Fix soundcard import. Add Twillio TURN server.

Author: paulovcmedeiros

pyproject.toml

@@ -4,7 +4,7 @@
   license = "MIT"
   name = "pyrobbot"
   readme = "README.md"
-  version = "0.7.0"
+  version = "0.7.1"
 
 [build-system]
   build-backend = "poetry.core.masonry.api"
@@ -24,7 +24,7 @@
   # Other dependencies
   loguru = "^0.7.2"
   numpy = "^1.26.1"
-  openai = "^1.12.0"
+  openai = "^1.13.3"
   pandas = "^2.2.0"
   pillow = "^10.2.0"
   pydantic = "^2.6.1"
@@ -34,18 +34,19 @@
   audio-recorder-streamlit = "^0.0.8"
   beautifulsoup4 = "^4.12.3"
   chime = "^0.7.0"
-  duckduckgo-search = {git = "https://github.com/deedy5/duckduckgo_search"}
+  duckduckgo-search = {url = "https://github.com/deedy5/duckduckgo_search/archive/refs/tags/v5.0b1.tar.gz"}
   gtts = "^2.5.1"
   httpx = "^0.26.0"
   ipinfo = "^5.0.1"
   pydub = "^0.25.1"
   pygame = "^2.5.2"
-  setuptools = "^68.2.2"              # Needed by webrtcvad-wheels
+  setuptools = "^68.2.2"                                                                                    # Needed by webrtcvad-wheels
   sounddevice = "^0.4.6"
   soundfile = "^0.12.1"
   speechrecognition = "^3.10.0"
   streamlit-mic-recorder = "^0.0.4"
   streamlit-webrtc = "^0.47.1"
+  twilio = "^9.0.0"
   tzlocal = "^5.2"
   unidecode = "^1.3.7"
   webrtcvad-wheels = "^2.0.11.post1"
@@ -58,7 +59,7 @@
   flakeheaven = "^3.3.0"
   isort = "^5.13.2"
   pydoclint = "^0.4.0"
-  ruff = "^0.2.1"
+  ruff = "^0.3.0"
 
 [tool.poetry.group.test.dependencies]
   pytest = "^8.0.0"

pyrobbot/app/app_utils.py

@@ -2,6 +2,7 @@
 
 import contextlib
 import datetime
+import os
 import queue
 import threading
 from typing import TYPE_CHECKING
@@ -11,6 +12,7 @@
 from PIL import Image
 from pydub import AudioSegment
 from streamlit.runtime.scriptrunner import add_script_run_ctx
+from twilio.rest import Client as TwilioClient
 
 from pyrobbot import GeneralDefinitions
 from pyrobbot.chat import AssistantResponseChunk
@@ -149,6 +151,24 @@ def stream_text_and_audio_reply(self):
         return {"text": full_response, "audio": full_audio_fpath}
 
 
+@st.cache_data
+def get_ice_servers():
+    """Use Twilio's TURN server as recommended by the streamlit-webrtc developers."""
+    try:
+        account_sid = os.environ["TWILIO_ACCOUNT_SID"]
+        auth_token = os.environ["TWILIO_AUTH_TOKEN"]
+    except KeyError:
+        logger.warning(
+            "Twilio credentials are not set. Cannot use their TURN servers. "
+            "Falling back to a free STUN server from Google."
+        )
+        return [{"urls": ["stun:stun.l.google.com:19302"]}]
+
+    client = TwilioClient(account_sid, auth_token)
+    token = client.tokens.create()
+    return token.ice_servers
+
+
 def filter_page_info_from_queue(app_page: "AppPage", the_queue: queue.Queue):
     """Filter `app_page`'s data from `queue` inplace. Return queue of items in `app_page`.
 

