@runtypelabs/persona 4.22.0 → 4.22.1

package/src/types.ts

@@ -991,15 +991,15 @@ export type AgentMessageMetadata = {
    */
   parentStepId?: string;
   /**
-   * Set to `true` on a tool-variant message produced from a `step_await`
+   * Set to `true` on a tool-variant message produced from an `await`
    * event (`awaitReason: "local_tool_required"`). Signals to UI code that
    * the tool call is a LOCAL tool and the server is paused waiting for a
    * `POST /v1/dispatch/resume` with the user's answer keyed by tool name.
    */
   awaitingLocalTool?: boolean;
   /**
-   * The provider per-call id (`toolu_…`) carried on the `step_await` /
-   * `flow_await` events for a LOCAL tool (core#3878). Present only when the
+   * The provider per-call id (`toolu_…`) carried on the `await`
+   * event for a LOCAL tool (core#3878). Present only when the
    * server emits it. Two PARALLEL calls to the same tool in one turn share a
    * `toolName` (and a collapsed `toolId`) but get DISTINCT `webMcpToolCallId`s,
    * so this is the key the widget batches a single `/resume` on: preferred
@@ -3418,7 +3418,7 @@ export type EventStreamConfig = {
   /**
    * Custom badge color mappings by event type prefix or exact type.
    * Keys are matched as exact match first, then prefix match (keys ending with "_").
-   * @example { "flow_": { bg: "#dcfce7", text: "#166534" }, "error": { bg: "#fecaca", text: "#991b1b" } }
+   * @example { "execution_": { bg: "#dcfce7", text: "#166534" }, "error": { bg: "#fecaca", text: "#991b1b" } }
    */
   badgeColors?: Record<string, EventStreamBadgeColor>;
   /**
@@ -4983,10 +4983,12 @@ export interface VoiceProvider {
    * `isFinal=false` is a live interim update (user partials, or assistant deltas
    * on providers that stream them); `isFinal=true` finalizes that role's text.
    * On the realtime `runtype` path, interim updates fire for the `user` only and
-   * the `assistant` arrives as a single final.
+   * the `assistant` arrives as a single final. Providers with overlapping turns
+   * must include the same turnId on user and assistant finals; without IDs,
+   * providers must discard cancelled output before emitting the next user final.
    */
   onTranscript?(
-    callback: (role: 'user' | 'assistant', text: string, isFinal: boolean) => void,
+    callback: (role: 'user' | 'assistant', text: string, isFinal: boolean, metadata?: { turnId?: string }) => void,
   ): void;
 
   /** Register a callback for per-turn latency metrics (realtime path). */
@@ -7561,16 +7563,16 @@ export type AgentWidgetConfig = {
    *
    * @example
    * ```typescript
-   * // For Runtype API format
+   * // For a custom backend (Runtype unified events are handled natively)
    * config: {
    *   parseSSEEvent: (data) => {
-   *     if ((data.type === 'step_delta' || data.type === 'step_chunk') && (data.delta || data.chunk)) {
-   *       return { text: data.delta ?? data.chunk };
+   *     if (data.type === 'token' && typeof data.value === 'string') {
+   *       return { text: data.value };
    *     }
-   *     if (data.type === 'flow_complete') {
+   *     if (data.type === 'done') {
    *       return { done: true };
    *     }
-   *     if (data.type === 'step_error') {
+   *     if (data.type === 'failure') {
    *       return { error: data.error };
    *     }
    *     return null; // Ignore other events
@@ -7787,7 +7789,7 @@ export type AgentWidgetReasoning = {
   /**
    * Reasoning channel scope (wire spec). `"turn"` is ordinary per-turn
    * thinking; `"loop"` is a cross-iteration agent reflection (the fold that
-   * replaced the legacy `agent_reflection` event). Absent for legacy streams.
+   * replaced the legacy `agent_reflection` event). Optional when the stream does not specify a scope.
    */
   scope?: "turn" | "loop";
   startedAt?: number;
@@ -7802,6 +7804,9 @@ export type AgentWidgetToolCall = {
   args?: unknown;
   chunks?: string[];
   result?: unknown;
+  /** Whether the tool succeeded; failure does not necessarily end the execution. */
+  success?: boolean;
+  error?: string;
   duration?: number;
   startedAt?: number;
   completedAt?: number;
@@ -7835,7 +7840,7 @@ export type AgentWidgetMessageVariant = "assistant" | "reasoning" | "tool" | "ap
 
