@opengeni/sdk 1.3.0 → 1.4.2

package/src/azure-live-transport.ts

@@ -0,0 +1,286 @@
+import { startCodexRealtimeWebrtc, type StartCodexRealtimeWebrtcOptions } from "./codex-realtime";
+import type { RealtimeControllerTransportStarter } from "./codex-realtime-controller";
+
+type Envelope = Record<string, unknown>;
+const MAX_FRAGMENT_BYTES = 64_000;
+
+/** The most recent speech wins when the user talks for a long time between delegations. */
+function takeUtf8Tail(text: string, maxBytes = MAX_FRAGMENT_BYTES): string {
+  const encoder = new TextEncoder();
+  if (encoder.encode(text).byteLength <= maxBytes) return text;
+  let tail = text.slice(-maxBytes);
+  while (encoder.encode(tail).byteLength > maxBytes) tail = tail.slice(1);
+  return tail;
+}
+
+/** Azure Live has timed transcript fragments, not provider-finalized turns. */
+export class AzureLiveDataChannel extends EventTarget {
+  private pending: {
+    role: "user" | "assistant";
+    text: string;
+    startMs: number;
+    endMs: number;
+  } | null = null;
+  private timer: ReturnType<typeof setTimeout> | null = null;
+  private ordinal = 0;
+  private readonly prefix = crypto.randomUUID();
+  private closed = false;
+  private closeSent = false;
+  private providerClosed = false;
+  private readonly delegations = new Set<string>();
+  // Azure delegations carry no content, so the user's words since the previous
+  // delegation become the delegation input.
+  private userSinceDelegation = "";
+  private draining: Promise<void> | null = null;
+  private finishDrain: ((error?: Error) => void) | null = null;
+  private readonly onOpen = () => this.dispatchEvent(new Event("open"));
+  private readonly onClose = () => {
+    this.providerClosed = true;
+    this.flush();
+    this.finishDrain?.();
+    this.dispatchEvent(new Event("close"));
+  };
+  private readonly onError = () => this.dispatchEvent(new Event("error"));
+  private readonly onMessage = (event: MessageEvent) => {
+    if (typeof event.data !== "string" || event.data.length > 1_048_576) return;
+    let value: unknown;
+    try {
+      value = JSON.parse(event.data);
+    } catch {
+      return;
+    }
+    if (!value || typeof value !== "object") return;
+    this.receive(value as Envelope);
+  };
+  constructor(
+    private readonly raw: RTCDataChannel,
+    private readonly drainTimeoutMs = 5_000,
+  ) {
+    super();
+    raw.addEventListener("open", this.onOpen);
+    raw.addEventListener("close", this.onClose);
+    raw.addEventListener("error", this.onError);
+    raw.addEventListener("message", this.onMessage);
+  }
+  get readyState(): RTCDataChannelState {
+    return this.closed || this.providerClosed
+      ? "closed"
+      : this.closeSent
+        ? "closing"
+        : this.raw.readyState;
+  }
+  asRtcDataChannel(): RTCDataChannel {
+    return this as unknown as RTCDataChannel;
+  }
+  private emit(value: Envelope): void {
+    if (!this.closed)
+      this.dispatchEvent(new MessageEvent("message", { data: JSON.stringify(value) }));
+  }
+  private flush(): void {
+    if (this.timer) clearTimeout(this.timer);
+    this.timer = null;
+    const pending = this.pending;
+    this.pending = null;
+    if (!pending?.text) return;
+    this.emit({
+      type: "transcript.segment",
+      turn: {
+        id: `${this.prefix}:${++this.ordinal}`,
+        role: pending.role,
+        transcript: pending.text,
+      },
+      start_ms: pending.startMs,
+      end_ms: pending.endMs,
+    });
+  }
+  private receive(value: Envelope): void {
+    if (this.closed) return;
+    if (
+      value.type === "session.input_transcript.delta" ||
+      value.type === "session.output_transcript.delta"
+    ) {
+      if (
+        typeof value.delta !== "string" ||
+        typeof value.start_ms !== "number" ||
+        typeof value.end_ms !== "number"
+      )
+        return;
+      const role = value.type === "session.input_transcript.delta" ? "user" : "assistant";
+      if (role === "user")
+        this.userSinceDelegation = takeUtf8Tail(this.userSinceDelegation + value.delta);
+      if (
+        this.pending &&
+        (this.pending.role !== role ||
+          new TextEncoder().encode(this.pending.text + value.delta).byteLength > MAX_FRAGMENT_BYTES)
+      )
+        this.flush();
+      if (new TextEncoder().encode(value.delta).byteLength > MAX_FRAGMENT_BYTES) {
+        this.emit({
+          type: "error",
+          error: { message: "Voice transcript fragment exceeded the supported size" },
+        });
+        return;
+      }
+      this.pending = {
+        role,
+        text: (this.pending?.text ?? "") + value.delta,
+        startMs: this.pending?.startMs ?? value.start_ms,
+        endMs: value.end_ms,
+      };
+      // Application grouping only: never present silence as a provider turn boundary.
+      if (!this.timer) this.timer = setTimeout(() => this.flush(), 1_000);
+      return;
+    }
+    if (value.type === "session.delegation.created") {
+      const delegation = value.delegation as Envelope | undefined;
+      if (delegation?.target !== "client" || typeof delegation.id !== "string") return;
