@auticlabs/bulut 1.2.2 → 1.2.4

package/dist/embed.js

@@ -729,16 +729,15 @@ const MAX_LINKS = 20;
 const MAX_INTERACTABLES = 24;
 const MAX_HEADINGS = 10;
 const MAX_TEXT_SNIPPETS = 4;
-const MAX_OUTER_HTML_DIGEST = 760;
 const MAX_CACHED_PAGES = 20;
 const MAX_PAGE_SCAN_ELEMENTS = 2e3;
 const MAX_EVENT_HINTS_PER_ELEMENT = 4;
 const MAX_BRANCH_SAMPLES = 4;
 const MAX_BRANCH_DEPTH = 2;
-const MAX_CONTEXT_SUMMARY_CHARS = 3400;
-const MAX_CONTEXT_WITH_HISTORY_CHARS = 4200;
-const PAGE_CONTEXT_CACHE_VERSION = 2;
-const PAGE_CONTEXT_CACHE_KEY = "auticbot_page_context_cache_v2";
+const MAX_STYLESHEET_SELECTORS = 80;
+const MAX_STYLESHEET_RULES = 700;
+const PAGE_CONTEXT_CACHE_VERSION = 3;
+const PAGE_CONTEXT_CACHE_KEY = "auticbot_page_context_cache_v3";
 const NON_CONTENT_TAGS = /* @__PURE__ */ new Set([
   "script",
   "style",
@@ -814,22 +813,10 @@ const ARIA_INTERACTION_ATTRS = [
   "aria-selected"
 ];
 const DATA_INTERACTION_PATTERN = /(action|click|press|toggle|target|trigger|nav|open|close|menu|modal|command|submit)/i;
+const STYLESHEET_SELECTOR_PATTERN = /(:hover|:focus|:active|button|a\b|input|textarea|select|\[role=|\[aria-|\[data-|\.btn|\.link)/i;
 const pageContextCache = /* @__PURE__ */ new Map();
 let cacheHydrated = false;
 const normalizeWhitespace = (value) => value.replace(/\s+/g, " ").trim();
-const truncate = (value, maxChars) => {
-  if (value.length <= maxChars) {
-    return value;
-  }
-  const suffix = "\n...[truncated]";
-  return `${value.slice(0, Math.max(0, maxChars - suffix.length))}${suffix}`;
-};
-const truncateInline = (value, maxChars) => {
-  if (value.length <= maxChars) {
-    return value;
-  }
-  return `${value.slice(0, Math.max(0, maxChars - 3))}...`;
-};
 const canonicalUrl = (rawUrl) => {
   try {
     return new URL(rawUrl, rawUrl).href;
@@ -863,9 +850,9 @@ const parseTabIndex = (value) => {
   const parsed = Number.parseInt(value, 10);
   return Number.isNaN(parsed) ? null : parsed;
 };
-const compactToken = (value, maxChars = 18) => {
+const compactToken = (value) => {
   const compact = value.replace(/\s+/g, "-").replace(/[^a-zA-Z0-9_-]/g, "");
-  return compact ? truncateInline(compact, maxChars) : "";
+  return compact || "";
 };
 const getElementDepth = (element) => {
   let depth = 0;
@@ -950,16 +937,13 @@ const buildSummaryWithHistory = (current) => {
   const historySection = [
     "Recent Page Memory:",
     ...recentPages.map((entry) => {
-      const compactSummary = normalizeWhitespace(entry.summary).slice(0, 180);
+      const compactSummary = normalizeWhitespace(entry.summary);
       return `- ${entry.url} :: ${compactSummary}`;
     })
   ].join("\n");
-  return truncate(
-    `${current.summary}
+  return `${current.summary}
 
-${historySection}`,
-    MAX_CONTEXT_WITH_HISTORY_CHARS
-  );
+${historySection}`;
 };
 const isVisible = (element) => {
   if (element.getAttribute("aria-hidden") === "true") {
@@ -988,20 +972,25 @@ const escapeCssValue = (value) => {
   }
   return value.replace(/([ #;&,.+*~':"!^$\[\]()=>|\/@])/g, "\\$1");
 };
-const buildSelector = (element) => {
+const buildSelectorSegment = (element) => {
   const tag = element.tagName.toLowerCase();
   if (element.id) {
     return `#${escapeCssValue(element.id)}`;
   }
-  const name = element.getAttribute("name");
-  if (name) {
-    return `${tag}[name="${escapeCssValue(name)}"]`;
-  }
-  const ariaLabel = element.getAttribute("aria-label");
-  if (ariaLabel) {
-    return `${tag}[aria-label="${escapeCssValue(ariaLabel)}"]`;
+  const attrCandidates = [
+    ["name", element.getAttribute("name")],
+    ["data-testid", element.getAttribute("data-testid")],
+    ["data-test-id", element.getAttribute("data-test-id")],
+    ["aria-label", element.getAttribute("aria-label")],
+    ["role", element.getAttribute("role")],
+    ["type", element.getAttribute("type")]
+  ];
+  for (const [attrName, attrValue] of attrCandidates) {
+    if (attrValue) {
+      return `${tag}[${attrName}="${escapeCssValue(attrValue)}"]`;
+    }
   }
-  const classes = Array.from(element.classList).filter(Boolean).slice(0, 2).map((className) => `.${escapeCssValue(className)}`).join("");
+  const classes = Array.from(element.classList).filter(Boolean).slice(0, 3).map((className) => `.${escapeCssValue(className)}`).join("");
   if (classes) {
     return `${tag}${classes}`;
   }
@@ -1015,6 +1004,21 @@ const buildSelector = (element) => {
   const index = siblingsOfTag.indexOf(element) + 1;
   return `${tag}:nth-of-type(${index})`;
 };
+const buildSelector = (element) => {
+  const segments = [];
+  let cursor = element;
+  let depth = 0;
+  while (cursor && depth < 4) {
+    const segment = buildSelectorSegment(cursor);
+    segments.unshift(segment);
+    if (segment.startsWith("#")) {
+      break;
+    }
+    cursor = cursor.parentElement;
+    depth += 1;
+  }
+  return segments.join(" > ");
+};
 const getElementLabel = (element) => {
   const text = normalizeWhitespace(
     (element instanceof HTMLElement ? element.innerText : element.textContent) || ""
@@ -1026,7 +1030,7 @@ const getElementLabel = (element) => {
   );
   const name = normalizeWhitespace(element.getAttribute("name") || "");
   const value = element instanceof HTMLInputElement || element instanceof HTMLTextAreaElement || element instanceof HTMLButtonElement ? normalizeWhitespace(element.value || "") : "";
-  const classHint = Array.from(element.classList).map((item) => compactToken(item, 16)).find(Boolean);
+  const classHint = Array.from(element.classList).map((item) => compactToken(item)).find(Boolean);
   const fallback = element.id && `#${element.id}` || classHint && `.${classHint}` || buildSelector(element);
   const label = text || ariaLabel || title || placeholder || value || name || fallback;
   if (element.tagName.toLowerCase() === "input") {
@@ -1058,23 +1062,35 @@ const getAriaInteractionHints = (element) => ARIA_INTERACTION_ATTRS.filter((attr
 const getDataInteractionHints = (element) => element.getAttributeNames().filter(
   (attrName) => attrName.startsWith("data-") && DATA_INTERACTION_PATTERN.test(attrName)
 ).slice(0, 2).map((attrName) => attrName.replace("data-", ""));
-const getStyleHints = (style) => {
-  const styleHints = [];
-  if (style.cursor === "pointer") {
-    styleHints.push("cursor:pointer");
+const getComputedStyleSignals = (style) => {
+  const signals = [];
+  if (style.cursor && style.cursor !== "auto") {
+    signals.push(`cursor:${style.cursor}`);
+  }
+  if (style.display) {
+    signals.push(`display:${style.display}`);
+  }
+  if (style.position) {
+    signals.push(`position:${style.position}`);
+  }
+  if (style.zIndex && style.zIndex !== "auto") {
+    signals.push(`z-index:${style.zIndex}`);
   }
-  if (style.display === "flex" || style.display === "grid" || style.display === "inline-flex" || style.display === "inline-grid") {
-    styleHints.push(`display:${style.display}`);
+  if (style.pointerEvents && style.pointerEvents !== "auto") {
+    signals.push(`pointer-events:${style.pointerEvents}`);
   }
-  if (style.position === "fixed" || style.position === "sticky") {
-    styleHints.push(`position:${style.position}`);
+  if (style.visibility && style.visibility !== "visible") {
+    signals.push(`visibility:${style.visibility}`);
   }
-  return styleHints.slice(0, 2);
+  if (style.opacity && style.opacity !== "1") {
+    signals.push(`opacity:${style.opacity}`);
+  }
