updating speech to speech plugin

Author: maxkahan

examples/env.example

@@ -13,3 +13,4 @@ ELEVENLABS_VOICE_ID=voice_id_to_use
 # OpenAI API credentials
 OPENAI_API_KEY=your_openai_api_key_here
 OPENAI_MODEL=model_to_use
+OPENAI_REALTIME_MODEL=gpt-4o-realtime-preview

examples/openai_realtime_speech_to_speech/main.py

@@ -7,8 +7,6 @@
 from getstream import Stream
 from getstream.models import StartClosedCaptionsResponse
 from getstream.plugins.sts.openai_realtime import OpenAIRealtime
-from dataclasses import asdict
-import json
 
 
 logging.basicConfig(
@@ -26,7 +24,6 @@ async def main():
 
     load_dotenv(os.path.join(os.path.dirname(__file__), "..", ".env"))
 
-    # Initialize Stream client from env vars (STREAM_API_KEY / SECRET / BASE_URL)
     client = Stream.from_env()
 
     user_id = f"user-{uuid4()}"
@@ -52,15 +49,14 @@ async def main():
 
     sts_bot = OpenAIRealtime(
         api_key=os.getenv("OPENAI_API_KEY"),
-        model="gpt-4o-realtime-preview",
+        model=os.getenv("OPENAI_REALTIME_MODEL"),
         instructions="You are a friendly assistant; reply verbally in a short sentence.",
         voice="alloy",
     )
 
     try:
         logging.info("Connecting to OpenAI Realtime...")
 
-        # Check if API key is set
         if not os.getenv("OPENAI_API_KEY"):
             logging.error("❌ OPENAI_API_KEY not found in environment")
             return
@@ -88,6 +84,8 @@ async def main():
                 },
                 tools=tools,
             )
+        
+            await sts_bot.send_user_message("Give a short greeting to the user.")
 
             logging.info("🎧 Listening for responses... (Press Ctrl+C to stop)")
             logging.info("💡 Try speaking in the browser – ask it something like 'start closed captions' to trigger the function call.")
@@ -111,14 +109,10 @@ async def start_closed_captions() -> StartClosedCaptionsResponse:
                         logging.info("🛠  Assistant requested start_closed_captions()")
 
                         result = await start_closed_captions()
-
-                        # Send the tool result back to the assistant
                         await sts_bot.send_function_call_output(tool_call_id, result.to_json())
-
+                        await sts_bot.request_assistant_response()
                         logging.info("🛠  Replied to tool call with result: %s", result)
 
-            
-
     except KeyboardInterrupt:  # noqa: WPS420
         logging.info("\n⏹️  Stopping OpenAI Realtime Speech to Speech bot…")
     except Exception as e:  # noqa: BLE001

getstream/plugins/sts/openai_realtime/sts.py

@@ -180,6 +180,13 @@ async def send_user_message(self, text: str):
                 ],
             }
         )
+        await self.request_assistant_response()
+    
+    async def request_assistant_response(self):
+        """Ask OpenAI to generate the next assistant turn."""
+        if not self._is_connected or not self._connection:
+            raise RuntimeError("Not connected")
+        
         await self._connection.response.create()
 
 

getstream/plugins/sts/openai_realtime/tests/pyproject.toml

@@ -97,6 +97,17 @@ async def test_send_function_call_output(mock_call, mock_connection):
     )
 
 
+@pytest.mark.asyncio
+async def test_request_assistant_response(mock_call, mock_connection):
+    mock_call.connect_openai.return_value = mock_connection  # type: ignore[attr-defined]
+    sts = OpenAIRealtime(api_key="abc")
+    await sts.connect(mock_call)
+
+    await sts.request_assistant_response()
+
+    mock_connection.response.create.assert_awaited_once()
+
+
 @pytest.mark.asyncio
 async def test_methods_raise_if_not_connected():
     sts = OpenAIRealtime(api_key="abc")
@@ -105,4 +116,6 @@ async def test_methods_raise_if_not_connected():
     with pytest.raises(RuntimeError):
         await sts.send_user_message("hi")
     with pytest.raises(RuntimeError):
-        await sts.send_function_call_output("id", "out") 
+        await sts.send_function_call_output("id", "out")
+    with pytest.raises(RuntimeError):
+        await sts.request_assistant_response() 

getstream/plugins/sts/openai_realtime/tests/test_openai_realtime.py

@@ -171,12 +171,12 @@ def patch_realtime_connect(client):
     try:
         # First, try to import the AsyncRealtimeConnection class
         try:
-            from openai.resources.beta.realtime.connection import (
+            from openai.resources.beta.realtime.connection import ( # type: ignore
                 AsyncRealtimeConnection,
             )
         except ImportError:
             # If that fails, try another possible location
-            from openai.types.beta.realtime.connection import AsyncRealtimeConnection
+            from openai.types.beta.realtime.connection import AsyncRealtimeConnection # type: ignore
 
         # Store the original recv method and replace it with our patched version
         if not hasattr(AsyncRealtimeConnection, "recv"):
@@ -202,7 +202,7 @@ def patch_realtime_connect(client):
             )
         except ImportError:
             # If that fails, try another possible location
-            from openai.types.beta.realtime.realtime import (
+            from openai.types.beta.realtime.realtime import ( # type: ignore
                 AsyncRealtimeConnectionManager,
             )
 
@@ -251,7 +251,7 @@ def patch_realtime_connect(client):
             )
         except ImportError:
             # If that fails, try another possible location
-            from openai.types.beta.realtime.realtime import (
+            from openai.types.beta.realtime.realtime import ( # type: ignore
                 AsyncRealtimeConnectionManager,
             )