+      this.delegations.add(delegation.id);
+      this.flush();
+      const spoken = this.userSinceDelegation.trim();
+      this.userSinceDelegation = "";
+      this.emit({
+        type: "delegation.created",
+        item: {
+          id: delegation.id,
+          type: "delegation",
+          target: "client",
+          content: spoken ? [{ type: "input_text", text: spoken }] : [],
+        },
+        offset_ms: value.offset_ms,
+      });
+      return;
+    }
+    if (value.type === "session.closed") {
+      this.providerClosed = true;
+      this.flush();
+      this.finishDrain?.();
+      return;
+    }
+    if (
+      value.type === "session.started" ||
+      value.type === "session.updated" ||
+      value.type === "error"
+    )
+      this.emit(value);
+  }
+  send(payload: string): void {
+    if (this.readyState !== "open") throw Error("Hosted voice channel is not open");
+    const value = JSON.parse(payload) as Envelope;
+    if (value.type !== "session.context.append" && value.type !== "delegation.context.append")
+      throw Error("Unsupported hosted voice context event");
+    const content = Array.isArray(value.content)
+      ? value.content.map((part) => (typeof part?.text === "string" ? part.text : "")).join("")
+      : "";
+    // Existing bridge chunks are <=500 UTF-8 bytes, below Live's 500-token bound.
+    this.raw.send(
+      JSON.stringify({
+        type:
+          value.channel === "speakable" ? "session.commentary.append" : "session.thinking.append",
+        content,
+        // Replayed tasks from an earlier provider session are general context.
+        delegation_id:
+          value.type === "delegation.context.append" &&
+          typeof value.delegation_item_id === "string" &&
+          this.delegations.has(value.delegation_item_id)
+            ? value.delegation_item_id
+            : null,
+      }),
+    );
+  }
+  drain(): Promise<void> {
+    if (this.draining) return this.draining;
+    if (this.providerClosed || this.raw.readyState === "closed") {
+      this.flush();
+      return Promise.resolve();
+    }
+    if (this.closed || this.raw.readyState !== "open") {
+      return Promise.reject(new Error("Hosted voice close is not confirmed"));
+    }
+    const pending = new Promise<void>((resolve, reject) => {
+      const timeout = setTimeout(() => {
+        this.flush();
+        this.finishDrain?.(
+          new Error("Hosted voice is still closing; retry to preserve final speech"),
+        );
+      }, this.drainTimeoutMs);
+      this.finishDrain = (error) => {
+        clearTimeout(timeout);
+        this.finishDrain = null;
+        if (error) reject(error);
+        else resolve();
+      };
+      if (!this.closeSent) {
+        this.closeSent = true;
+        try {
+          this.raw.send(JSON.stringify({ type: "session.close" }));
+        } catch {
+          this.finishDrain(new Error("Hosted voice close could not be sent"));
+        }
+      }
+    });
+    this.draining = pending;
+    // An uncertain close remains retryable; never resend the provider mutation.
+    void pending.then(
+      () => {
+        if (this.draining === pending) this.draining = null;
+      },
+      () => {
+        if (this.draining === pending) this.draining = null;
+      },
+    );
+    return pending;
+  }
+  close(): void {
+    if (this.closed) return;
+    if (this.timer) clearTimeout(this.timer);
+    this.timer = null;
+    this.pending = null;
+    this.finishDrain?.(new Error("Hosted voice channel was disposed before close confirmation"));
+    this.closed = true;
+    this.raw.removeEventListener("open", this.onOpen);
+    this.raw.removeEventListener("close", this.onClose);
+    this.raw.removeEventListener("error", this.onError);
+    this.raw.removeEventListener("message", this.onMessage);
+    this.dispatchEvent(new Event("close"));
+  }
+}
+
+export function createAzureLiveTransportStarter(
+  options: Pick<StartCodexRealtimeWebrtcOptions, "remoteAudio" | "createPeerConnection"> = {},
+): RealtimeControllerTransportStarter {
+  return async (input) => {
+    let channel: AzureLiveDataChannel | undefined;
+    try {
+      const transport = await startCodexRealtimeWebrtc({
+        ...input,
+        ...options,
+        activateRemoteAudio: false,
+        expectedModel: "opengeni-azure/gpt-live-1",
+        onEventsCreated(raw) {
+          channel = new AzureLiveDataChannel(raw);
+          input.onEventsCreated(channel.asRtcDataChannel());
+        },
+        negotiate: (request, requestOptions) =>
+          input.client.negotiateCodexRealtimeWebrtc(
+            input.workspaceId,
+            input.sessionId,
+            request,
+            requestOptions,
+          ),
+      });
+      const events = channel!;
+      return {
+        ...transport,
+        events: events.asRtcDataChannel(),
+        drain: () => events.drain(),
+        stop() {
+          events.close();
+          transport.stop();
+        },
+      };
+    } catch (error) {
+      channel?.close();
+      throw error;
+    }
+  };
+}