+  return Array.from(new Set(signals));
 };
 const buildBlueprintToken = (element) => {
   const tag = element.tagName.toLowerCase();
   const idToken = element.id ? `#${compactToken(element.id)}` : "";
-  const classToken = Array.from(element.classList).map((item) => compactToken(item, 16)).find(Boolean);
+  const classToken = Array.from(element.classList).map((item) => compactToken(item)).find(Boolean);
   return `${tag}${idToken}${classToken ? `.${classToken}` : ""}`;
 };
 const buildBranchDigest = (element, depth) => {
@@ -1094,7 +1110,7 @@ const collectDomBranchDigest = () => {
   const root = document.body ?? document.documentElement;
   const topLevelNodes = Array.from(root.children).filter((child) => !NON_CONTENT_TAGS.has(child.tagName.toLowerCase())).filter((child) => isVisible(child)).slice(0, MAX_BRANCH_SAMPLES);
   return topLevelNodes.map(
-    (child) => truncateInline(buildBranchDigest(child, MAX_BRANCH_DEPTH), 140)
+    (child) => buildBranchDigest(child, MAX_BRANCH_DEPTH)
   );
 };
 const formatSection = (title, lines) => {
@@ -1110,7 +1126,7 @@ const buildOuterHtmlDigest = () => {
   const raw = ((_a = document.body) == null ? void 0 : _a.outerHTML) || document.documentElement.outerHTML;
   const withoutScripts = raw.replace(/<script[\s\S]*?<\/script>/gi, "").replace(/<style[\s\S]*?<\/style>/gi, "").replace(/<noscript[\s\S]*?<\/noscript>/gi, "").replace(/<!--[\s\S]*?-->/g, "").replace(/\s+/g, " ").trim();
   const structural = withoutScripts.replace(/>[^<]*</g, "><").replace(/\s+/g, " ").trim();
-  return truncate(structural, MAX_OUTER_HTML_DIGEST);
+  return structural;
 };
 const collectTextSnippets = () => {
   const root = document.querySelector("main, article, [role='main']") ?? document.body;
@@ -1125,12 +1141,11 @@ const collectTextSnippets = () => {
     if (!text || text.length < 20) {
       continue;
     }
-    const compact = truncateInline(text, 180);
-    if (seen.has(compact)) {
+    if (seen.has(text)) {
       continue;
     }
-    seen.add(compact);
-    snippets.push(`- ${compact}`);
+    seen.add(text);
+    snippets.push(`- ${text}`);
     if (snippets.length >= MAX_TEXT_SNIPPETS) {
       break;
     }
@@ -1161,6 +1176,51 @@ const collectLandmarkSnapshot = () => {
     ({ label, selector }) => `- ${label}: ${document.querySelectorAll(selector).length}`
   );
 };
+const collectSelectorsFromRuleList = (rules, selectors, scanned) => {
+  for (const rule of Array.from(rules)) {
+    if (scanned.count >= MAX_STYLESHEET_RULES || selectors.size >= MAX_STYLESHEET_SELECTORS) {
+      return;
+    }
+    scanned.count += 1;
+    if (rule instanceof CSSStyleRule) {
+      const parts = rule.selectorText.split(",").map((selector) => normalizeWhitespace(selector)).filter(Boolean);
+      for (const selector of parts) {
+        if (!STYLESHEET_SELECTOR_PATTERN.test(selector)) {
+          continue;
+        }
+        selectors.add(selector);
+        if (selectors.size >= MAX_STYLESHEET_SELECTORS) {
+          return;
+        }
+      }
+      continue;
+    }
+    if ("cssRules" in rule) {
+      try {
+        const nestedRules = rule.cssRules;
+        collectSelectorsFromRuleList(nestedRules, selectors, scanned);
+      } catch {
+      }
+    }
+  }
+};
+const collectStylesheetSelectors = () => {
+  const selectors = /* @__PURE__ */ new Set();
+  const scanned = { count: 0 };
+  for (const stylesheet of Array.from(document.styleSheets)) {
+    if (scanned.count >= MAX_STYLESHEET_RULES || selectors.size >= MAX_STYLESHEET_SELECTORS) {
+      break;
+    }
+    try {
+      if (!stylesheet.cssRules) {
+        continue;
+      }
+      collectSelectorsFromRuleList(stylesheet.cssRules, selectors, scanned);
+    } catch {
+    }
+  }
+  return Array.from(selectors).map((selector) => `- ${selector}`);
+};
 const collectPageSignalSnapshot = () => {
   const allElements = Array.from(document.querySelectorAll("*"));
   const sampledElements = allElements.slice(0, MAX_PAGE_SCAN_ELEMENTS);
@@ -1172,6 +1232,7 @@ const collectPageSignalSnapshot = () => {
   const eventCounts = /* @__PURE__ */ new Map();
   const displayCounts = /* @__PURE__ */ new Map();
   const positionCounts = /* @__PURE__ */ new Map();
+  const styleSignalCounts = /* @__PURE__ */ new Map();
   let visibleElements = 0;
   let maxDepth = 0;
   let semanticInteractables = 0;
@@ -1207,6 +1268,10 @@ const collectPageSignalSnapshot = () => {
     if (TRACKED_POSITION_VALUES.has(style.position)) {
       bumpCount(positionCounts, style.position);
     }
+    const computedStyleSignals = getComputedStyleSignals(style);
+    for (const styleSignal of computedStyleSignals) {
+      bumpCount(styleSignalCounts, styleSignal);
+    }
     const eventHints = getEventHints(element);
     if (eventHints.length > 0) {
       for (const eventName of eventHints) {
@@ -1241,7 +1306,7 @@ const collectPageSignalSnapshot = () => {
     if (tag === "a" && href && !href.startsWith("#") && !href.startsWith("javascript:")) {
       const absoluteHref = toAbsoluteUrl(href);
       const label2 = getElementLabel(element) || absoluteHref;
-      const line2 = `- ${truncateInline(label2, 90)} -> ${truncateInline(absoluteHref, 140)}`;
+      const line2 = `- ${label2} -> ${absoluteHref}`;
       if (!linkSet.has(line2)) {
         linkSet.add(line2);
         links.push(line2);
@@ -1257,8 +1322,8 @@ const collectPageSignalSnapshot = () => {
       nonSemanticInteractables += 1;
     }
     const selector = buildSelector(element);
-    const label = truncateInline(getElementLabel(element), 90);
-    const styleHints = getStyleHints(style);
+    const label = getElementLabel(element);
+    const styleSignals = computedStyleSignals;
     const signalTokens = [];
     if (eventHints.length > 0) {
       signalTokens.push(`evt:${eventHints.join("|")}`);
@@ -1275,16 +1340,13 @@ const collectPageSignalSnapshot = () => {
     if (ariaHints.length > 0) {
       signalTokens.push(`aria:${ariaHints.join("|")}`);
     }
-    if (styleHints.length > 0) {
-      signalTokens.push(`css:${styleHints.join("|")}`);
+    if (styleSignals.length > 0) {
+      signalTokens.push(`css:${styleSignals.join("|")}`);
     } else if (hasPointerCursor) {
       signalTokens.push("css:cursor:pointer");
     }
     const signalBlock = signalTokens.length > 0 ? ` [${signalTokens.join("; ")}]` : "";
-    const line = truncateInline(
-      `- ${tag} ${selector}${signalBlock} (${label})`,
-      240
-    );
+    const line = `- ${tag} ${selector}${signalBlock} (${label})`;
     const score = eventHints.length * 5 + (isNativeInteractive ? 5 : 0) + (isRoleInteractive ? 4 : 0) + (hasTabStop ? 2 : 0) + (hasPointerCursor ? 2 : 0) + (dataHints.length > 0 ? 2 : 0) + (ariaHints.length > 0 ? 1 : 0) + (isContentEditable ? 2 : 0);
     const existing = interactableCandidates.get(line);
     if (!existing || score > existing.score) {
@@ -1302,8 +1364,9 @@ const collectPageSignalSnapshot = () => {
     `- listener hints: ${formatTopCounts(eventCounts, 8)}`,
     `- interaction cues: tabindex>=0=${tabStopElements}, pointer-cursor=${pointerCursorElements}, data-hints=${dataHintElements}, aria-hints=${ariaHintElements}`,
     `- role hints: ${formatTopCounts(interactiveRoleCounts, 8)}`,
+    `- css footprint: ${formatTopCounts(styleSignalCounts, 10)}`,
     "- listener scope: inline/on* handlers are detected directly; addEventListener handlers are inferred via cues."