pyrobbot/app/multipage.py

@@ -18,15 +18,20 @@
 from pydantic import ValidationError
 from pydub import AudioSegment
 from streamlit.runtime.scriptrunner import add_script_run_ctx
-from streamlit_webrtc import RTCConfiguration, WebRtcMode
+from streamlit_webrtc import WebRtcMode
 
 from pyrobbot import GeneralDefinitions
 from pyrobbot.chat_configs import VoiceChatConfigs
 from pyrobbot.general_utils import trim_beginning
 from pyrobbot.openai_utils import OpenAiClientWrapper
 
 from .app_page_templates import AppPage, ChatBotPage, _RecoveredChat
-from .app_utils import WebAppChat, filter_page_info_from_queue, get_avatar_images
+from .app_utils import (
+    WebAppChat,
+    filter_page_info_from_queue,
+    get_avatar_images,
+    get_ice_servers,
+)
 
 incoming_frame_queue = queue.Queue()
 possible_speech_chunks_queue = queue.Queue()
@@ -308,6 +313,11 @@ def __init__(self, **kwargs) -> None:
         self.text_prompt_queue = text_prompt_queue
         self.reply_ongoing = reply_ongoing
 
+    @property
+    def ice_servers(self):
+        """Return the ICE servers for WebRTC."""
+        return get_ice_servers()
+
     @property
     def continuous_audio_input_engine_is_running(self):
         """Return whether the continuous audio input engine is running."""
@@ -318,12 +328,7 @@ def continuous_audio_input_engine_is_running(self):
         )
 
     def render_continuous_audio_input_widget(self):
-        """Render the continuous audio input widget."""
-        # Definitions related to webrtc_streamer
-        rtc_configuration = RTCConfiguration(
-            {"iceServers": [{"urls": ["stun:stun.l.google.com:19302"]}]}
-        )
-
+        """Render the continuous audio input widget using webrtc_streamer."""
         try:
             selected_page = self.selected_page
         except StopIteration:
@@ -377,7 +382,7 @@ def audio_frame_callback(frame):
             self.stream_audio_context = streamlit_webrtc.component.webrtc_streamer(
                 key="sendonly-audio",
                 mode=WebRtcMode.SENDONLY,
-                rtc_configuration=rtc_configuration,
+                rtc_configuration={"iceServers": self.ice_servers},
                 media_stream_constraints={"audio": True, "video": False},
                 desired_playing_state=True,
                 audio_frame_callback=audio_frame_callback,

pyrobbot/voice_chat.py

@@ -25,7 +25,7 @@
 try:
     import sounddevice as sd
 except OSError as error:
-    logger.error(error)
+    logger.exception(error)
     logger.error(
         "Can't use module `sounddevice`. Please check your system's PortAudio install."
     )
@@ -516,8 +516,8 @@ def _assistant_still_replying(self):
 def _check_needed_imports():
     """Check if the needed modules are available."""
     if not _sounddevice_imported:
-        raise ImportError(
-            "Module `sounddevice`, needed for audio recording, is not available."
+        logger.error(
+            "Module `sounddevice`, needed for local audio recording, is not available."
         )
 
     if not _pydub_imported:

tests/conftest.py

@@ -150,12 +150,15 @@ def _mock_input(*args, **kwargs):  # noqa: ARG001
     )
 
 
-@pytest.fixture(params=ChatOptions.get_allowed_values("model")[:3])
+@pytest.fixture(params=ChatOptions.get_allowed_values("model")[:2])
 def llm_model(request):
     return request.param
 
 
-@pytest.fixture(params=ChatOptions.get_allowed_values("context_model")[:3])
+context_model_values = ChatOptions.get_allowed_values("context_model")
+
+
+@pytest.fixture(params=[context_model_values[0], context_model_values[2]])
 def context_model(request):
     return request.param
 

tests/unit/test_voice_chat.py

@@ -10,11 +10,12 @@
 from pyrobbot.voice_chat import VoiceChat
 
 
-def test_cannot_instanciate_assistant_if_soundcard_not_imported(mocker):
+def test_soundcard_import_check(mocker, caplog):
     """Test that the voice chat cannot be instantiated if soundcard is not imported."""
     mocker.patch("pyrobbot.voice_chat._sounddevice_imported", False)
-    with pytest.raises(ImportError, match="Module `sounddevice`"):
-        VoiceChat(configs=VoiceChatConfigs())
+    _ = VoiceChat(configs=VoiceChatConfigs())
+    msg = "Module `sounddevice`, needed for local audio recording, is not available."
+    assert msg in caplog.text
 
 
 @pytest.mark.parametrize("param_name", ["sample_rate", "frame_duration"])