Accept all media types

custom_components/stream_assist/config_flow.py

@@ -79,6 +79,7 @@ async def async_step_init(self, user_input: dict = None):
                     vol.Optional("player_entity_id"): cv.multi_select(players),
                     vol.Optional("stt_start_media"): str,
                     vol.Optional("pipeline_id"): vol.In(pipelines),
+                    vol.Optional("allow_all_mediatypes"): bool,
                 },
                 defaults,
             ),

custom_components/stream_assist/core/__init__.py

@@ -67,6 +67,8 @@ async def stream_run(hass: HomeAssistant, data: dict, stt_stream: Stream) -> Non
             stream_kwargs["file"] = await get_stream_source(hass, entity)
         else:
             return
+    if "allow_all_mediatypes" not in stream_kwargs:
+        stream_kwargs["allow_all_mediatypes"] = data.get("allow_all_mediatypes", False)
 
     stt_stream.open(**stream_kwargs)
 

custom_components/stream_assist/core/stream.py

@@ -15,9 +15,7 @@ def __init__(self):
         self.enabled: bool = False
         self.queue: asyncio.Queue[bytes] = asyncio.Queue()
 
-    def open(self, file: str, **kwargs):
-        _LOGGER.debug(f"stream open")
-
+    def open(self, file: str, allow_all_mediatypes: bool, **kwargs):
         if "options" not in kwargs:
             kwargs["options"] = {
                 "fflags": "nobuffer",
@@ -27,15 +25,21 @@ def open(self, file: str, **kwargs):
 
             if file.startswith("rtsp"):
                 kwargs["options"]["rtsp_flags"] = "prefer_tcp"
-                kwargs["options"]["allowed_media_types"] = "audio"
+                if not allow_all_mediatypes:
+                    kwargs["options"]["allowed_media_types"] = "audio"
 
         kwargs.setdefault("timeout", 5)
+        _LOGGER.debug("Starting stream with kwargs: %s", kwargs)
 
-        # https://pyav.org/docs/9.0.2/api/_globals.html
-        self.container = av.open(file, **kwargs)
+        try:
+            # https://pyav.org/docs/9.0.2/api/_globals.html
+            self.container = av.open(file, **kwargs)
+            _LOGGER.info("Stream started")
+        except Exception as e:
+            _LOGGER.error("Can't open stream: %s", e)
 
     def run(self, end=True):
-        _LOGGER.debug("stream start")
+        _LOGGER.info("stream start")
 
         resampler = AudioResampler(format="s16", layout="mono", rate=16000)
 
@@ -49,18 +53,18 @@ def run(self, end=True):
                     chunk = frame_raw.to_ndarray().tobytes()
                     self.queue.put_nowait(chunk)
         except Exception as e:
-            _LOGGER.debug(f"stream exception {type(e)}: {e}")
+            _LOGGER.error(f"stream exception {type(e)}: {e}")
         finally:
             self.container.close()
             self.container = None
 
         if end and self.enabled:
             self.queue.put_nowait(b"")
 
-        _LOGGER.debug("stream end")
+        _LOGGER.info("stream end")
 
     def close(self):
-        _LOGGER.debug(f"stream close")
+        _LOGGER.info(f"stream close")
         self.closed = True
 
     def start(self):

custom_components/stream_assist/services.yaml

@@ -52,3 +52,9 @@ run:
       description: Advanced parameters
       selector:
         object:
+
+    allow_all_mediatypes:
+      name: Allow all media types
+      description: Activate if RTSP camera doesn't support filtering media types
+      selector:
+        boolean:

custom_components/stream_assist/translations/en.json

@@ -21,7 +21,8 @@
           "camera_entity_id": "Camera Entity",
           "player_entity_id": "Player Entity",
           "stt_start_media": "STT start media",
-          "pipeline_id": "Pipeline (Settings > Voice Assistant)"
+          "pipeline_id": "Pipeline (Settings > Voice Assistant)",
+          "allow_all_mediatypes": "Allow all Media Types"
         }
       }
     }