-  ].map((line) => truncateInline(line, 250));
+  ];
   const branchDigest = collectDomBranchDigest();
   const pageBlueprint = [
     `- nodes: total=${allElements.length}, scanned=${sampledElements.length}, visible=${visibleElements}, max-depth=${maxDepth}${allElements.length > sampledElements.length ? ", sampling=on" : ""}`,
@@ -1311,15 +1374,16 @@ const collectPageSignalSnapshot = () => {
     `- role density: ${formatTopCounts(roleCounts, 8)}`,
     `- layout density: display(${formatTopCounts(displayCounts, 6)}), position(${formatTopCounts(positionCounts, 4)})`,
     `- branch digest: ${branchDigest.length > 0 ? branchDigest.join(" || ") : "none"}`
-  ].map((line) => truncateInline(line, 260));
+  ];
   return {
     links: links.slice(0, MAX_LINKS),
     interactables,
     interactionSignals,
+    styleSelectors: collectStylesheetSelectors(),
     pageBlueprint
   };
 };
-const buildPageContextSummary = (input, maxChars = MAX_CONTEXT_SUMMARY_CHARS) => {
+const buildPageContextSummary = (input) => {
   const sections = [
     formatSection("Meta", [
       `- URL: ${input.url || "unknown"}`,
@@ -1329,6 +1393,7 @@ const buildPageContextSummary = (input, maxChars = MAX_CONTEXT_SUMMARY_CHARS) =>
     formatSection("Headings", input.headings),
     formatSection("Landmark Snapshot", input.landmarks),
     formatSection("Interaction Signals", input.interactionSignals),
+    formatSection("Stylesheet Selector Snapshot", input.styleSelectors),
     formatSection("Compressed Page Blueprint", input.pageBlueprint),
     formatSection("Top Links", input.links),
     formatSection("Top Interactables", input.interactables),
@@ -1337,7 +1402,7 @@ const buildPageContextSummary = (input, maxChars = MAX_CONTEXT_SUMMARY_CHARS) =>
       `- ${input.outerHtmlDigest || "unavailable"}`
     ])
   ];
-  return truncate(sections.join("\n\n"), maxChars);
+  return sections.join("\n\n");
 };
 const getPageContext = () => {
   if (typeof window === "undefined" || typeof document === "undefined") {
@@ -1361,7 +1426,7 @@ const getPageContext = () => {
   console.info(`[Autic] context cache miss url=${url}`);
   const snapshot = collectPageSignalSnapshot();
   const headings = Array.from(document.querySelectorAll("h1, h2, h3")).filter((element) => isVisible(element)).map(
-    (element) => `- ${truncateInline(normalizeWhitespace(element.textContent || ""), 120)}`
+    (element) => `- ${normalizeWhitespace(element.textContent || "")}`
   ).filter((line) => line !== "- ").slice(0, MAX_HEADINGS);
   const summary = buildPageContextSummary({
     url,
@@ -1372,6 +1437,7 @@ const getPageContext = () => {
     links: snapshot.links,
     interactables: snapshot.interactables,
     interactionSignals: snapshot.interactionSignals,
+    styleSelectors: snapshot.styleSelectors,
     pageBlueprint: snapshot.pageBlueprint,
     textSnippets: collectTextSnippets(),
     outerHtmlDigest: buildOuterHtmlDigest()
@@ -1516,6 +1582,18 @@ const getCursorPosition = (cursor) => ({
 const setCursorVisibility = (cursor, visible) => {
   cursor.style.opacity = visible ? "1" : "0";
 };
+const hideAgentCursor = () => {
+  if (typeof document === "undefined" || typeof window === "undefined") {
+    return;
+  }
+  const cursor = document.getElementById(AGENT_CURSOR_ID);
+  if (!(cursor instanceof HTMLElement)) {
+    return;
+  }
+  const { x: x2, y: y2 } = getCursorPosition(cursor);
+  setCursorVisibility(cursor, false);
+  persistCursorState(x2, y2, false);
+};
 let cursorHoverTrackingInitialized = false;
 const initializeCursorHoverTracking = () => {
   if (cursorHoverTrackingInitialized) {
@@ -1954,6 +2032,9 @@ if (typeof document !== "undefined") {
   }
 }
 const TTS_WS_RETRY_DELAYS_MS = [250, 750, 1500];
+const BULUT_AUDIO_STOP_EVENT = "bulut:audio-stop";
+const activeAudioElements = /* @__PURE__ */ new Set();
+let audioPlaybackGeneration = 0;
 const normalizeBaseUrl = (baseUrl) => {
   const trimmed = baseUrl.trim().replace(/\/+$/, "");
   if (/^https?:\/\//i.test(trimmed)) {
@@ -2017,6 +2098,27 @@ const parseErrorBody = async (response) => {
 const sleep = (ms) => new Promise((resolve) => {
   setTimeout(resolve, ms);
 });
+const registerActiveAudioElement = (audioElement) => {
+  activeAudioElements.add(audioElement);
+};
+const unregisterActiveAudioElement = (audioElement) => {
+  activeAudioElements.delete(audioElement);
+};
+const wasPlaybackStoppedAfter = (generationAtStart) => audioPlaybackGeneration !== generationAtStart;
+const getAudioPlaybackGeneration = () => audioPlaybackGeneration;
+const stopActiveAudioPlayback = () => {
+  audioPlaybackGeneration += 1;
+  const active = Array.from(activeAudioElements);
+  for (const audioElement of active) {
+    try {
+      audioElement.dispatchEvent(new Event(BULUT_AUDIO_STOP_EVENT));
+      audioElement.pause();
+      audioElement.removeAttribute("src");
+      audioElement.load();
+    } catch {
+    }
+  }
+};
 const base64ToUint8Array = (base64) => {
   const cleanBase64 = base64.replace(/^data:audio\/\w+;base64,/, "");
   const binaryString = atob(cleanBase64);
@@ -2072,20 +2174,31 @@ const waitForPlaybackEnd = async (audioElement) => {
       cleanup();
       reject(new Error("Ses oynatma hatası oluştu."));
     };
+    const onForcedStop = () => {
+      cleanup();
+      resolve();
+    };
     const cleanup = () => {
       window.clearInterval(watchdog);
       audioElement.removeEventListener("ended", onEnded);
       audioElement.removeEventListener("error", onError);
+      audioElement.removeEventListener(BULUT_AUDIO_STOP_EVENT, onForcedStop);
     };
     audioElement.addEventListener("ended", onEnded);
     audioElement.addEventListener("error", onError);
+    audioElement.addEventListener(BULUT_AUDIO_STOP_EVENT, onForcedStop);
   });
 };
 const playBufferedAudio = async (chunks, mimeType, sampleRate = 16e3, onAudioStateChange) => {
+  const playbackGeneration = getAudioPlaybackGeneration();
   if (chunks.length === 0) {
     onAudioStateChange == null ? void 0 : onAudioStateChange("done");
     return;
   }
+  if (wasPlaybackStoppedAfter(playbackGeneration)) {
+    onAudioStateChange == null ? void 0 : onAudioStateChange("done");
+    return;
+  }
   const totalBytes = chunks.reduce((acc, c2) => acc + c2.byteLength, 0);
   console.log(`[Bulut] Playing buffered audio: ${chunks.length} chunks, ${totalBytes} bytes, type=${mimeType}`);
   onAudioStateChange == null ? void 0 : onAudioStateChange("fallback");
@@ -2122,10 +2235,15 @@ const playBufferedAudio = async (chunks, mimeType, sampleRate = 16e3, onAudioSta
   const audioElement = new Audio();
   const objectUrl = URL.createObjectURL(blob);
   try {
+    registerActiveAudioElement(audioElement);
     audioElement.preload = "auto";
     audioElement.autoplay = true;
     audioElement.setAttribute("playsinline", "true");
     audioElement.src = objectUrl;
+    if (wasPlaybackStoppedAfter(playbackGeneration)) {
+      onAudioStateChange == null ? void 0 : onAudioStateChange("done");
+      return;
+    }
     await audioElement.play();
     onAudioStateChange == null ? void 0 : onAudioStateChange("playing");
     await waitForPlaybackEnd(audioElement);
@@ -2135,6 +2253,7 @@ const playBufferedAudio = async (chunks, mimeType, sampleRate = 16e3, onAudioSta
     onAudioStateChange == null ? void 0 : onAudioStateChange("done");