 /**
  * Per-turn / per-step stop reason emitted by the runtime on
- * `agent_turn_complete` and `step_complete` SSE events. The vocabulary is
+ * `turn_complete` and `step_complete` SSE events. The vocabulary is
  * owned by the upstream Runtype API: do not extend without coordination.
  *
  * - `end_turn`: natural completion (no affordance needed)
@@ -7967,7 +7972,7 @@ export type AgentWidgetMessage = {
    */
   agentMetadata?: AgentMessageMetadata;
   /**
-   * Per-turn stop reason reported by the runtime on `agent_turn_complete`
+   * Per-turn stop reason reported by the runtime on `turn_complete`
    * (agent-loop path) or the last `step_complete` for a prompt step
    * (dispatch / flow path). Absent when the API did not report a value.
    *

package/src/ui.ask-user-question-plugin.test.ts

@@ -174,7 +174,7 @@ describe("renderAskUserQuestion plugin hook", () => {
       ok: true,
       body: new ReadableStream({
         start(c) {
-          c.enqueue(new TextEncoder().encode('data: {"type":"flow_complete","success":true}\n\n'));
+          c.enqueue(new TextEncoder().encode('data: {"type":"execution_complete","success":true}\n\n'));
           c.close();
         },
       }),
@@ -217,7 +217,7 @@ describe("renderAskUserQuestion plugin hook", () => {
       ok: true,
       body: new ReadableStream({
         start(c) {
-          c.enqueue(new TextEncoder().encode('data: {"type":"flow_complete","success":true}\n\n'));
+          c.enqueue(new TextEncoder().encode('data: {"type":"execution_complete","success":true}\n\n'));
           c.close();
         },
       }),
@@ -284,7 +284,7 @@ describe("renderAskUserQuestion plugin hook", () => {
       ok: true,
       body: new ReadableStream({
         start(c) {
-          c.enqueue(new TextEncoder().encode('data: {"type":"flow_complete","success":true}\n\n'));
+          c.enqueue(new TextEncoder().encode('data: {"type":"execution_complete","success":true}\n\n'));
           c.close();
         },
       }),
@@ -366,7 +366,7 @@ describe("renderAskUserQuestion plugin hook", () => {
       ok: true,
       body: new ReadableStream({
         start(c) {
-          c.enqueue(new TextEncoder().encode('data: {"type":"flow_complete","success":true}\n\n'));
+          c.enqueue(new TextEncoder().encode('data: {"type":"execution_complete","success":true}\n\n'));
           c.close();
         },
       }),
@@ -434,7 +434,7 @@ describe("renderAskUserQuestion plugin hook", () => {
       ok: true,
       body: new ReadableStream({
         start(c) {
-          c.enqueue(new TextEncoder().encode('data: {"type":"flow_complete","success":true}\n\n'));
+          c.enqueue(new TextEncoder().encode('data: {"type":"execution_complete","success":true}\n\n'));
           c.close();
         },
       }),
@@ -631,7 +631,7 @@ describe("renderAskUserQuestion plugin hook", () => {
       ok: true,
       body: new ReadableStream({
         start(c) {
-          c.enqueue(new TextEncoder().encode('data: {"type":"flow_complete","success":true}\n\n'));
+          c.enqueue(new TextEncoder().encode('data: {"type":"execution_complete","success":true}\n\n'));
           c.close();
         },
       }),
@@ -645,7 +645,7 @@ describe("renderAskUserQuestion plugin hook", () => {
       ok: true,
       body: new ReadableStream({
         start(c) {
-          c.enqueue(new TextEncoder().encode('data: {"type":"flow_complete","success":true}\n\n'));
+          c.enqueue(new TextEncoder().encode('data: {"type":"execution_complete","success":true}\n\n'));
           c.close();
         },
       }),

