Standardize model names to gpt-realtime and add telemetry enablement sample (#46552)

- Update all sample code and configuration to use 'gpt-realtime' instead of 'gpt-4o-realtime-preview'
- Add sample_voicelive_with_telemetry_enablement.py demonstrating minimal code changes for OpenTelemetry integration
- Updated files: .env.template, BASIC_VOICE_ASSISTANT.md, README.md, async samples, and telemetry samples
- Reflects latest model naming and simplified telemetry integration pattern

Co-authored-by: Xiting Zhang <xitzhang@microsoft.com>

Author: xitzhang

sdk/voicelive/azure-ai-voicelive/.env.template

@@ -6,7 +6,7 @@ AZURE_VOICELIVE_API_KEY=your-voicelive-api-key
 AZURE_VOICELIVE_ENDPOINT=wss://api.voicelive.com/v1
 
 # Optional configuration
-AZURE_VOICELIVE_MODEL=gpt-4o-realtime-preview
+AZURE_VOICELIVE_MODEL=gpt-realtime
 AZURE_VOICELIVE_VOICE=alloy
 AZURE_VOICELIVE_INSTRUCTIONS=You are a helpful assistant. Keep your responses concise.
 

sdk/voicelive/azure-ai-voicelive/samples/BASIC_VOICE_ASSISTANT.md

@@ -51,7 +51,7 @@ Create a `.env` file with your credentials:
 ```bash
 AZURE_VOICELIVE_API_KEY=your-api-key
 AZURE_VOICELIVE_ENDPOINT=your-endpoint
-AZURE_VOICELIVE_MODEL=gpt-4o-realtime-preview
+AZURE_VOICELIVE_MODEL=gpt-realtime
 AZURE_VOICELIVE_VOICE=en-US-AvaNeural
 AZURE_VOICELIVE_INSTRUCTIONS=You are a helpful AI assistant. Respond naturally and conversationally.
 ```
@@ -66,7 +66,7 @@ Optional command-line arguments:
 
 ```bash
 python basic_voice_assistant_async.py \
-    --model gpt-4o-realtime-preview \
+    --model gpt-realtime \
     --voice en-US-AvaNeural \
     --instructions "You are a helpful assistant" \
     --verbose

sdk/voicelive/azure-ai-voicelive/samples/README.md

@@ -44,7 +44,7 @@ This directory contains sample applications demonstrating various capabilities o
    ```ini
    AZURE_VOICELIVE_API_KEY=your-voicelive-api-key
    AZURE_VOICELIVE_ENDPOINT=wss://api.voicelive.com/v1
-   AZURE_VOICELIVE_MODEL=gpt-4o-realtime-preview
+   AZURE_VOICELIVE_MODEL=gpt-realtime
    AZURE_VOICELIVE_VOICE=alloy
    AZURE_VOICELIVE_INSTRUCTIONS=You are a helpful assistant. Keep your responses concise.
    ```
@@ -93,7 +93,7 @@ python basic_voice_assistant_async.py
 Most samples support additional command-line arguments. For example:
 
 ```bash
-python basic_voice_assistant_async.py --model gpt-4o-realtime-preview --voice alloy
+python basic_voice_assistant_async.py --model gpt-realtime --voice alloy
 
 # With telemetry tracing
 python basic_voice_assistant_async.py --enable-tracing
@@ -117,6 +117,7 @@ These samples are in the `telemetry/` folder and demonstrate OpenTelemetry-based
 
 | Sample | Description |
 |---|---|
+| `sample_voicelive_with_telemetry_enablement.py` | Minimal telemetry setup on top of existing VoiceLive code. Shows the smallest practical code delta to emit spans to console. |
 | `sample_voicelive_with_console_tracing.py` | Basic tracing with console output. All connection, send, and receive operations produce OpenTelemetry spans printed to stdout. |
 | `sample_voicelive_with_azure_monitor_tracing.py` | Traces exported to Azure Monitor / Application Insights. View results in the "Tracing" tab. |
 | `sample_voicelive_with_console_tracing_custom_attributes.py` | Adds custom `SpanProcessor` to inject application-specific attributes (session ID, etc.) into every span. |

sdk/voicelive/azure-ai-voicelive/samples/async_function_calling_sample.py

@@ -727,7 +727,7 @@ async def main():
     client = AsyncFunctionCallingClient(
         endpoint=endpoint,
         credential=credential,
-        model="gpt-4o-realtime-preview",
+        model="gpt-realtime",
         voice="en-US-AvaNeural",
         instructions="You are a helpful AI assistant with access to functions. "
         "Use the functions when appropriate to provide accurate, real-time information. "

