         const r = n.getBoundingClientRect();
         const style = window.getComputedStyle(n);
         return r.width > 0 && r.height > 0 && style.visibility !== 'hidden' && style.display !== 'none';
@@ -706,6 +745,60 @@ export class ChatGPTController {
     await this.page.setFileInputFiles(absFiles);
   }
 
+  // Providers disable send while attachment uploads are in flight. Sending
+  // before they settle races the composer: every send path fails and the run
+  // dies with send_not_triggered while the file tiles are still spinning.
+  // Ready = a visible, enabled send button and no in-flight upload indicator
+  // in the composer, held stable across polls. Budget scales with bytes.
+  async #waitForUploadsSettled({ attachments = [], timeoutMs = null, pollMs = 500 } = {}) {
+    if (!attachments?.length) return;
+    let totalBytes = 0;
+    for (const f of attachments) {
+      try {
+        totalBytes += (await fs.stat(path.resolve(f))).size;
+      } catch {}
+    }
+    const budget = timeoutMs || Math.min(15 * 60_000, 120_000 + Math.ceil(totalBytes / 1_000_000) * 60_000);
+    await this.#emitProgress({ phase: 'uploading_files' });
+    const sendSel = JSON.stringify(this.selectors.sendButton);
+    const start = Date.now();
+    let readySince = null;
+    let lastSnap = null;
+    while (Date.now() - start < budget) {
+      this.#throwIfStopRequested();
+      const snap = await this.#eval(`(() => {
+        const visible = (n) => {
+          if (!n) return false;
+          const r = n.getBoundingClientRect();
+          const style = window.getComputedStyle(n);
+          return r.width > 0 && r.height > 0 && style.visibility !== 'hidden' && style.display !== 'none';
+        };
+        const send = Array.from(document.querySelectorAll(${sendSel})).find(visible) || null;
+        const sendEnabled = send ? !(send.disabled || String(send.getAttribute('aria-disabled') || '').toLowerCase() === 'true') : false;
+        const scope = send?.closest('form') || document.querySelector('main form') || document.querySelector('main') || document.body;
+        const uploading = !!scope.querySelector('[role=\"progressbar\"], progress, [aria-busy=\"true\"]');
+        const failed = /failed to upload|upload failed|couldn't upload|could not upload|error uploading/i.test((scope.innerText || '').slice(0, 4000));
+        return { sendEnabled, uploading, failed };
+      })()`).catch(() => null);
+      lastSnap = snap;
+      if (snap?.failed) {
+        const err = new Error('attachment_upload_failed');
+        err.data = { attachments: attachments.length, totalBytes };
+        throw err;
+      }
+      if (snap && snap.sendEnabled && !snap.uploading) {
+        if (readySince == null) readySince = Date.now();
+        else if (Date.now() - readySince >= 900) return;
+      } else {
+        readySince = null;
+      }
+      await sleep(pollMs);
+    }
+    const err = new Error('attachment_upload_timeout');
+    err.data = { attachments: attachments.length, totalBytes, waitedMs: Date.now() - start, last: lastSnap };
+    throw err;
+  }
+
   async #waitForAssistantStable({ timeoutMs = 5 * 60_000, stableMs = 1500, pollMs = 400 } = {}) {
     await this.#emitProgress({ phase: 'waiting_for_response', blocked: false, blockedKind: null, blockedTitle: null });
     const assistantSel = JSON.stringify(this.selectors.assistantMessage);
@@ -716,11 +809,13 @@ export class ChatGPTController {
     let lastChange = Date.now();
     let stopGoneAt = null;
     let continueClicks = 0;
+    let baseCount = null;
 
     while (Date.now() - start < timeoutMs) {
       this.#throwIfStopRequested();
       const snap = await this.#eval(`(() => {
-        const stop = !!document.querySelector(${stopSel});
+        const inScope = (n) => !n.closest('nav, aside, [role="navigation"], [data-testid*="history" i]');
+        const stop = Array.from(document.querySelectorAll(${stopSel})).some(inScope);
         const send = Array.from(document.querySelectorAll(${sendSel})).find((n) => {
           const r = n.getBoundingClientRect();
           const style = window.getComputedStyle(n);
@@ -734,18 +829,24 @@ export class ChatGPTController {
         const hasContinue = Array.from(document.querySelectorAll('button, a')).some(b => /continue generating/i.test((b.textContent||'').trim()));
         const hasRegenerate = Array.from(document.querySelectorAll('button, a')).some(b => /regenerate/i.test((b.textContent||'').trim()));
         const hasError = /something went wrong|try again|error/i.test(txt) && txt.length < 500;
-        return { stop, sendEnabled, txt, count: nodes.length, usedFallback: !lastNode, hasError, hasContinue, hasRegenerate };
+        return { stop, sendVisible: !!send, sendEnabled, txt, count: nodes.length, usedFallback: !lastNode, hasError, hasContinue, hasRegenerate };
       })()`);
 
