chore: gemini 3 changes.

Author: thorwebdev

gemini-live-ephemeral-tokens-websocket/frontend/geminilive.js

@@ -323,18 +323,18 @@ class GeminiLiveAPI {
         },
         systemInstruction: { parts: [{ text: this.systemInstructions }] },
         tools: { functionDeclarations: tools },
-        proactivity: this.proactivity,
-
-        realtimeInputConfig: {
-          automaticActivityDetection: {
-            disabled: this.automaticActivityDetection.disabled,
-            silenceDurationMs: this.automaticActivityDetection.silence_duration_ms,
-            prefixPaddingMs: this.automaticActivityDetection.prefix_padding_ms,
-            endOfSpeechSensitivity: this.automaticActivityDetection.end_of_speech_sensitivity,
-            startOfSpeechSensitivity: this.automaticActivityDetection.start_of_speech_sensitivity,
-          },
-          activityHandling: this.activityHandling,
-        },
+        // proactivity: this.proactivity,
+
+        // realtimeInputConfig: {
+        //   automaticActivityDetection: {
+        //     disabled: this.automaticActivityDetection.disabled,
+        //     silenceDurationMs: this.automaticActivityDetection.silence_duration_ms,
+        //     prefixPaddingMs: this.automaticActivityDetection.prefix_padding_ms,
+        //     endOfSpeechSensitivity: this.automaticActivityDetection.end_of_speech_sensitivity,
+        //     startOfSpeechSensitivity: this.automaticActivityDetection.start_of_speech_sensitivity,
+        //   },
+        //   activityHandling: this.activityHandling,
+        // },
       },
     };
 
@@ -356,9 +356,9 @@ class GeminiLiveAPI {
     }
 
     // Add affective dialog if enabled
-    if (this.enableAffectiveDialog) {
-      sessionSetupMessage.setup.generationConfig.enableAffectiveDialog = true;
-    }
+    // if (this.enableAffectiveDialog) {
+    //   sessionSetupMessage.setup.generationConfig.enableAffectiveDialog = true;
+    // }
 
     // Store the setup message for later access
     this.lastSetupMessage = sessionSetupMessage;
@@ -394,16 +394,15 @@ class GeminiLiveAPI {
   }
 
   sendRealtimeInputMessage(data, mimeType) {
-    const message = {
-      realtimeInput: {
-        mediaChunks: [
-          {
-            mimeType: mimeType,
-            data: data,
-          },
-        ],
-      },
-    };
+    const blob = { mimeType, data };
+    const message = { realtimeInput: {} };
+
+    if (mimeType.startsWith("audio/")) {
+      message.realtimeInput.audio = blob;
+    } else if (mimeType.startsWith("image/") || mimeType.startsWith("video/")) {
+      message.realtimeInput.video = blob;
+    }
+
     this.sendMessage(message);
     this.addToBytesSent(data);
   }

gemini-live-genai-python-sdk/gemini_live.py

@@ -1,6 +1,7 @@
 import asyncio
 import inspect
 import logging
+import traceback
 
 logger = logging.getLogger(__name__)
 from google import genai
@@ -24,7 +25,7 @@ def __init__(self, api_key, model, input_sample_rate, tools=None, tool_mapping=N
         self.api_key = api_key
         self.model = model
         self.input_sample_rate = input_sample_rate
-        self.client = genai.Client(api_key=api_key, http_options={"api_version": "v1alpha"})
+        self.client = genai.Client(api_key=api_key)
         self.tools = tools or []
         self.tool_mapping = tool_mapping or {}
 
@@ -41,12 +42,15 @@ async def start_session(self, audio_input_queue, video_input_queue, text_input_q
             system_instruction=types.Content(parts=[types.Part(text="You are a helpful AI assistant. Keep your responses concise. Speak in a friendly Irish accent. You can see the user's camera or screen which is shared as realtime input images with you.")]),
             input_audio_transcription=types.AudioTranscriptionConfig(),
             output_audio_transcription=types.AudioTranscriptionConfig(),
-            proactivity=types.ProactivityConfig(proactive_audio=True),
-            enable_affective_dialog=True,
+            #proactivity=types.ProactivityConfig(proactive_audio=True),
+            #enable_affective_dialog=True,
             tools=self.tools,
         )
 