package/src/ui.composer-defer-one.test.ts

@@ -3,42 +3,14 @@
 import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 
 import { createAgentExperience } from "./ui";
-import { createUnifiedEventWrite } from "./utils/__fixtures__/unified-translator.oracle";
-
-const legacyEvent = (type: string, data: Record<string, unknown>) =>
-  `event: ${type}\ndata: ${JSON.stringify({ type, ...data })}\n\n`;
+import { buildAssistantTurnFrames } from "./testing/mock-stream";
 
 /** One complete assistant turn, in the same wire shape the widget expects. */
-const unifiedFrames = (): string => {
-  let out = "";
-  const write = createUnifiedEventWrite((chunk) => {
-    out += chunk;
-  });
-  write(legacyEvent("flow_start", { flowId: "flow_1", flowName: "T", totalSteps: 1 }));
-  write(
-    legacyEvent("step_start", {
-      id: "step_1",
-      name: "Prompt",
-      stepType: "prompt",
-      index: 0,
-      totalSteps: 1,
-    })
-  );
-  write(legacyEvent("text_start", { messageId: "message_1" }));
-  write(legacyEvent("step_delta", { id: "step_1", text: "answer" }));
-  write(legacyEvent("text_end", { messageId: "message_1" }));
-  write(
-    legacyEvent("step_complete", {
-      id: "step_1",
-      name: "Prompt",
-      stepType: "prompt",
-      success: true,
-      result: { response: "answer" },
-    })
-  );
-  write(legacyEvent("flow_complete", { flowId: "flow_1", success: true }));
-  return out;
-};
+const unifiedFrames = (): string => [
+  { type: "execution_start", executionId: "exec-1", kind: "agent" },
+  ...buildAssistantTurnFrames({ executionId: "exec-1", text: "answer" }),
+  { type: "execution_complete", executionId: "exec-1", kind: "agent", success: true },
+].map((frame) => `data: ${JSON.stringify(frame)}\n\n`).join("");
 
 const mounts: HTMLElement[] = [];
 const controllers: ReturnType<typeof createAgentExperience>[] = [];

package/src/ui.composer-draft-persistence.test.ts

@@ -17,7 +17,7 @@ const streamResponse = () => {
     body: new ReadableStream({
       start(controller) {
         controller.enqueue(
-          encoder.encode('data: {"type":"flow_complete","success":true}\n\n')
+          encoder.encode('data: {"type":"execution_complete","success":true}\n\n')
         );
         controller.close();
       },

package/src/ui.composer-quote.test.ts

@@ -17,7 +17,7 @@ const streamResponse = () => {
     body: new ReadableStream({
       start(controller) {
         controller.enqueue(
-          encoder.encode('data: {"type":"flow_complete","success":true}\n\n')
+          encoder.encode('data: {"type":"execution_complete","success":true}\n\n')
         );
         controller.close();
       },

package/src/ui.message-history-actions.test.ts

@@ -17,7 +17,7 @@ const streamResponse = () => {
     body: new ReadableStream({
       start(controller) {
         controller.enqueue(
-          encoder.encode('data: {"type":"flow_complete","success":true}\n\n')
+          encoder.encode('data: {"type":"execution_complete","success":true}\n\n')
         );
         controller.close();
       },

package/src/ui.streaming-coalescing.test.ts

@@ -4,7 +4,6 @@ import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 
 import { createAgentExperience } from "./ui";
 import type { AgentWidgetStoredState } from "./types";
-import { createUnifiedEventWrite } from "./utils/__fixtures__/unified-translator.oracle";
 