package/src/codex-realtime-controller.ts

@@ -120,8 +120,39 @@ export type CodexRealtimeControllerSnapshot = {
   reconnectAttempt: number;
   diagnostic: CodexRealtimeDiagnostic | null;
   error: string | null;
+  /**
+   * Set when Opengeni refused or ended deployment-funded voice for credits or
+   * availability. Codes match voice input: `insufficient_credits`,
+   * `allowance_exhausted`, `monthly_model_cost_limit`, plus
+   * `realtime_voice_unavailable`. Cleared on the next start.
+   */
+  refusal?: CodexRealtimeRefusal | null;
 };
 
+export type CodexRealtimeRefusal = { code: string; message: string };
+
+const REALTIME_REFUSAL_CODES = new Set([
+  "insufficient_credits",
+  "allowance_exhausted",
+  "monthly_model_cost_limit",
+  "realtime_voice_unavailable",
+]);
+
+/** A definitive credit/availability refusal from Opengeni, with its plain message. */
+export function codexRealtimeRefusal(error: unknown): CodexRealtimeRefusal | null {
+  if (!(error instanceof OpenGeniApiError) || !error.code) return null;
+  if (!REALTIME_REFUSAL_CODES.has(error.code)) return null;
+  const message = apiErrorMessage(error);
+  return {
+    code: error.code,
+    message:
+      message ??
+      (error.code === "insufficient_credits"
+        ? "Live voice needs Opengeni credits. Add credits to continue."
+        : "Live voice is unavailable right now."),
+  };
+}
+
 export type CodexRealtimeControllerClient = {
   beginSessionRealtime(
     workspaceId: string,
@@ -178,6 +209,7 @@ export type CodexRealtimeOwnerStorage = Pick<Storage, "getItem" | "setItem" | "r
 
