Merge pull request #1689 from pipecat-ai/mb/handle-http-smart-turn-errors

Handle case where Fal Smart Turn returns a 500 error

CHANGELOG.md

@@ -60,9 +60,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Deprecated
 
-- Function calls with parameters `(function_name, tool_call_id, args, llm,
-  context, result_callback)` are deprectated, use a single `FunctionCallParams`
-  parameter instead.
+- Function calls with parameters
+  `(function_name, tool_call_id, args, llm, context, result_callback)` are
+  deprectated, use a single `FunctionCallParams` parameter instead.
 
 - `TransportParams.camera_*` parameters are now deprecated, use
   `TransportParams.video_*` instead.
@@ -75,6 +75,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- Fixed an issue with HTTP Smart Turn handling, where the service returns a 500
+  error. Previously, this would cause an unhandled exception. Now, a 500 error
+  is treated as an incomplete response.
+
 - Fixed a TTS services issue that could cause assistant output not to be
   aggregated to the context when also using `TTSSpeakFrame`s.
 

src/pipecat/audio/turn/smart_turn/http_smart_turn.py

@@ -40,7 +40,7 @@ class HttpSmartTurnAnalyzer(BaseSmartTurn):
     async def _send_raw_request(self, data_bytes: bytes) -> Dict[str, Any]:
         headers = {"Content-Type": "application/octet-stream"}
         headers.update(self._headers)
-        logger.trace(f"Sending {len(data_bytes)} bytes as raw body to {self._url}...")
+
         try:
             timeout = aiohttp.ClientTimeout(total=self._params.stop_secs)
 
@@ -50,23 +50,30 @@ class HttpSmartTurnAnalyzer(BaseSmartTurn):
                 logger.trace("\n--- Response ---")
                 logger.trace(f"Status Code: {response.status}")
 
-                if response.status == 200:
-                    try:
-                        json_data = await response.json()
-                        logger.trace("Response JSON:")
-                        logger.trace(json_data)
-                        return json_data
-                    except aiohttp.ContentTypeError:
-                        # Non-JSON response
-                        text = await response.text()
-                        logger.trace("Response Content (non-JSON):")
-                        logger.trace(text)
-                        raise Exception(f"Non-JSON response: {text}")
-                else:
+                # Check if successful
+                if response.status != 200:
                     error_text = await response.text()
                     logger.trace("Response Content (Error):")
                     logger.trace(error_text)
-                    response.raise_for_status()
+
+                    if response.status == 500:
+                        logger.warning(f"Smart turn service returned 500 error: {error_text}")
+                        raise Exception(f"Server returned HTTP 500: {error_text}")
+                    else:
+                        response.raise_for_status()
+
+                # Process successful response
+                try:
+                    json_data = await response.json()
+                    logger.trace("Response JSON:")
+                    logger.trace(json_data)
+                    return json_data
+                except aiohttp.ContentTypeError:
+                    # Non-JSON response
+                    text = await response.text()
+                    logger.trace("Response Content (non-JSON):")
+                    logger.trace(text)
+                    raise Exception(f"Non-JSON response: {text}")
 
         except asyncio.TimeoutError:
             logger.error(f"Request timed out after {self._params.stop_secs} seconds")
@@ -76,5 +83,14 @@ class HttpSmartTurnAnalyzer(BaseSmartTurn):
             raise Exception("Failed to send raw request to Daily Smart Turn.")
 
     async def _predict_endpoint(self, audio_array: np.ndarray) -> Dict[str, Any]:
-        serialized_array = self._serialize_array(audio_array)
-        return await self._send_raw_request(serialized_array)
+        try:
+            serialized_array = self._serialize_array(audio_array)
+            return await self._send_raw_request(serialized_array)
+        except Exception as e:
+            logger.error(f"Smart turn prediction failed: {str(e)}")
+            # Return an incomplete prediction when a failure occurs
+            return {
+                "prediction": 0,
+                "probability": 0.0,
+                "metrics": {"inference_time": 0.0, "total_time": 0.0},
+            }