 // The approval UI ships in a lazy chunk; provide it eagerly so this file's
 // synchronous render assertions hold. Per-file module isolation keeps the
@@ -48,7 +47,7 @@ const flushMicrotasks = async (times = 20) => {
   }
 };
 
-const legacyEvent = (type: string, data: Record<string, unknown>) =>
+const sseEvent = (type: string, data: Record<string, unknown>) =>
   `event: ${type}\ndata: ${JSON.stringify({ type, ...data })}\n\n`;
 
 const createStreamHarness = () => {
@@ -59,9 +58,7 @@ const createStreamHarness = () => {
     new Response(new ReadableStream<Uint8Array>({
       start(streamController) {
         controller = streamController;
-        write = createUnifiedEventWrite((chunk) => {
-          streamController.enqueue(encoder.encode(chunk));
-        });
+        write = (chunk) => streamController.enqueue(encoder.encode(chunk));
       },
     }))
   );
@@ -69,7 +66,7 @@ const createStreamHarness = () => {
     fetchMock,
     send(type: string, data: Record<string, unknown>) {
       if (!write) throw new Error("stream not started");
-      write(legacyEvent(type, data));
+      write(sseEvent(type, data));
     },
     close() {
       controller?.close();
@@ -85,7 +82,7 @@ const startTurn = async (mount: HTMLElement) => {
 };
 
 const startText = (stream: ReturnType<typeof createStreamHarness>, deltas: string[]) => {
-  stream.send("flow_start", { flowId: "flow_1", flowName: "Test", totalSteps: 1 });
+  stream.send("execution_start", { kind: "flow", executionId: "exec-1", flowId: "flow_1", flowName: "Test", totalSteps: 1 });
   stream.send("step_start", {
     id: "step_1",
     name: "Prompt",
@@ -93,12 +90,12 @@ const startText = (stream: ReturnType<typeof createStreamHarness>, deltas: strin
     index: 0,
     totalSteps: 1,
   });
-  stream.send("text_start", { messageId: "message_1" });
-  deltas.forEach((text) => stream.send("step_delta", { id: "step_1", text }));
+  stream.send("text_start", { id: "message_1" });
+  deltas.forEach((text) => stream.send("text_delta", { id: "message_1", delta: text }));
 };
 
 const finishText = (stream: ReturnType<typeof createStreamHarness>) => {
-  stream.send("text_end", { messageId: "message_1" });
+  stream.send("text_complete", { id: "message_1" });
   stream.send("step_complete", {
     id: "step_1",
     name: "Prompt",
@@ -106,7 +103,7 @@ const finishText = (stream: ReturnType<typeof createStreamHarness>) => {
     success: true,
     result: { response: "Hello world" },
   });
-  stream.send("flow_complete", { flowId: "flow_1", success: true });
+  stream.send("execution_complete", { kind: "flow", executionId: "exec-1", flowId: "flow_1", success: true });
   stream.close();
 };
 
@@ -145,8 +142,8 @@ describe("pure streaming text update coalescing", () => {
     startText(stream, ["Hel"]);
     await flushMicrotasks();
     const afterFirstDelta = saves.length;
-    stream.send("step_delta", { id: "step_1", text: "lo" });
-    stream.send("step_delta", { id: "step_1", text: " world" });
+    stream.send("text_delta", { id: "message_1", delta: "lo" });
+    stream.send("text_delta", { id: "message_1", delta: " world" });
     await flushMicrotasks();
 
     expect(saves).toHaveLength(afterFirstDelta);

package/src/ui.ts

@@ -1,3 +1,4 @@
+import { usesSessionVoice } from "./utils/voice-support";
 import { escapeHtml, createMarkdownProcessorFromConfig } from "./postprocessors";
 import { resolveSanitizer } from "./utils/sanitize";
 import { stabilizeStreamingTables } from "./utils/streaming-table";
@@ -3272,7 +3273,7 @@ export const createAgentExperience = (
 
     removeAskUserQuestionSheet(askUserOverlay, toolCallId);
 