 /** Canonical browser-owner storage namespace for a public realtime model. */
 export function sessionRealtimeOwnerStorageNamespace(model: SessionRealtimeModel): string {
+  if (model === "opengeni-azure/gpt-live-1") return "azure-live-owner";
   if (model === "gpt-live-1-boulder-alpha") return "codex-realtime-owner";
   if (model === "supergrok/grok-voice-think-fast-2.0") return "xai-realtime-owner";
   return "gateway-realtime-owner";
@@ -355,6 +387,7 @@ export function createCodexRealtimeController(
     reconnectAttempt: 0,
     diagnostic: null,
     error: null,
+    refusal: null,
   };
   let active: ConnectionRuntime | null = null;
   let pendingAbort: AbortController | null = null;
@@ -371,6 +404,9 @@ export function createCodexRealtimeController(
   let generation = 0;
   let reconnectAttempt = 0;
   let recoveryTerminal = false;
+  // Whether this mode ever reached a live provider connection. A definitive
+  // failure before that ends the mode instead of leaving an empty call open.
+  let connectedInMode = false;
   let mutationTail = Promise.resolve();
   let connectionTask: Promise<void> | null = null;
   const acceptedDelegationItemIds = new Set(
@@ -699,6 +735,12 @@ export function createCodexRealtimeController(
     if (closed || stopping || isAbortError(error)) return;
     const message = safeError(error);
     if (error instanceof CodexRealtimeMicrophoneError) {
+      if (error.code !== "track_ended" && !connectedInMode) {
+        // No conversation exists yet: end the call and say how to fix the
+        // microphone instead of holding an empty "reconnecting" call open.
+        await endAfterFailure(microphoneFailureMessage(error), null);
+        return;
+      }
       if (error.code === "track_ended" && state.mode?.state === "active") {
         reconnectAttempt += 1;
         publish({
@@ -720,6 +762,11 @@ export function createCodexRealtimeController(
       transitionEnded("Realtime owner no longer exists");
       return;
     }
+    const refusal = codexRealtimeRefusal(error);
+    if (refusal) {
+      await endForRefusal(refusal);
+      return;
+    }
     if (error instanceof OpenGeniApiError && error.status === 409 && error.retryable && owner) {
       try {
         const reconciled = await begin(owner, true, false);
@@ -737,6 +784,9 @@ export function createCodexRealtimeController(
           return;
         }
       }
+    } else if (error instanceof OpenGeniApiError && !error.retryable && !connectedInMode) {
+      await endAfterFailure(message, null);
+      return;
     } else if (error instanceof OpenGeniApiError && !error.retryable) {
       stopTimers();
       if (!active) releaseMicrophone();
@@ -883,6 +933,15 @@ export function createCodexRealtimeController(
     });
     let connected: CodexRealtimeWebrtcSession;
     try {
+      // Persist a fragment provider's tail before the replacement broker reads
+      // its startup history, while the old connection still owns the ledger.
+      if (active?.transport.drain) {
+        await active.transport.drain();
+        await active.bridge.flush();
+      }
+      if (closed || stopping || abort.signal.aborted || pendingGeneration !== targetGeneration) {
+        throw abort.signal.reason ?? new DOMException("Aborted", "AbortError");
+      }
       const media = await ensureMicrophone(replaceMicrophone, abort.signal);
       const operationId = randomUUID();
       const commonTransportInput = {
@@ -1016,6 +1075,7 @@ export function createCodexRealtimeController(
         onFatal: (fatal) => onBridgeFatal(targetGeneration, bridge, fatal),
       });
       active = { generation: targetGeneration, transport: connected, bridge };
+      connectedInMode = true;
       connected.setOutputMuted(state.outputMuted);
       recoveryTerminal = false;
       pendingAbort = null;
@@ -1098,6 +1158,7 @@ export function createCodexRealtimeController(
   };
 