     throw err;
   } finally {
+    unregisterActiveAudioElement(audioElement);
     audioElement.pause();
     audioElement.removeAttribute("src");
     audioElement.load();
@@ -2175,6 +2294,7 @@ async function transcribeAudio(baseUrl, file, projectId, sessionId, language, on
 }
 const startSttWebSocketStream = (baseUrl, config, events = {}) => {
   const wsUrl = toWebSocketUrl(baseUrl, "/chat/stt/ws");
+  console.info("[Bulut] STT WS connecting to", wsUrl);
   const socket = new WebSocket(wsUrl);
   let seq = 0;
   let finalText = "";
@@ -2197,6 +2317,7 @@ const startSttWebSocketStream = (baseUrl, config, events = {}) => {
   const rejectAll = (error) => {
     if (settled) return;
     settled = true;
+    console.warn("[Bulut] STT WS rejected:", error.message);
     rejectStart == null ? void 0 : rejectStart(error);
     rejectDone == null ? void 0 : rejectDone(error);
   };
@@ -2210,15 +2331,9 @@ const startSttWebSocketStream = (baseUrl, config, events = {}) => {
       session_id: finalSessionId
     });
   };
-  const enqueueSend = (payload) => {
-    sendQueue = sendQueue.then(() => {
-      if (stopped || socket.readyState !== WebSocket.OPEN) return;
-      socket.send(JSON.stringify(payload));
-    });
-    return sendQueue;
-  };
   socket.onopen = () => {
     var _a;
+    console.info("[Bulut] STT WS connected, sending start");
     (_a = events.onRequestSent) == null ? void 0 : _a.call(events);
     socket.send(
       JSON.stringify({
@@ -2235,6 +2350,7 @@ const startSttWebSocketStream = (baseUrl, config, events = {}) => {
     const payload = parseSttWsEventPayload(String(event.data));
     if (!payload) return;
     if (payload.type === "start_ack" && typeof payload.session_id === "string") {
+      console.info("[Bulut] STT WS start_ack received, session:", payload.session_id);
       finalSessionId = payload.session_id;
       (_a = events.onSessionId) == null ? void 0 : _a.call(events, payload.session_id);
       resolveStart == null ? void 0 : resolveStart();
@@ -2248,6 +2364,7 @@ const startSttWebSocketStream = (baseUrl, config, events = {}) => {
       return;
     }
     if (payload.type === "final" && typeof payload.text === "string") {
+      console.info("[Bulut] STT WS final text received:", payload.text.slice(0, 80));
       finalText = payload.text;
       if (typeof payload.session_id === "string") {
         finalSessionId = payload.session_id;
@@ -2256,20 +2373,24 @@ const startSttWebSocketStream = (baseUrl, config, events = {}) => {
       return;
     }
     if (payload.type === "done") {
+      console.info("[Bulut] STT WS done");
       resolveDoneIfPossible();
       socket.close();
       return;
     }
     if (payload.type === "error") {
+      console.error("[Bulut] STT WS server error:", payload.error);
       const err = buildError(payload.error || "stt_ws_error", payload.retryable !== false);
       rejectAll(err);
       socket.close();
     }
   };
-  socket.onerror = () => {
+  socket.onerror = (ev) => {
+    console.error("[Bulut] STT WS transport error", ev);
     rejectAll(buildError("stt_ws_transport_error", true));
   };
-  socket.onclose = () => {
+  socket.onclose = (ev) => {
+    console.info("[Bulut] STT WS closed code=%d reason=%s", ev.code, ev.reason);
     if (settled) return;
     if (finalText && finalSessionId) {
       resolveDoneIfPossible();
@@ -2278,23 +2399,30 @@ const startSttWebSocketStream = (baseUrl, config, events = {}) => {
     rejectAll(buildError("stt_ws_closed_before_done", true));
   };
   return {
-    pushChunk: async (chunk) => {
-      if (stopped || chunk.size === 0) return;
-      await startPromise;
-      if (stopped) return;
-      const audio = await blobToBase64(chunk);
-      seq += 1;
-      await enqueueSend({
-        type: "chunk",
-        seq,
-        audio
+    pushChunk: (chunk) => {
+      if (stopped || chunk.size === 0) return Promise.resolve();
+      sendQueue = sendQueue.then(async () => {
+        if (stopped) return;
+        await startPromise;
+        if (stopped) return;
+        const audio = await blobToBase64(chunk);
+        seq += 1;
+        if (stopped || socket.readyState !== WebSocket.OPEN) return;
+        console.debug("[Bulut] STT WS sending chunk seq=%d size=%d", seq, chunk.size);
+        socket.send(JSON.stringify({ type: "chunk", seq, audio }));
       });
+      return sendQueue;
     },
-    stop: async () => {
-      await startPromise;
-      if (!stopped) {
-        await enqueueSend({ type: "stop" });
-      }
+    stop: () => {
+      console.info("[Bulut] STT WS stop requested, draining %d pending chunks", seq);
+      sendQueue = sendQueue.then(async () => {
+        await startPromise;
+        if (stopped) return;
+        if (socket.readyState === WebSocket.OPEN) {
+          console.info("[Bulut] STT WS sending stop after seq=%d", seq);
+          socket.send(JSON.stringify({ type: "stop" }));
+        }
+      });
       return donePromise;
     },
     cancel: () => {
@@ -2499,6 +2627,50 @@ const collectTtsViaWebSocket = async (baseUrl, assistantText, voice, accessibili
   }
   throw buildError("tts_ws_exhausted", true);
 };
+const speakText = async (baseUrl, text, voice, accessibilityMode, onAudioStateChange) => {
+  const trimmed = text.trim();
+  if (!trimmed) return;
+  const playbackGeneration = getAudioPlaybackGeneration();
+  console.info(`[Bulut] speakText start (${trimmed.length} chars)`);
+  onAudioStateChange == null ? void 0 : onAudioStateChange("rendering");
+  let ttsResult;
+  const neverStopped = () => false;
+  try {
+    ttsResult = await collectTtsViaWebSocket(
+      baseUrl,
+      trimmed,
+      voice,
+      accessibilityMode,
+      neverStopped,
+      () => {
+      }
+    );
+  } catch {
+    ttsResult = await collectTtsViaSse(
+      baseUrl,
+      trimmed,
+      voice,
+      accessibilityMode,
+      neverStopped,
+      () => {
+      }
+    );
+  }
+  if (wasPlaybackStoppedAfter(playbackGeneration)) {
+    onAudioStateChange == null ? void 0 : onAudioStateChange("done");
+    return;
+  }
+  if (ttsResult.chunks.length > 0) {
+    await playBufferedAudio(
+      ttsResult.chunks,
+      ttsResult.mimeType,
+      ttsResult.sampleRate,
+      onAudioStateChange
+    );
+  } else {
+    onAudioStateChange == null ? void 0 : onAudioStateChange("done");
+  }
+};
 const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config, events, executeTool) => {
   let isStopped = false;
   let activeSocket = null;
@@ -2534,6 +2706,7 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
         activeSocket = socket;
         let finalReply = "";
         let resolved = false;
+        let accumulatedDelta = "";
         const finish = (reply) => {
           if (resolved) return;
           resolved = true;
@@ -2557,7 +2730,7 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
           }));
         };
         socket.onmessage = async (event) => {
-          var _a2, _b2, _c2, _d2, _e, _f, _g, _h;
+          var _a2, _b2, _c2, _d2, _e, _f, _g, _h, _i;
           let data;
           try {
             data = JSON.parse(String(event.data));
@@ -2580,12 +2753,17 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
             return;
           }