-    // Branch: LOCAL-tool pause (step_await) resumes via /resume with structured
+    // Branch: LOCAL-tool pause (await) resumes via /resume with structured
     // toolOutputs; legacy path sends as a plain user message.
     const sourceMessage = sessionRef.current
       .getMessages()
@@ -5245,7 +5246,7 @@ export const createAgentExperience = (
       setTimeout(() => {
         if (!voiceState.active) {
           voiceState.manuallyDeactivated = false;
-          if (config.voiceRecognition?.provider?.type === 'runtype') {
+          if (usesSessionVoice(config.voiceRecognition?.provider)) {
             session.toggleVoice().then(() => {
               voiceState.active = session.isVoiceActive();
               emitVoiceState("restore");
@@ -6319,20 +6320,23 @@ export const createAgentExperience = (
     // reasoning removal, no threshold variant.
     const dropCompletedToolCalls =
       config.features?.toolCallDisplay?.completedVisibility === "removed";
-
-    messages.forEach((message) => {
+    /** A finished reasoning/tool row that `completedVisibility` removes from the transcript. */
+    const isDroppedCompletedRow = (message: AgentWidgetMessage): boolean => {
       if (
         message.variant === "reasoning" &&
-        message.reasoning?.status === "complete" &&
-        dropReasoningRow(message.reasoning)
+        message.reasoning?.status === "complete"
       ) {
-        return;
+        return dropReasoningRow(message.reasoning);
       }
-      if (
+      return (
         dropCompletedToolCalls &&
         message.variant === "tool" &&
         message.toolCall?.status === "complete"
-      ) {
+      );
+    };
+
+    messages.forEach((message) => {
+      if (isDroppedCompletedRow(message)) {
         return;
       }
       activeMessageIds.add(message.id);
@@ -6408,11 +6412,11 @@ export const createAgentExperience = (
       if (cachedWrapper) {
         tempContainer.appendChild(cachedWrapper.cloneNode(true));
         // Keep the overlay sheet alive only while the server is actively
-        // waiting on the user (awaitingLocalTool === true). Before step_await
+        // waiting on the user (awaitingLocalTool === true). Before await
         // fires, or after the answer resumes the flow, omit from
         // liveAskToolIds so the prune loop below removes any stale DOM sheet.
         // Guards against lingering skeleton sheets from tool_start events
-        // that never get a matching step_await (e.g. LLM-hallucinated trailing
+        // that never get a matching await (e.g. LLM-hallucinated trailing
         // ask_user_question calls at end-of-turn).
         if (
           isAskUserQuestionMessage(message) &&
@@ -7002,16 +7006,40 @@ export const createAgentExperience = (
     // Add standalone typing indicator only if streaming but no assistant message is streaming yet
     // (This shows while waiting for the stream to start)
     // Check for ANY streaming assistant message, even if empty (to avoid duplicate bubbles)
+    //
+    // Tool and reasoning rows hidden by `features.showToolCalls: false` /
+    // `features.showReasoning: false`, or removed once complete by their
+    // `completedVisibility` setting, never render, so they must not count as
+    // "the assistant is already visibly responding": otherwise the transcript
+    // goes blank for the whole tool phase. When those rows are shown, their own
+    // bubble carries the loading animation and the standalone dots stay hidden.
+    const isHiddenActivityRow = (msg: AgentWidgetMessage): boolean =>
+      (msg.variant === "tool" && !showToolCalls) ||
+      (msg.variant === "reasoning" && !showReasoning) ||
+      isDroppedCompletedRow(msg);
     const hasStreamingAssistantMessage = messages.some(
-      (msg) => msg.role === "assistant" && msg.streaming
+      (msg) => msg.role === "assistant" && msg.streaming && !isHiddenActivityRow(msg)
     );
 