   const heartbeat = async (): Promise<void> => {
+    let stopInstruction: CodexRealtimeRefusal | null = null;
     await exclusive(async () => {
       const current = state.mode;
       if (!owner || !current || current.state !== "active") {
@@ -1121,6 +1182,43 @@ export function createCodexRealtimeController(
         return;
       }
       publish({ mode: result.mode, realtimeId: result.mode.id, error: null });
+      stopInstruction = result.stop ?? null;
+    });
+    // The server stopped extending the lease (for example, out of credits):
+    // drain and end gracefully inside the remaining lease.
+    const instruction = stopInstruction as CodexRealtimeRefusal | null;
+    stopInstruction = null;
+    if (instruction && !stopping && !closed) await endForRefusal(instruction);
+  };
+
+  /**
+   * End the call because Opengeni refused it, keep final speech, and leave a
+   * terminal, non-retrying state that explains why.
+   */
+  const endForRefusal = async (refusal: CodexRealtimeRefusal): Promise<void> =>
+    await endAfterFailure(refusal.message, refusal);
+
+  const endAfterFailure = async (
+    message: string,
+    refusal: CodexRealtimeRefusal | null,
+  ): Promise<void> => {
+    try {
+      await controller.stop();
+    } catch {
+      // The lease is no longer extended; a failed graceful end lapses on its own.
+      stopping = false;
+      closeBrowserResources();
+      clearOwner();
+    }
+    recoveryTerminal = false;
+    publish({
+      status: "error",
+      realtimeId: null,
+      mode: null,
+      bridge: null,
+      diagnostic: diagnostic("terminal_stop", message, false),
+      error: message,
+      refusal,
     });
   };
 
@@ -1176,6 +1274,7 @@ export function createCodexRealtimeController(
       closed = false;
       stopping = false;
       recoveryTerminal = false;
+      connectedInMode = false;
       const record: OwnerRecord = {
         version: OWNER_RECORD_VERSION,
         workspaceId: options.workspaceId,
@@ -1186,11 +1285,12 @@ export function createCodexRealtimeController(
       };
       owner = record;
       storage?.setItem(storageKey, JSON.stringify(record));
-      publish({ status: "starting", error: null, diagnostic: null });
+      publish({ status: "starting", error: null, diagnostic: null, refusal: null });
       try {
         await begin(record, false);
       } catch (error) {
         closeBrowserResources(false);
+        const refusal = codexRealtimeRefusal(error);
         const failedMode = state.mode as SessionRealtimeMode | null;
         if (failedMode?.state === "active") {
           await handleConnectionFailure(error, "reconnect", false);
@@ -1200,9 +1300,13 @@ export function createCodexRealtimeController(
             status: "error",
             realtimeId: null,
             mode: null,
-            error: safeError(error),
+            error: refusal?.message ?? safeError(error),
+            ...(refusal
+              ? { refusal, diagnostic: diagnostic("terminal_stop", refusal.message, false) }
+              : {}),
           });
         }
+        if (refusal) return;
         throw error;
       }
     },
@@ -1211,6 +1315,8 @@ export function createCodexRealtimeController(
         if (recoveryTerminal && state.mode?.state === "active") return;
         const record = readOwnerRecord(storage, storageKey, options);
         if (!record) {
+          // Keep a terminal failure readable after the server records the end.
+          if (state.status === "error" && !state.mode) return;
           transitionEnded();
           return;
         }
@@ -1234,7 +1340,10 @@ export function createCodexRealtimeController(
       if (lifecycle.state === "ended") {
         const record = readOwnerRecord(storage, storageKey, options);
         if (record && record.operationId !== lifecycle.operationId) {
-          if (connectionTask || state.status === "active") return;
+          // An earlier call's end is not news while this browser is starting
+          // its own call (for example, waiting on the microphone prompt):
+          // re-beginning would advance the lease and fail the pending start.
+          if (connectionTask || state.status === "active" || state.status === "starting") return;
           closed = false;
           stopping = false;
           owner = record;
@@ -1250,6 +1359,9 @@ export function createCodexRealtimeController(
           await connectionTask;
           return;
         }
+        // The server's end of a call that failed to start must not wipe the
+        // failure the user still needs to read.
+        if (state.status === "error" && !state.mode) return;
         if (state.realtimeId === null || lifecycle.realtimeId === state.realtimeId) {
           transitionEnded(`Realtime ended: ${lifecycle.reason}`);
         }
@@ -1337,6 +1449,7 @@ export function createCodexRealtimeController(
       connectionTask = null;
       void retiredTask?.catch(() => undefined);
       try {
+        await active?.transport.drain?.();
         await active?.bridge.sealAndFlush();
         closeBrowserResources();
         let current = state.mode;
@@ -1708,6 +1821,37 @@ function isAbortError(error: unknown): boolean {
   return error instanceof Error && error.name === "AbortError";
 }
 