           if (msgType === "reply_delta" && typeof data.delta === "string") {
+            accumulatedDelta += data.delta;
             (_c2 = events.onAssistantDelta) == null ? void 0 : _c2.call(events, data.delta);
             return;
           }
           if (msgType === "tool_calls" && Array.isArray(data.calls)) {
             const calls = data.calls;
-            (_d2 = events.onToolCalls) == null ? void 0 : _d2.call(events, calls);
+            if (accumulatedDelta.trim()) {
+              (_d2 = events.onIntermediateReply) == null ? void 0 : _d2.call(events, accumulatedDelta.trim());
+            }
+            accumulatedDelta = "";
+            (_e = events.onToolCalls) == null ? void 0 : _e.call(events, calls);
             const results = [];
             for (const call of calls) {
               const isNavigate = call.tool === "navigate";
@@ -2608,7 +2786,7 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
               if (isNavigate) {
                 clearPendingAgentResume();
               }
-              (_e = events.onToolResult) == null ? void 0 : _e.call(events, call.call_id, call.tool, result.result);
+              (_f = events.onToolResult) == null ? void 0 : _f.call(events, call.call_id, call.tool, result.result);
               results.push(result);
             }
             if (socket.readyState === WebSocket.OPEN) {
@@ -2621,9 +2799,9 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
           }
           if (msgType === "agent_done") {
             finalReply = data.final_reply || "";
-            (_f = events.onAssistantDone) == null ? void 0 : _f.call(events, finalReply);
+            (_g = events.onAssistantDone) == null ? void 0 : _g.call(events, finalReply);
             if (typeof data.session_id === "string") {
-              (_g = events.onSessionId) == null ? void 0 : _g.call(events, data.session_id);
+              (_h = events.onSessionId) == null ? void 0 : _h.call(events, data.session_id);
             }
             finish(finalReply);
             return;
@@ -2631,7 +2809,7 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
           if (msgType === "error") {
             const errMsg = data.error || "Agent error";
             errorEmitted = true;
-            (_h = events.onError) == null ? void 0 : _h.call(events, errMsg);
+            (_i = events.onError) == null ? void 0 : _i.call(events, errMsg);
             fail(new Error(errMsg));
             return;
           }
@@ -2650,6 +2828,9 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
       });
       activeSocket = null;
       if (isStopped || !assistantText) {
+        if (!isStopped) {
+          hideAgentCursor();
+        }
         return resolve();
       }
       console.info(
@@ -2694,6 +2875,9 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
       } else {
         (_c = events.onAudioStateChange) == null ? void 0 : _c.call(events, "done");
       }
+      if (!isStopped) {
+        hideAgentCursor();
+      }
       resolve();
     } catch (err) {
       if (!errorEmitted) {
@@ -2713,6 +2897,7 @@ const agentVoiceChatStream = (baseUrl, audioFile, projectId, sessionId, config,
   return {
     stop: () => {
       isStopped = true;
+      stopActiveAudioPlayback();
       if (activeReader) {
         activeReader.cancel().catch(() => {
         });
@@ -2744,6 +2929,7 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
         let finalReply = "";
         let resolved = false;
         let effectiveSessionId = sessionId || "";
+        let accumulatedDelta = "";
         const finish = (reply) => {
           if (resolved) return;
           resolved = true;
@@ -2766,7 +2952,7 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
           }));
         };
         socket.onmessage = async (event) => {
-          var _a2, _b2, _c2, _d, _e, _f, _g, _h;
+          var _a2, _b2, _c2, _d, _e, _f, _g, _h, _i;
           let data;
           try {
             data = JSON.parse(String(event.data));
@@ -2788,12 +2974,17 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
             return;
           }
           if (msgType === "reply_delta" && typeof data.delta === "string") {
+            accumulatedDelta += data.delta;
             (_c2 = events.onAssistantDelta) == null ? void 0 : _c2.call(events, data.delta);
             return;
           }
           if (msgType === "tool_calls" && Array.isArray(data.calls)) {
             const calls = data.calls;
-            (_d = events.onToolCalls) == null ? void 0 : _d.call(events, calls);
+            if (accumulatedDelta.trim()) {
+              (_d = events.onIntermediateReply) == null ? void 0 : _d.call(events, accumulatedDelta.trim());
+            }
+            accumulatedDelta = "";
+            (_e = events.onToolCalls) == null ? void 0 : _e.call(events, calls);
             const results = [];
             for (const call of calls) {
               const isNavigate = call.tool === "navigate";
@@ -2816,7 +3007,7 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
               if (isNavigate) {
                 clearPendingAgentResume();
               }
-              (_e = events.onToolResult) == null ? void 0 : _e.call(events, call.call_id, call.tool, result.result);
+              (_f = events.onToolResult) == null ? void 0 : _f.call(events, call.call_id, call.tool, result.result);
               results.push(result);
             }
             if (socket.readyState === WebSocket.OPEN) {
@@ -2829,9 +3020,9 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
           }
           if (msgType === "agent_done") {
             finalReply = data.final_reply || "";
-            (_f = events.onAssistantDone) == null ? void 0 : _f.call(events, finalReply);
+            (_g = events.onAssistantDone) == null ? void 0 : _g.call(events, finalReply);
             if (typeof data.session_id === "string") {
-              (_g = events.onSessionId) == null ? void 0 : _g.call(events, data.session_id);
+              (_h = events.onSessionId) == null ? void 0 : _h.call(events, data.session_id);
             }
             finish(finalReply);
             return;
@@ -2839,7 +3030,7 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
           if (msgType === "error") {
             const errMsg = data.error || "Agent error";
             errorEmitted = true;
-            (_h = events.onError) == null ? void 0 : _h.call(events, errMsg);
+            (_i = events.onError) == null ? void 0 : _i.call(events, errMsg);
             fail(new Error(errMsg));
             return;
           }
@@ -2853,7 +3044,12 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
         socket.onclose = () => finish(finalReply);
       });
       activeSocket = null;
-      if (isStopped || !assistantText) return resolve();
+      if (isStopped || !assistantText) {
+        if (!isStopped) {
+          hideAgentCursor();
+        }
+        return resolve();
+      }
       (_a = events.onAudioStateChange) == null ? void 0 : _a.call(events, "rendering");
       let ttsResult;
       try {
@@ -2890,6 +3086,9 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
       } else {
         (_b = events.onAudioStateChange) == null ? void 0 : _b.call(events, "done");
       }
+      if (!isStopped) {
+        hideAgentCursor();