     // Also check if there's a recently completed assistant message (streaming just ended)
     // This prevents flicker when the message completes but isStreaming hasn't updated yet
     // Approval-variant messages are UI controls, not content: exclude them so the typing
-    // indicator still shows while the agent resumes after approval
-    const lastMessage = messages[messages.length - 1];
-    const hasRecentAssistantResponse = lastMessage?.role === "assistant" && !lastMessage.streaming && lastMessage.variant !== "approval";
+    // indicator still shows while the agent resumes after approval. Hidden tool/reasoning
+    // rows are skipped for the same reason (the gap between a tool finishing and the
+    // model's next text would otherwise show nothing), and a hidden row *after* the
+    // last visible message means the agent has moved on to more work, so that visible
+    // message is not the turn's final response.
+    let lastVisibleIndex = messages.length - 1;
+    while (lastVisibleIndex >= 0 && isHiddenActivityRow(messages[lastVisibleIndex])) {
+      lastVisibleIndex--;
+    }
+    const lastMessage = lastVisibleIndex >= 0 ? messages[lastVisibleIndex] : undefined;
+    const hasHiddenActivityAfterLastVisible = lastVisibleIndex < messages.length - 1;
+    const hasRecentAssistantResponse =
+      lastMessage?.role === "assistant" &&
+      !lastMessage.streaming &&
+      lastMessage.variant !== "approval" &&
+      !hasHiddenActivityAfterLastVisible;
 
     if (isStreaming && messages.some((msg) => msg.role === "user") && !hasStreamingAssistantMessage && !hasRecentAssistantResponse) {
       // Get loading indicator using priority chain: plugin -> config -> default
@@ -7925,6 +7953,10 @@ export const createAgentExperience = (
   };
 
   const isComposerInputDisabled = (): boolean => composerLock.inputDisabled;
+  const isActiveSessionVoice = (): boolean =>
+    usesSessionVoice(config.voiceRecognition?.provider) &&
+    !!session && (session.isBargeInActive() ||
+      ["listening", "processing", "speaking"].includes(session.getVoiceStatus()));
   /** Any submission path is blocked; stop is not a submission. */
   const isComposerSendBlocked = (): boolean =>
     composerLock.inputDisabled || composerLock.sendDisabled;
@@ -8033,7 +8065,7 @@ export const createAgentExperience = (
         button.disabled = locked;
       });
     }
-    if (micButton) micButton.disabled = blocked;
+    if (micButton) micButton.disabled = locked || (streaming && !isActiveSessionVoice());
 
     suggestionManagers.forEach((manager) => {
       manager.elements.forEach((element) => {
@@ -8860,7 +8892,7 @@ export const createAgentExperience = (
       // per-state UI (e.g. a listening/speaking status dock). Fires for every
       // provider; the mic-button styling below is runtype-specific.
       eventBus.emit("voice:status", { status, timestamp: Date.now() });
-      if (config.voiceRecognition?.provider?.type !== 'runtype') return;
+      if (status !== 'disconnected' && !usesSessionVoice(config.voiceRecognition?.provider)) return;
 
       switch (status) {
         case 'listening':
@@ -8983,7 +9015,7 @@ export const createAgentExperience = (
   scrollSendSeeded = true;
 
   // Setup Runtype voice provider when configured (connects WebSocket for server-side STT)
-  if (config.voiceRecognition?.provider?.type === 'runtype') {
+  if (config.voiceRecognition?.enabled === true && usesSessionVoice(config.voiceRecognition.provider)) {
     try {
       session.setupVoice();
     } catch (err) {
@@ -12622,7 +12654,7 @@ export const createAgentExperience = (
     // Dictation is composition: a locked input takes none of it.
     if (isComposerInputDisabled()) return;
     // Runtype provider: use session.toggleVoice() (WebSocket-based STT)
-    if (config.voiceRecognition?.provider?.type === 'runtype') {
+    if (usesSessionVoice(config.voiceRecognition?.provider)) {
       const voiceStatus = session.getVoiceStatus();
       const interruptionMode = session.getVoiceInterruptionMode();
 