sdk/voicelive/azure-ai-voicelive/samples/async_mcp_sample.py

@@ -668,7 +668,7 @@ async def main():
     client = AsyncMCPCallClient(
         endpoint=endpoint,
         credential=credential,
-        model="gpt-4o-realtime-preview",
+        model="gpt-realtime",
         voice="en-US-AvaNeural",
         instructions="You are a helpful AI assistant with access to some mcp server. ",
     )

sdk/voicelive/azure-ai-voicelive/samples/telemetry/sample_voicelive_with_azure_monitor_tracing.py

@@ -20,7 +20,7 @@
     Set these environment variables with your own values:
     1) AZURE_VOICELIVE_ENDPOINT - The Azure VoiceLive endpoint URL.
     2) AZURE_VOICELIVE_API_KEY - The Azure VoiceLive API key.
-    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-4o-realtime-preview).
+    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-realtime).
     4) AZURE_EXPERIMENTAL_ENABLE_GENAI_TRACING - Set to "true" to enable tracing.
     5) APPLICATIONINSIGHTS_CONNECTION_STRING - The connection string for your Application Insights resource.
     6) OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT - Optional. Set to "true" to trace the content of
@@ -67,7 +67,7 @@
 async def main() -> None:
     endpoint = os.environ["AZURE_VOICELIVE_ENDPOINT"]
     api_key = os.environ["AZURE_VOICELIVE_API_KEY"]
-    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-4o-realtime-preview")
+    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-realtime")
 
     credential = AzureKeyCredential(api_key)
 

sdk/voicelive/azure-ai-voicelive/samples/telemetry/sample_voicelive_with_console_tracing.py

@@ -26,7 +26,7 @@
     Set these environment variables with your own values:
     1) AZURE_VOICELIVE_ENDPOINT - The Azure VoiceLive endpoint URL.
     2) AZURE_VOICELIVE_API_KEY - The Azure VoiceLive API key.
-    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-4o-realtime-preview).
+    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-realtime).
     4) AZURE_EXPERIMENTAL_ENABLE_GENAI_TRACING - Set to "true" to enable tracing.
     5) OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT - Optional. Set to "true" to trace the content of
        messages, which may contain personal data. False by default.
@@ -74,7 +74,7 @@
 async def main() -> None:
     endpoint = os.environ["AZURE_VOICELIVE_ENDPOINT"]
     api_key = os.environ["AZURE_VOICELIVE_API_KEY"]
-    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-4o-realtime-preview")
+    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-realtime")
 
     credential = AzureKeyCredential(api_key)
 

sdk/voicelive/azure-ai-voicelive/samples/telemetry/sample_voicelive_with_console_tracing_custom_attributes.py

@@ -21,7 +21,7 @@
     Set these environment variables with your own values:
     1) AZURE_VOICELIVE_ENDPOINT - The Azure VoiceLive endpoint URL.
     2) AZURE_VOICELIVE_API_KEY - The Azure VoiceLive API key.
-    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-4o-realtime-preview).
+    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-realtime).
     4) AZURE_EXPERIMENTAL_ENABLE_GENAI_TRACING - Set to "true" to enable tracing.
     5) OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT - Optional. Set to "true" to trace the content of
        messages, which may contain personal data. False by default.
@@ -102,7 +102,7 @@ def on_end(self, span: ReadableSpan):
 async def main() -> None:
     endpoint = os.environ["AZURE_VOICELIVE_ENDPOINT"]
     api_key = os.environ["AZURE_VOICELIVE_API_KEY"]
-    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-4o-realtime-preview")
+    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-realtime")
 
     credential = AzureKeyCredential(api_key)
 

sdk/voicelive/azure-ai-voicelive/samples/telemetry/sample_voicelive_with_content_recording.py

@@ -24,7 +24,7 @@
     Set these environment variables with your own values:
     1) AZURE_VOICELIVE_ENDPOINT - The Azure VoiceLive endpoint URL.
     2) AZURE_VOICELIVE_API_KEY - The Azure VoiceLive API key.
-    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-4o-realtime-preview).
+    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-realtime).
     4) AZURE_EXPERIMENTAL_ENABLE_GENAI_TRACING - Set to "true" to enable tracing.
 """
 