+function microphoneFailureMessage(error: CodexRealtimeMicrophoneError): string {
+  switch (error.code) {
+    case "permission_denied":
+      return "Microphone access is blocked. Allow it in site settings, then try again.";
+    case "device_not_found":
+      return "No microphone was found. Connect one, then try again.";
+    case "device_unavailable":
+      return "Your microphone is busy or unavailable. Close other apps using it, then try again.";
+    default:
+      return /timed out/i.test(error.message)
+        ? "Microphone access wasn't granted in time. Allow it when your browser asks, then try again."
+        : "The microphone could not start. Try again.";
+  }
+}
+
+/** The server's own message, without the transport prefix or reference id. */
+function apiErrorMessage(error: OpenGeniApiError): string | null {
+  try {
+    const body = JSON.parse(error.body) as Record<string, unknown>;
+    const nested =
+      body.error && typeof body.error === "object" ? (body.error as Record<string, unknown>) : body;
+    return typeof nested.message === "string" && nested.message ? nested.message : null;
+  } catch {
+    return null;
+  }
+}
+
 function safeError(error: unknown): string {
-  return error instanceof Error ? error.message : "Codex realtime browser controller failed";
+  if (error instanceof OpenGeniApiError) {
+    const message = apiErrorMessage(error);
+    if (message) return /[.!?]$/.test(message) ? message : `${message}.`;
+  }
+  return error instanceof Error ? error.message : "Live voice failed in this browser.";
 }

package/src/codex-realtime-v3-wire.ts

@@ -3,7 +3,8 @@
  *
  * These shapes are pinned to openai/codex@fa1d4c40. V3 delegates work with
  * `delegation.created`; it does not expose the ordinary Realtime V2
- * `function_call` protocol.
+ * `function_call` protocol. `transcript.segment` is an application adapter
+ * extension for providers that emit fragments without semantic turn boundaries.
  */
 
 export const CODEX_REALTIME_INITIAL_ITEMS_MAX_COUNT = 128;
@@ -40,7 +41,7 @@ export type CodexRealtimeV3Event =
       endMs: number | null;
     })
   | (ProviderEventIdentity & {
-      type: "turn.done";
+      type: "turn.done" | "transcript.segment";
       turnId: string;
       role: "user" | "assistant";
       transcript: string;
@@ -173,7 +174,7 @@ export function parseCodexRealtimeV3Event(payload: string): CodexRealtimeV3Parse
     };
   }
 
-  if (type === "turn.done") {
+  if (type === "turn.done" || type === "transcript.segment") {
     const turn = record(event.turn);
     const turnIdField = boundedStringField(turn, "id", CODEX_REALTIME_V3_MAX_IDENTIFIER_BYTES);
     const transcriptField = boundedStringField(