+      }
       resolve();
     } catch (err) {
       if (!errorEmitted) {
@@ -2909,6 +3108,7 @@ const agentTextChatStream = (baseUrl, userText, projectId, sessionId, config, ev
   return {
     stop: () => {
       isStopped = true;
+      stopActiveAudioPlayback();
       if (activeReader) activeReader.cancel().catch(() => {
       });
       if (activeSocket && activeSocket.readyState <= WebSocket.OPEN) {
@@ -2954,6 +3154,7 @@ Sayfa bağlamı: ${pageContext}`
         activeSocket = socket;
         let finalReply = "";
         let resolved = false;
+        let accumulatedDelta = "";
         const finish = (reply) => {
           if (resolved) return;
           resolved = true;
@@ -2978,7 +3179,7 @@ Sayfa bağlamı: ${pageContext}`
           }));
         };
         socket.onmessage = async (event) => {
-          var _a2, _b2, _c2, _d, _e, _f, _g, _h;
+          var _a2, _b2, _c2, _d, _e, _f, _g, _h, _i;
           let data;
           try {
             data = JSON.parse(String(event.data));
@@ -3000,12 +3201,17 @@ Sayfa bağlamı: ${pageContext}`
             return;
           }
           if (msgType === "reply_delta" && typeof data.delta === "string") {
+            accumulatedDelta += data.delta;
             (_c2 = events.onAssistantDelta) == null ? void 0 : _c2.call(events, data.delta);
             return;
           }
           if (msgType === "tool_calls" && Array.isArray(data.calls)) {
             const calls = data.calls;
-            (_d = events.onToolCalls) == null ? void 0 : _d.call(events, calls);
+            if (accumulatedDelta.trim()) {
+              (_d = events.onIntermediateReply) == null ? void 0 : _d.call(events, accumulatedDelta.trim());
+            }
+            accumulatedDelta = "";
+            (_e = events.onToolCalls) == null ? void 0 : _e.call(events, calls);
             const results = [];
             for (const call of calls) {
               const isNavigate = call.tool === "navigate";
@@ -3028,7 +3234,7 @@ Sayfa bağlamı: ${pageContext}`
               if (isNavigate) {
                 clearPendingAgentResume();
               }
-              (_e = events.onToolResult) == null ? void 0 : _e.call(events, call.call_id, call.tool, result.result);
+              (_f = events.onToolResult) == null ? void 0 : _f.call(events, call.call_id, call.tool, result.result);
               results.push(result);
             }
             if (socket.readyState === WebSocket.OPEN) {
@@ -3038,9 +3244,9 @@ Sayfa bağlamı: ${pageContext}`
           }
           if (msgType === "agent_done") {
             finalReply = data.final_reply || "";
-            (_f = events.onAssistantDone) == null ? void 0 : _f.call(events, finalReply);
+            (_g = events.onAssistantDone) == null ? void 0 : _g.call(events, finalReply);
             if (typeof data.session_id === "string") {
-              (_g = events.onSessionId) == null ? void 0 : _g.call(events, data.session_id);
+              (_h = events.onSessionId) == null ? void 0 : _h.call(events, data.session_id);
             }
             finish(finalReply);
             return;
@@ -3048,7 +3254,7 @@ Sayfa bağlamı: ${pageContext}`
           if (msgType === "error") {
             const errMsg = data.error || "Agent error";
             errorEmitted = true;
-            (_h = events.onError) == null ? void 0 : _h.call(events, errMsg);
+            (_i = events.onError) == null ? void 0 : _i.call(events, errMsg);
             fail(new Error(errMsg));
             return;
           }
@@ -3062,7 +3268,12 @@ Sayfa bağlamı: ${pageContext}`
         socket.onclose = () => finish(finalReply);
       });
       activeSocket = null;
-      if (isStopped || !assistantText) return resolve();
+      if (isStopped || !assistantText) {
+        if (!isStopped) {
+          hideAgentCursor();
+        }
+        return resolve();
+      }
       console.info(`[Bulut] TTS start mode=resume voice=${resumeState.voice}`);
       (_a = events.onAudioStateChange) == null ? void 0 : _a.call(events, "rendering");
       let ttsResult;
@@ -3103,6 +3314,9 @@ Sayfa bağlamı: ${pageContext}`
       } else {
         (_b = events.onAudioStateChange) == null ? void 0 : _b.call(events, "done");
       }
+      if (!isStopped) {
+        hideAgentCursor();
+      }
       resolve();
     } catch (err) {
       if (!errorEmitted) {
@@ -3122,6 +3336,7 @@ Sayfa bağlamı: ${pageContext}`
   return {
     stop: () => {
       isStopped = true;
+      stopActiveAudioPlayback();
       if (activeReader) activeReader.cancel().catch(() => {
       });
       if (activeSocket && activeSocket.readyState <= WebSocket.OPEN) {
@@ -3214,6 +3429,9 @@ const resolveStatusText = (flags) => {
   if (flags.isBusy) return STATUS_LABELS.loading;
   return STATUS_LABELS.ready;
 };
+const hasActiveStatus = (flags, statusOverride) => Boolean(
+  statusOverride || flags.isBusy || flags.isRecording || flags.isTranscribing || flags.isThinking || flags.isRenderingAudio || flags.isPlayingAudio || flags.isRunningTools
+);
 const formatDurationMs = (durationMs) => {
   const totalSeconds = Math.max(0, Math.floor(durationMs / 1e3));
   const minutes = Math.floor(totalSeconds / 60).toString().padStart(2, "0");
@@ -3291,15 +3509,18 @@ const ChatWindow = ({
   const [isRunningTools, setIsRunningTools] = d(false);
   const [recordingDurationMs, setRecordingDurationMs] = d(0);
   const [statusOverride, setStatusOverride] = d(null);
-  const resolvedStatusText = resolveStatusText({
+  const statusFlags = {
     isBusy,
     isRecording,
     isTranscribing,
     isThinking,
+    isRenderingAudio,
     isPlayingAudio,
     isRunningTools
-  });
-  const statusText = statusOverride ?? resolvedStatusText;
+  };
+  const resolvedStatusText = resolveStatusText(statusFlags);
+  const showStatus = hasActiveStatus(statusFlags, statusOverride);
+  const statusText = showStatus ? statusOverride ?? resolvedStatusText : STATUS_LABELS.ready;
   const isBusyRef = A(isBusy);
   const isRecordingRef = A(isRecording);
   const nextMessageIdRef = A(getNextMessageId(messages));
@@ -3343,6 +3564,8 @@ const ChatWindow = ({
   const liveTranscriptionMessageIdRef = A(null);
   const liveTranscriptionTextRef = A("");
   const autoListenSuppressedRef = A(false);
+  const requestEpochRef = A(0);
+  const sttSendCuePlayedRef = A(false);
   y(() => {
     isBusyRef.current = isBusy;
   }, [isBusy]);
@@ -3366,7 +3589,7 @@ const ChatWindow = ({
       onPreviewChange((lastAssistant2 == null ? void 0 : lastAssistant2.text) ?? getGreetingText(config.agentName));
       return;
     }
-    if (isBusy || isTranscribing || isThinking || isRunningTools) {
+    if (showStatus) {
       const st = statusOverride ?? resolveStatusText({
         isBusy,
         isRecording,
@@ -3389,11 +3612,27 @@ const ChatWindow = ({
     isPlayingAudio,
     isRenderingAudio,
     statusOverride,
+    showStatus,
     messages
   ]);
   const playSfx = (name) => {
     playCue(name);
   };
+  const beginRequestEpoch = () => {
+    requestEpochRef.current += 1;
+    return requestEpochRef.current;
+  };
+  const invalidateRequestEpoch = () => {
+    requestEpochRef.current += 1;
+  };
+  const isCurrentRequestEpoch = (epoch) => requestEpochRef.current === epoch;
+  const playSttSentCueOnce = () => {