       const txt = String(snap?.txt || '');
       if (txt !== last) {
         last = txt;
         lastChange = Date.now();
       }
-
-      // Some providers expose unrelated visible "stop/cancel" controls.
-      // Treat "generating" as stop-visible only when send is not enabled.
-      const generating = !!snap?.stop && !snap?.sendEnabled;
+      // The assistant-message count when we started waiting; a real reply is a
+      // NEW node beyond this baseline (reasoning models can "think" for minutes
+      // before any text appears — pre-existing page text must not count).
+      if (baseCount == null && snap) baseCount = snap.count || 0;
+      const newMsg = snap ? (snap.count || 0) > (baseCount || 0) : false;
+
+      // Generating whenever a stop control is visible and no enabled send
+      // button is; providers replace send with stop while responding, so
+      // "no send button" must NOT read as send-enabled.
+      const generating = !!snap?.stop && !(snap?.sendVisible && snap?.sendEnabled);
       if (generating) stopGoneAt = null;
       else if (stopGoneAt == null) stopGoneAt = Date.now();
 
@@ -763,11 +864,15 @@ export class ChatGPTController {
         continue;
       }
 
-      const readyByNodes = (snap?.count || 0) > 0;
-      const fallbackWaited = !!snap?.usedFallback && (Date.now() - start >= 2500);
+      // A reply counts when a NEW assistant node appeared; the old count>0 test
+      // accepted pre-existing thread content as "the reply". Escape hatches:
+      // selector-less providers (usedFallback) after a long grace, and a
+      // baseline-glitch guard after 2 minutes of stable, non-generating quiet.
+      const readyByNodes = newMsg || ((snap?.count || 0) > 0 && Date.now() - start >= 120_000);
+      const fallbackWaited = !!snap?.usedFallback && (Date.now() - start >= 45_000);
       const fallbackStableLongEnough = txt.length > 0 && (Date.now() - lastChange >= Math.max(dynamicStableMs, 5000));
       const done =
-        (!generating && stopGoneLongEnough && snap?.sendEnabled && stable && txt.length > 0 && (readyByNodes || fallbackWaited)) ||
+        (!generating && stopGoneLongEnough && stable && txt.length > 0 && (readyByNodes || fallbackWaited)) ||
         (!generating && fallbackStableLongEnough && (readyByNodes || fallbackWaited));
       if (done) {
         const extra = await this.#eval(`(() => {
@@ -800,8 +905,9 @@ export class ChatGPTController {
       await this.ensureReady({ timeoutMs });
       await this.#attachFiles(attachments);
       await this.#typePrompt(prompt);
+      await this.#waitForUploadsSettled({ attachments });
       await this.#clickSend();
-      return await this.#waitForAssistantStable({ timeoutMs: Math.min(timeoutMs, 8 * 60_000) });
+      return await this.#waitForAssistantStable({ timeoutMs });   // reasoning models can think for an hour
     } finally {
       if (this.currentRun === run) this.currentRun = null;
     }

chrome-cdp-backend.mjs

@@ -403,6 +403,20 @@ class ChromeCdpPageAdapter {
   }
 
   async setFileInputFiles(files) {
+    // Pick the input whose `accept` is compatible with the files. Sites ship
+    // several file inputs (e.g. chatgpt.com: #upload-files, #upload-photos
+    // accept="image/*", #upload-camera accept="image/*"); feeding documents to
+    // an image-only input routes them into the site's image pipeline, whose
+    // decode fails and leaves the attachment stuck at 0% forever.
+    const isImagePath = (p) => /\.(png|jpe?g|gif|webp|bmp|heic|heif|svg|avif|tiff?)$/i.test(String(p));
+    const allImages = files.every(isImagePath);
+    const acceptCompatible = (accept) => {
+      const a = String(accept || '').trim().toLowerCase();
+      if (!a || a === '*' || a.includes('*/*')) return true;
+      const imageOnly = a.split(',').every((part) => part.trim().startsWith('image/') || /^\.(png|jpe?g|gif|webp|bmp|heic|heif|svg|avif|tiff?)$/.test(part.trim()));
+      return imageOnly ? allImages : true;
+    };
+
     let lastNodeIds = [];
     for (let attempt = 0; attempt < 10; attempt++) {
       const { root } = await this.client.send('DOM.getDocument', { depth: 12, pierce: true }, this.sessionId);
@@ -414,8 +428,22 @@ class ChromeCdpPageAdapter {
         continue;
       }
 