@@ -12693,7 +12725,7 @@ export const createAgentExperience = (
     // The click listener itself is registered by `wireComposerSurface`, so a
     // composer rebuild unwires it with the rest of the surface.
     destroyCallbacks.push(() => {
-      if (config.voiceRecognition?.provider?.type === 'runtype') {
+      if (usesSessionVoice(config.voiceRecognition?.provider)) {
         if (session.isVoiceActive()) session.toggleVoice();
         removeRuntypeMicStateStyles();
       } else {
@@ -12722,7 +12754,7 @@ export const createAgentExperience = (
     }
     setTimeout(() => {
       if (!voiceState.active && !voiceState.manuallyDeactivated) {
-        if (config.voiceRecognition?.provider?.type === 'runtype') {
+        if (usesSessionVoice(config.voiceRecognition?.provider)) {
           session.toggleVoice().then(() => {
             voiceState.active = session.isVoiceActive();
             emitVoiceState("auto");
@@ -14308,9 +14340,9 @@ export const createAgentExperience = (
         typeof window !== 'undefined' &&
         (typeof (window as any).webkitSpeechRecognition !== 'undefined' ||
          typeof (window as any).SpeechRecognition !== 'undefined');
-      const hasRuntypeProvider =
-        config.voiceRecognition?.provider?.type === 'runtype';
-      const hasVoiceInput = hasSpeechRecognition || hasRuntypeProvider;
+      const hasSessionProvider =
+        usesSessionVoice(config.voiceRecognition?.provider);
+      const hasVoiceInput = hasSpeechRecognition || hasSessionProvider;
 
       if (voiceRecognitionEnabled && hasVoiceInput) {
         // Create or update mic button
@@ -14343,7 +14375,7 @@ export const createAgentExperience = (
             );
 
             // Set disabled state
-            micButton.disabled = session.isStreaming();
+            micButton.disabled = isComposerInputDisabled() || (session.isStreaming() && !isActiveSessionVoice());
           }
         } else {
           // Update existing mic button with new config
@@ -14439,14 +14471,14 @@ export const createAgentExperience = (
 
           // Show and update disabled state
           micButtonWrapper.style.display = "";
-          micButton.disabled = session.isStreaming();
+          micButton.disabled = isComposerInputDisabled() || (session.isStreaming() && !isActiveSessionVoice());
         }
       } else {
         // Hide mic button
         if (micButton && micButtonWrapper) {
           micButtonWrapper.style.display = "none";
           // Stop any active recording if disabling
-          if (config.voiceRecognition?.provider?.type === 'runtype') {
+          if (usesSessionVoice(config.voiceRecognition?.provider)) {
             if (session.isVoiceActive()) session.toggleVoice();
           } else if (isRecording) {
             stopVoiceRecognition();
@@ -15015,8 +15047,8 @@ export const createAgentExperience = (
     },
     startVoiceRecognition(): boolean {
       if (session.isStreaming()) return false;
-      if (config.voiceRecognition?.provider?.type === 'runtype') {
-        if (session.isVoiceActive()) return true;
+      if (usesSessionVoice(config.voiceRecognition?.provider)) {
+        if (session.isVoiceActive() || session.isBargeInActive()) return true;
         if (!open && isPanelToggleable()) setOpenState(true, "system");
         voiceState.manuallyDeactivated = false;
         persistVoiceMetadata();
@@ -15037,9 +15069,12 @@ export const createAgentExperience = (
       return true;
     },
     stopVoiceRecognition(): boolean {
-      if (config.voiceRecognition?.provider?.type === 'runtype') {
-        if (!session.isVoiceActive()) return false;
-        session.toggleVoice().then(() => {
+      if (usesSessionVoice(config.voiceRecognition?.provider)) {
+        if (!session.isVoiceActive() && !session.isBargeInActive()) return false;
+        const stop = session.isBargeInActive()
+          ? session.deactivateBargeIn()
+          : session.toggleVoice();
+        stop.then(() => {
           voiceState.active = false;
           voiceState.manuallyDeactivated = true;
           persistVoiceMetadata();