+    if (sttSendCuePlayedRef.current) {
+      return;
+    }
+    sttSendCuePlayedRef.current = true;
+    playSfx("sent");
+  };
   y(() => {
     if (typeof localStorage !== "undefined") {
       localStorage.setItem(STORAGE_KEY, JSON.stringify(messages));
@@ -3497,9 +3736,11 @@ const ChatWindow = ({
   };
   y(
     () => () => {
+      invalidateRequestEpoch();
       clearMicHoldTimeout();
       pendingStopAfterStartRef.current = false;
       stopActiveStream();
+      stopActiveAudioPlayback();
       cancelActiveSttWs();
       cleanupVAD();
       stopStreamTracks();
@@ -3529,6 +3770,7 @@ const ChatWindow = ({
         localStorage.setItem(SESSION_ID_KEY, resumeState.sessionId);
       }
     }
+    const requestEpoch = beginRequestEpoch();
     setIsBusy(true);
     setIsRunningTools(true);
     setStatusOverride(STATUS_LABELS.thinking);
@@ -3547,6 +3789,7 @@ const ChatWindow = ({
       freshPageContext,
       {
         onSessionId: (sid) => {
+          if (!isCurrentRequestEpoch(requestEpoch)) return;
           if (sid && sid !== sessionIdRef.current) {
             sessionIdRef.current = sid;
             if (typeof localStorage !== "undefined") {
@@ -3555,6 +3798,7 @@ const ChatWindow = ({
           }
         },
         onAssistantDelta: (delta) => {
+          if (!isCurrentRequestEpoch(requestEpoch)) return;
           setIsRunningTools(false);
           setIsThinking(true);
           setStatusOverride(null);
@@ -3572,6 +3816,7 @@ const ChatWindow = ({
           }
         },
         onAssistantDone: (assistantText) => {
+          if (!isCurrentRequestEpoch(requestEpoch)) return;
           playSfx("completed");
           setStatusOverride(null);
           setIsThinking(false);
@@ -3590,7 +3835,18 @@ const ChatWindow = ({
             );
           }
         },
+        onIntermediateReply: (text) => {
+          if (!isCurrentRequestEpoch(requestEpoch)) return;
+          void speakText(
+            config.backendBaseUrl,
+            text,
+            config.voice,
+            accessibilityMode,
+            (state) => handleAudioStateChange(state, requestEpoch)
+          ).catch((err) => console.warn("[Bulut] intermediate TTS failed", err));
+        },
         onToolCalls: (calls) => {
+          if (!isCurrentRequestEpoch(requestEpoch)) return;
           if (calls.length > 0) {
             playSfx("toolCall");
           }
@@ -3616,12 +3872,16 @@ const ChatWindow = ({
         onToolResult: () => {
         },
         onIteration: () => {
+          if (!isCurrentRequestEpoch(requestEpoch)) return;
           playSfx("thinking");
           setIsThinking(true);
           setStatusOverride(STATUS_LABELS.thinking);
         },
-        onAudioStateChange: handleAudioStateChange,
+        onAudioStateChange: (state) => {
+          handleAudioStateChange(state, requestEpoch);
+        },
         onError: (err) => {
+          if (!isCurrentRequestEpoch(requestEpoch)) return;
           setStatusOverride(null);
           appendMessage(`Hata: ${err}`, false);
         }
@@ -3631,6 +3891,7 @@ const ChatWindow = ({
     activeStreamControllerRef.current = controller;
     controller.done.catch(() => {
     }).finally(() => {
+      if (!isCurrentRequestEpoch(requestEpoch)) return;
       setIsBusy(false);
       setIsRunningTools(false);
       setIsThinking(false);
@@ -3684,7 +3945,10 @@ const ChatWindow = ({
     liveTranscriptionMessageIdRef.current = null;
     liveTranscriptionTextRef.current = "";
   };
-  const handleAudioStateChange = (state) => {
+  const handleAudioStateChange = (state, requestEpoch) => {
+    if (typeof requestEpoch === "number" && !isCurrentRequestEpoch(requestEpoch)) {
+      return;
+    }
     if (state === "rendering") {
       setIsRenderingAudio(true);
       setIsPlayingAudio(false);
@@ -3708,7 +3972,10 @@ const ChatWindow = ({
     setIsRenderingAudio(false);
     setIsPlayingAudio(false);
   };
-  const finalizeStreamCycle = () => {
+  const finalizeStreamCycle = (requestEpoch) => {
+    if (typeof requestEpoch === "number" && !isCurrentRequestEpoch(requestEpoch)) {
+      return;
+    }
     awaitingAssistantResponseRef.current = false;
     setStatusOverride(null);
     setIsBusy(false);
@@ -3742,6 +4009,7 @@ const ChatWindow = ({
       appendMessage("Ses kaydı metne dönüştürülemedi. Lütfen tekrar deneyin.", false);
       return;
     }
+    const requestEpoch = beginRequestEpoch();
     setIsBusy(true);
     setIsTranscribing(false);
     setIsThinking(true);
@@ -3777,6 +4045,7 @@ const ChatWindow = ({
         },
         {
           onSessionId: (sid) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             if (sid && sid !== sessionIdRef.current) {
               sessionIdRef.current = sid;
               if (typeof localStorage !== "undefined") {
@@ -3785,6 +4054,7 @@ const ChatWindow = ({
             }
           },
           onAssistantDelta: (delta) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             setIsTranscribing(false);
             setIsThinking(true);
             setIsRunningTools(false);
@@ -3806,6 +4076,7 @@ const ChatWindow = ({
             }
           },
           onAssistantDone: (assistantText) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             playSfx("completed");
             awaitingAssistantResponseRef.current = false;
             setStatusOverride(null);
@@ -3825,7 +4096,18 @@ const ChatWindow = ({
               );
             }
           },
+          onIntermediateReply: (text) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
+            void speakText(
+              config.backendBaseUrl,
+              text,
+              config.voice,
+              accessibilityMode,
+              (state) => handleAudioStateChange(state, requestEpoch)
+            ).catch((err) => console.warn("[Bulut] intermediate TTS failed", err));
+          },
           onToolCalls: (calls) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             if (calls.length > 0) {
               playSfx("toolCall");
             }
@@ -3851,12 +4133,16 @@ const ChatWindow = ({
           onToolResult: () => {
           },
           onIteration: () => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             playSfx("thinking");
             setIsThinking(true);
             setStatusOverride(STATUS_LABELS.thinking);
           },
-          onAudioStateChange: handleAudioStateChange,
+          onAudioStateChange: (state) => {
+            handleAudioStateChange(state, requestEpoch);
+          },
           onError: (err) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             awaitingAssistantResponseRef.current = false;
             setStatusOverride(null);
             appendMessage(`Hata: ${err}`, false);
@@ -3867,13 +4153,14 @@ const ChatWindow = ({
       activeStreamControllerRef.current = controller;
       await controller.done;
     } catch (error) {
+      if (!isCurrentRequestEpoch(requestEpoch)) return;
       awaitingAssistantResponseRef.current = false;
       setStatusOverride(null);
       if (error instanceof Error) {
         appendMessage(`Hata: ${error.message}`, false);
       }
     } finally {
-      finalizeStreamCycle();
+      finalizeStreamCycle(requestEpoch);
     }
   };