+      const infos = [];
+      for (const nodeId of nodeIds) {
+        let accept = '';
+        try {
+          const attrs = await this.client.send('DOM.getAttributes', { nodeId }, this.sessionId);
+          const list = Array.isArray(attrs?.attributes) ? attrs.attributes : [];
+          for (let i = 0; i + 1 < list.length; i += 2) {
+            if (list[i] === 'accept') accept = list[i + 1];
+          }
+        } catch {}
+        infos.push({ nodeId, accept });
+      }
+      const ordered = [...infos.filter((i) => acceptCompatible(i.accept)), ...infos.filter((i) => !acceptCompatible(i.accept))];
+
       let lastErr = null;
-      for (const nodeId of [...nodeIds].reverse()) {
+      for (const { nodeId } of ordered) {
         try {
           await this.client.send('DOM.setFileInputFiles', { nodeId, files }, this.sessionId);
           lastErr = null;
@@ -594,6 +622,10 @@ export class ChromeCdpBrowserBackend {
         } catch {}
         this.onChanged?.();
       });
+      // Chrome only emits Target.targetDestroyed when discovery is enabled.
+      // Without this, closed tabs were never pruned and later queries hung
+      // on dead sessions until their full timeout.
+      await this.client.send('Target.setDiscoverTargets', { discover: true }).catch(() => {});
       this.started = true;
       return this.getState();
     } catch (error) {

electron-browser-backend.mjs

@@ -76,6 +76,19 @@ class ElectronPageAdapter {
       throw err;
     }
 
+    // Pick the input whose `accept` is compatible with the files — feeding
+    // documents to an image-only input (e.g. chatgpt.com's #upload-camera)
+    // routes them into the site's image pipeline, whose decode fails and
+    // leaves the attachment stuck at 0% forever.
+    const isImagePath = (p) => /\.(png|jpe?g|gif|webp|bmp|heic|heif|svg|avif|tiff?)$/i.test(String(p));
+    const allImages = files.every(isImagePath);
+    const acceptCompatible = (accept) => {
+      const a = String(accept || '').trim().toLowerCase();
+      if (!a || a === '*' || a.includes('*/*')) return true;
+      const imageOnly = a.split(',').every((part) => part.trim().startsWith('image/') || /^\.(png|jpe?g|gif|webp|bmp|heic|heif|svg|avif|tiff?)$/.test(part.trim()));
+      return imageOnly ? allImages : true;
+    };
+
     try {
       let lastNodeIds = [];
       for (let attempt = 0; attempt < 10; attempt++) {
@@ -88,8 +101,22 @@ class ElectronPageAdapter {
           continue;
         }
 
+        const infos = [];
+        for (const nodeId of nodeIds) {
+          let accept = '';
+          try {
+            const attrs = await wc.debugger.sendCommand('DOM.getAttributes', { nodeId });
+            const list = Array.isArray(attrs?.attributes) ? attrs.attributes : [];
+            for (let i = 0; i + 1 < list.length; i += 2) {
+              if (list[i] === 'accept') accept = list[i + 1];
+            }
+          } catch {}
+          infos.push({ nodeId, accept });
+        }
+        const ordered = [...infos.filter((i) => acceptCompatible(i.accept)), ...infos.filter((i) => !acceptCompatible(i.accept))];
+
         let lastErr = null;
-        for (const nodeId of [...nodeIds].reverse()) {
+        for (const { nodeId } of ordered) {
           try {
             await wc.debugger.sendCommand('DOM.setFileInputFiles', { nodeId, files });
             lastErr = null;

http-api.mjs

@@ -820,7 +820,8 @@ export function startHttpApi({
 
       if (url.pathname === '/query' && req.method === 'POST') {
         const body = await parseBody(req, { maxBytes: 5_000_000 });
-        const timeoutMs = positiveIntOr(body.timeoutMs, 10 * 60_000, 30 * 60_000);
+        // Reasoning models can take an hour before responding; allow up to 2h.
+        const timeoutMs = positiveIntOr(body.timeoutMs, 10 * 60_000, 2 * 60 * 60_000);
         const prompt = String(body.prompt || '');
         if (!prompt.trim()) throw new Error('missing_prompt');
         if (prompt.length > 200_000) throw new Error('prompt_too_large');