package/src/codex-realtime-v3.ts

@@ -324,6 +324,12 @@ export function createCodexRealtimeV3Bridge(
           clientAckThroughSequence = Math.max(clientAckThroughSequence ?? 0, entry.sequence);
         }
       }
+      // Persist final inbound fragments even after the provider starts closing.
+      // Pending outbound context remains replayable on the next connection.
+      if (options.events.readyState !== "open") {
+        publish();
+        continue;
+      }
       for (let index = 0; index < result.outbound.length;) {
         const entry = result.outbound[index]!;
         if (sentSequences.has(entry.sequence)) {
@@ -497,7 +503,7 @@ export function createCodexRealtimeV3Bridge(
       }
     } else if (event.type === "output_audio.delta") {
       speaking = true;
-    } else if (event.type === "turn.done") {
+    } else if (event.type === "turn.done" || event.type === "transcript.segment") {
       speaking = false;
       if (event.transcript.length > 0 && !finalizedTurnIds.has(event.turnId)) {
         const coveredByDelegationItemId =
@@ -570,7 +576,7 @@ export function createCodexRealtimeV3Bridge(
 
 function finalTranscript(
   randomUUID: () => string,
-  event: Extract<CodexRealtimeV3Event, { type: "turn.done" }>,
+  event: Extract<CodexRealtimeV3Event, { type: "turn.done" | "transcript.segment" }>,
   coveredByDelegationItemId: string | null,
   modelContext: string | undefined,
 ): SessionRealtimeInboundEntry {
@@ -581,6 +587,9 @@ function finalTranscript(
     text: event.transcript,
     payload: {
       turnId: event.turnId,
+      ...(event.type === "transcript.segment"
+        ? { transcriptSource: "provider_fragments", grouping: "application_segment" }
+        : {}),
       ...(coveredByDelegationItemId ? { coveredByDelegationItemId } : {}),
     },
     ...(modelContext ? { modelContext } : {}),

package/src/codex-realtime.ts

@@ -37,6 +37,8 @@ export type AcquireCodexRealtimeMicrophoneOptions = {
 
 export type StartCodexRealtimeWebrtcOptions = {
   negotiate: CodexRealtimeNegotiator;
+  /** Exact configured provider model expected from the server broker. */
+  expectedModel?: CodexRealtimeWebrtcResponse["model"];
   realtimeId: string;
   operationId: string;
   browserInstanceId: string;
@@ -77,6 +79,8 @@ export type CodexRealtimeWebrtcSession = {
   setOutputMuted(muted: boolean): void;
   activateRemoteAudio(): void;
   retryAudibleOutput(): Promise<boolean>;
+  /** Gracefully finish provider output before sealing the durable ledger. */
+  drain?(): Promise<void>;
   /** Idempotently close media, data channel, and peer transport. */
   stop(): void;
 };
@@ -260,7 +264,11 @@ export async function startCodexRealtimeWebrtc(
       { signal: options.signal },
     );
     throwIfAborted(options.signal);
-    if (answer.version !== "v3" || answer.model !== "gpt-live-1-boulder-alpha" || !answer.sdp) {
+    if (
+      answer.version !== "v3" ||
+      answer.model !== (options.expectedModel ?? "gpt-live-1-boulder-alpha") ||
+      !answer.sdp
+    ) {
       throw new Error("Opengeni returned an incompatible Codex realtime answer");
     }
     await peerConnection.setRemoteDescription({

package/src/index.ts

@@ -356,9 +356,11 @@ export type {
   CodexRealtimeControllerStatus,
   CodexRealtimeMicrophoneState,
   CodexRealtimeOwnerStorage,
+  CodexRealtimeRefusal,
   CreateCodexRealtimeControllerOptions,
   RealtimeControllerTransportStarter,
 } from "./codex-realtime-controller";
+export { codexRealtimeRefusal } from "./codex-realtime-controller";
 export {
   createGatewayRealtimeTransportStarter,
   createXaiSubscriptionRealtimeTransportStarter,
@@ -749,6 +751,7 @@ export type {
   SessionRealtimeMode,
   SessionRealtimeModel,
   SessionRealtimeMutationResponse,
+  SessionRealtimeStopInstruction,
   SessionRealtimeState,
   WorkspaceModelCatalogModel,
   WorkspaceModelCatalogResponse,