   const handleAudioBlob = async (blob) => {
@@ -3881,6 +4168,7 @@ const ChatWindow = ({
       appendMessage("Hata: Project ID yapılandırılmamış.", false);
       return;
     }
+    const requestEpoch = beginRequestEpoch();
     setIsBusy(true);
     setIsTranscribing(true);
     setIsThinking(false);
@@ -3919,9 +4207,11 @@ const ChatWindow = ({
         },
         {
           onSttRequestSent: () => {
-            playSfx("sent");
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
+            playSttSentCueOnce();
           },
           onTranscription: (data) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             if (data.session_id && data.session_id !== sessionIdRef.current) {
               sessionIdRef.current = data.session_id;
               if (typeof localStorage !== "undefined") {
@@ -3944,6 +4234,7 @@ const ChatWindow = ({
             setStatusOverride(STATUS_LABELS.thinking);
           },
           onSessionId: (sid) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             if (sid && sid !== sessionIdRef.current) {
               sessionIdRef.current = sid;
               if (typeof localStorage !== "undefined") {
@@ -3952,6 +4243,7 @@ const ChatWindow = ({
             }
           },
           onAssistantDelta: (delta) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             setIsTranscribing(false);
             setIsThinking(true);
             setIsRunningTools(false);
@@ -3973,6 +4265,7 @@ const ChatWindow = ({
             }
           },
           onAssistantDone: (assistantText) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             playSfx("completed");
             awaitingAssistantResponseRef.current = false;
             setStatusOverride(null);
@@ -3993,7 +4286,18 @@ const ChatWindow = ({
               );
             }
           },
+          onIntermediateReply: (text) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
+            void speakText(
+              config.backendBaseUrl,
+              text,
+              config.voice,
+              accessibilityMode,
+              (state) => handleAudioStateChange(state, requestEpoch)
+            ).catch((err) => console.warn("[Bulut] intermediate TTS failed", err));
+          },
           onToolCalls: (calls) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             if (calls.length > 0) {
               playSfx("toolCall");
             }
@@ -4019,12 +4323,16 @@ const ChatWindow = ({
           onToolResult: (_callId, _toolName, _result) => {
           },
           onIteration: (_iteration, _maxIterations) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             playSfx("thinking");
             setIsThinking(true);
             setStatusOverride(STATUS_LABELS.thinking);
           },
-          onAudioStateChange: handleAudioStateChange,
+          onAudioStateChange: (state) => {
+            handleAudioStateChange(state, requestEpoch);
+          },
           onError: (err) => {
+            if (!isCurrentRequestEpoch(requestEpoch)) return;
             awaitingAssistantResponseRef.current = false;
             setStatusOverride(null);
             appendMessage(`Hata: ${err}`, false);
@@ -4035,10 +4343,11 @@ const ChatWindow = ({
       activeStreamControllerRef.current = controller;
       await controller.done;
     } catch (error) {
+      if (!isCurrentRequestEpoch(requestEpoch)) return;
       awaitingAssistantResponseRef.current = false;
       setStatusOverride(null);
     } finally {
-      finalizeStreamCycle();
+      finalizeStreamCycle(requestEpoch);
     }
   };
   const stopRecording = (options) => {
@@ -4132,6 +4441,7 @@ const ChatWindow = ({
       recorderRef.current = recorder;
       audioChunksRef.current = [];
       clearLiveUserTranscriptionState();
+      sttSendCuePlayedRef.current = false;
       const sttMimeType = (recorder.mimeType || recorderOptions.mimeType || "audio/webm").split(";")[0].trim() || "audio/webm";
       const sttWsController = startSttWebSocketStream(
         config.backendBaseUrl,
@@ -4142,9 +4452,6 @@ const ChatWindow = ({
           mimeType: sttMimeType
         },
         {
-          onRequestSent: () => {
-            playSfx("sent");
-          },
           onSessionId: (sid) => {
             if (!sid || sid === sessionIdRef.current) {
               return;
@@ -4210,6 +4517,7 @@ const ChatWindow = ({
         setStatusOverride(STATUS_LABELS.transcribing);
         try {
           if (currentSttWs) {
+            playSttSentCueOnce();
             const sttResult = await currentSttWs.stop();
             if (sttResult.session_id && sttResult.session_id !== sessionIdRef.current) {
               sessionIdRef.current = sttResult.session_id;
@@ -4226,11 +4534,12 @@ const ChatWindow = ({
           }
         } catch (error) {
           console.warn(
-            `[Bulut] STT WS finalization failed, falling back to /chat/stt: ${error instanceof Error ? error.message : String(error)}`
+            `[Bulut] STT WS finalization failed, falling back to HTTP POST /chat/stt: ${error instanceof Error ? error.message : String(error)}`
           );
         } finally {
           clearLiveUserTranscriptionState();
         }
+        console.info("[Bulut] Using HTTP POST fallback for STT (streaming WS did not succeed)");
         setStatusOverride(STATUS_LABELS.thinking);
         await handleAudioBlob(blob);
       };
@@ -4333,9 +4642,12 @@ const ChatWindow = ({
     handleMicPointerUp(event);
   };
   const handleRestart = () => {
+    invalidateRequestEpoch();
+    sttSendCuePlayedRef.current = false;
     resetMicGesture();
     pendingStopAfterStartRef.current = false;
     stopActiveStream();
+    stopActiveAudioPlayback();
     cancelActiveSttWs();
     if (recorderRef.current && recorderRef.current.state !== "inactive") {
       stopRecording({ discard: true });
@@ -4369,7 +4681,10 @@ const ChatWindow = ({
     return () => window.clearTimeout(timer);
   }, [accessibilityMode]);
   const stopTask = () => {
+    invalidateRequestEpoch();
+    sttSendCuePlayedRef.current = false;
     stopActiveStream();
+    stopActiveAudioPlayback();
     cancelActiveSttWs();
     stopRecording({ discard: true });
     cleanupVAD();
@@ -4384,6 +4699,7 @@ const ChatWindow = ({
         void startRecording("vad");
       },
       cancelRecording: () => {
+        stopActiveAudioPlayback();
         cancelActiveSttWs();
         const recorder = recorderRef.current;
         if (recorder && recorder.state !== "inactive") {
@@ -4642,7 +4958,7 @@ const ChatWindow = ({
       message.id
     )) }) }),
     /* @__PURE__ */ u$1("div", { style: footerStyle, children: [
-      /* @__PURE__ */ u$1("div", { style: { ...statusPanelStyle, transition: "opacity 0.2s ease-out" }, children: statusText !== STATUS_LABELS.ready ? /* @__PURE__ */ u$1("span", { className: "bulut-status-dots", title: statusText, children: statusText }) : onAccessibilityToggle ? /* @__PURE__ */ u$1(
+      /* @__PURE__ */ u$1("div", { style: { ...statusPanelStyle, transition: "opacity 0.2s ease-out" }, children: showStatus ? /* @__PURE__ */ u$1("span", { className: "bulut-status-dots", title: statusText, children: statusText }) : onAccessibilityToggle ? /* @__PURE__ */ u$1(
         "div",
         {
           style: {