@@ -76,7 +76,7 @@
 async def main() -> None:
     endpoint = os.environ["AZURE_VOICELIVE_ENDPOINT"]
     api_key = os.environ["AZURE_VOICELIVE_API_KEY"]
-    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-4o-realtime-preview")
+    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-realtime")
 
     credential = AzureKeyCredential(api_key)
 

sdk/voicelive/azure-ai-voicelive/samples/telemetry/sample_voicelive_with_telemetry_enablement.py

@@ -0,0 +1,110 @@
+# pylint: disable=line-too-long,useless-suppression
+# ------------------------------------
+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT License.
+# ------------------------------------
+
+"""
+DESCRIPTION:
+    This sample shows the minimum code changes needed to enable VoiceLive telemetry
+    on top of existing VoiceLive client code.
+
+USAGE:
+    python sample_voicelive_with_telemetry_enablement.py
+
+    Before running the sample:
+
+    pip install azure-ai-voicelive azure-identity opentelemetry-sdk azure-core-tracing-opentelemetry
+
+    Set these environment variables with your own values:
+    1) AZURE_VOICELIVE_ENDPOINT - The Azure VoiceLive endpoint URL.
+    2) AZURE_VOICELIVE_API_KEY - The Azure VoiceLive API key.
+    3) AZURE_VOICELIVE_MODEL - The model deployment name (e.g., gpt-realtime).
+
+    Optional telemetry variables:
+    4) AZURE_EXPERIMENTAL_ENABLE_GENAI_TRACING - Set to "true" to enable tracing.
+    5) OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT - Optional. Set to "true" to include
+       message content in telemetry events. False by default.
+"""
+
+import asyncio
+import os
+
+from azure.core.credentials import AzureKeyCredential
+from azure.core.settings import settings
+from azure.ai.voicelive.aio import connect
+from azure.ai.voicelive.models import (
+    InputTextContentPart,
+    Modality,
+    OutputAudioFormat,
+    RequestSession,
+    ServerEventType,
+    ServerVad,
+    UserMessageItem,
+)
+from azure.ai.voicelive.telemetry import VoiceLiveInstrumentor
+
+from opentelemetry import trace
+from opentelemetry.sdk.trace import TracerProvider
+from opentelemetry.sdk.trace.export import SimpleSpanProcessor, ConsoleSpanExporter
+
+
+def enable_telemetry_minimal():
+    """Enable telemetry with the smallest practical setup for existing code."""
+    # Add these lines to your existing app to enable VoiceLive telemetry.
+    # 1) Tell azure-core to use OpenTelemetry tracing.
+    settings.tracing_implementation = "opentelemetry"
+
+    # 2) Configure a tracer provider and exporter (console in this sample).
+    provider = TracerProvider()
+    provider.add_span_processor(SimpleSpanProcessor(ConsoleSpanExporter()))
+    trace.set_tracer_provider(provider)
+
+    # 3) Enable the VoiceLive instrumentation gate and instrument the SDK.
+    os.environ.setdefault("AZURE_EXPERIMENTAL_ENABLE_GENAI_TRACING", "true")
+    VoiceLiveInstrumentor().instrument()
+
+    return provider
+
+
+async def main() -> None:
+    tracer_provider = enable_telemetry_minimal()
+
+    endpoint = os.environ["AZURE_VOICELIVE_ENDPOINT"]
+    api_key = os.environ["AZURE_VOICELIVE_API_KEY"]
+    model = os.environ.get("AZURE_VOICELIVE_MODEL", "gpt-realtime")
+
+    credential = AzureKeyCredential(api_key)
+
+    async with connect(
+        endpoint=endpoint,
+        credential=credential,
+        model=model,
+    ) as connection:
+        print(f"Connected to VoiceLive at {endpoint}")
+
+        session_config = RequestSession(
+            modalities=[Modality.TEXT],
+            instructions="You are a helpful assistant. Keep the answer short.",
+            turn_detection=ServerVad(threshold=0.5, prefix_padding_ms=300, silence_duration_ms=500),
+            output_audio_format=OutputAudioFormat.PCM16,
+        )
+        await connection.session.update(session=session_config)
+
+        await connection.conversation.item.create(
+            item=UserMessageItem(content=[InputTextContentPart(text="Say hello in one sentence")])
+        )
+        await connection.response.create()
+
+        async for event in connection:
+            event_type = getattr(event, "type", None)
+            print(f"Received event: {event_type}")
+            if event_type == ServerEventType.RESPONSE_DONE:
+                break
+
+    tracer_provider.force_flush()
+    tracer_provider.shutdown()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())