-        async with self.client.aio.live.connect(model=self.model, config=config) as session:
+        logger.info(f"Connecting to Gemini Live with model={self.model}")
+        try:
+          async with self.client.aio.live.connect(model=self.model, config=config) as session:
+            logger.info("Gemini Live session opened successfully")
 
             async def send_audio():
                 try:
@@ -56,18 +60,22 @@ async def send_audio():
                             audio=types.Blob(data=chunk, mime_type=f"audio/pcm;rate={self.input_sample_rate}")
                         )
                 except asyncio.CancelledError:
-                    pass
+                    logger.debug("send_audio task cancelled")
+                except Exception as e:
+                    logger.error(f"send_audio error: {e}\n{traceback.format_exc()}")
 
             async def send_video():
                 try:
                     while True:
                         chunk = await video_input_queue.get()
                         logger.info(f"Sending video frame to Gemini: {len(chunk)} bytes")
                         await session.send_realtime_input(
-                            media=types.Blob(data=chunk, mime_type="image/jpeg")
+                            video=types.Blob(data=chunk, mime_type="image/jpeg")
                         )
                 except asyncio.CancelledError:
-                    pass
+                    logger.debug("send_video task cancelled")
+                except Exception as e:
+                    logger.error(f"send_video error: {e}\n{traceback.format_exc()}")
 
             async def send_text():
                 try:
@@ -76,7 +84,9 @@ async def send_text():
                         logger.info(f"Sending text to Gemini: {text}")
                         await session.send_realtime_input(text=text)
                 except asyncio.CancelledError:
-                    pass
+                    logger.debug("send_text task cancelled")
+                except Exception as e:
+                    logger.error(f"send_text error: {e}\n{traceback.format_exc()}")
 
             event_queue = asyncio.Queue()
 
@@ -85,6 +95,13 @@ async def receive_loop():
                     while True:
                         async for response in session.receive():
                             logger.debug(f"Received response from Gemini: {response}")
+                            
+                            # Log the raw response type for debugging
+                            if response.go_away:
+                                logger.warning(f"Received GoAway from Gemini: {response.go_away}")
+                            if response.session_resumption_update:
+                                logger.info(f"Session resumption update: {response.session_resumption_update}")
+                            
                             server_content = response.server_content
                             tool_call = response.tool_call
 
@@ -139,10 +156,17 @@ async def receive_loop():
                                         await event_queue.put({"type": "tool_call", "name": func_name, "args": args, "result": result})
 
                                 await session.send_tool_response(function_responses=function_responses)
+                        
+                        # session.receive() iterator ended (e.g. after turn_complete) — re-enter to keep listening
+                        logger.debug("Gemini receive iterator completed, re-entering receive loop")
 
+                except asyncio.CancelledError:
+                    logger.debug("receive_loop task cancelled")
                 except Exception as e:
-                    await event_queue.put({"type": "error", "error": str(e)})
+                    logger.error(f"receive_loop error: {type(e).__name__}: {e}\n{traceback.format_exc()}")
+                    await event_queue.put({"type": "error", "error": f"{type(e).__name__}: {e}"})
                 finally:
+                    logger.info("receive_loop exiting")
                     await event_queue.put(None)
 
             send_audio_task = asyncio.create_task(send_audio())
@@ -161,7 +185,13 @@ async def receive_loop():
                         break 
                     yield event
             finally:
+                logger.info("Cleaning up Gemini Live session tasks")
                 send_audio_task.cancel()
                 send_video_task.cancel()
                 send_text_task.cancel()
                 receive_task.cancel()
+        except Exception as e:
+            logger.error(f"Gemini Live session error: {type(e).__name__}: {e}\n{traceback.format_exc()}")
+            raise
+        finally:
+            logger.info("Gemini Live session closed")

gemini-live-genai-python-sdk/main.py

@@ -14,8 +14,10 @@
 # Load environment variables
 load_dotenv()
 
-# Configure logging
+# Configure logging - DEBUG for our modules, INFO for everything else
 logging.basicConfig(level=logging.INFO)
+logging.getLogger("gemini_live").setLevel(logging.DEBUG)
+logging.getLogger(__name__).setLevel(logging.DEBUG)
 logger = logging.getLogger(__name__)
 
 # Configuration
@@ -106,7 +108,8 @@ async def run_session():
     try:
         await run_session()
     except Exception as e:
-        logger.error(f"Error in Gemini session: {e}")
+        import traceback
+        logger.error(f"Error in Gemini session: {type(e).__name__}: {e}\n{traceback.format_exc()}")
     finally:
         receive_task.cancel()
         # Ensure websocket is closed if not already