Enhance LiveKit integration: Add audio input and VAD configurations, update settings, and improve Streamlit app for better agent dispatch and error handling.

.env.example

@@ -21,3 +21,15 @@ LIVEKIT_API_SECRET=your_livekit_api_secret_here
 LIVEKIT_AGENT_NAME=open-voice-agent
 LIVEKIT_NUM_IDLE_PROCESSES=2
 
+# LiveKit Audio Input Configuration - OPTIMIZED FOR FALSE DETECTION FIX
+LIVEKIT_SAMPLE_RATE=24000
+LIVEKIT_NUM_CHANNELS=1
+LIVEKIT_FRAME_SIZE_MS=20  # Smaller = faster VAD response, less latency
+LIVEKIT_PRE_CONNECT_AUDIO=true
+LIVEKIT_PRE_CONNECT_TIMEOUT=3.0
+
+# Voice Activity Detection (VAD) Configuration - OPTIMIZED FOR FALSE DETECTION FIX
+VAD_MIN_SPEECH_DURATION=0.25  # Require 250ms of speech before activation
+VAD_MIN_SILENCE_DURATION=0.5  # Require 500ms of silence before deactivation
+VAD_THRESHOLD=0.6  # Higher = less sensitive to noise (0.5 is default)
+

.streamlit/config.toml

@@ -0,0 +1,9 @@
+[theme]
+primaryColor="#5b8cff"              # Blue accent (matches visualizer bars)
+backgroundColor="#0f1115"            # Dark background
+secondaryBackgroundColor="#151923"   # Panel background
+textColor="#f2f2f2"                 # Light text
+font="sans serif"
+
+[server]
+headless=true 

pyproject.toml

@@ -21,14 +21,14 @@ dependencies = [
   "livekit-agents[silero,turn-detector]>=1.3.3,<1.4",
   "livekit-plugins-noise-cancellation~=0.2",
   "livekit-api==1.1.0",
-  "livekit-plugins-langchain",
-  "langchain-nvidia-ai-endpoints",
-  "langgraph",
-  "pocket-tts",
-  "scipy",
-  "python-dotenv",
-  "pydantic-settings",
-  "streamlit",
+  "livekit-plugins-langchain>=1.3.3",
+  "langchain-nvidia-ai-endpoints>=1.0.3",
+  "langgraph>=1.0.8",
+  "pocket-tts>=1.0.3",
+  "scipy>=1.17.0",
+  "python-dotenv>=1.2.1",
+  "pydantic-settings>=2.12.0",
+  "streamlit>=1.54.0",
   "transformers @ git+https://github.com/huggingface/transformers.git@7769f660935b5d48b73bf6711d0a78b6f8f98739",
-
+  "pytest>=8.0.0",
 ]

src/agent/agent.py

@@ -34,7 +34,11 @@ async def session_handler(ctx: agents.JobContext) -> None:
             temperature=settings.voice.POCKET_TTS_TEMPERATURE,
             lsd_decode_steps=settings.voice.POCKET_TTS_LSD_DECODE_STEPS,
         ),
-        vad=silero.VAD.load(),
+        vad=silero.VAD.load(
+            min_speech_duration=settings.voice.VAD_MIN_SPEECH_DURATION,
+            min_silence_duration=settings.voice.VAD_MIN_SILENCE_DURATION,
+            activation_threshold=settings.voice.VAD_THRESHOLD,
+        ),
         turn_detection=MultilingualModel(),
     )
     await session.start(
@@ -42,6 +46,11 @@ async def session_handler(ctx: agents.JobContext) -> None:
         agent=Assistant(),
         room_options=room_io.RoomOptions(
             audio_input=room_io.AudioInputOptions(
+                sample_rate=settings.voice.LIVEKIT_SAMPLE_RATE,
+                num_channels=settings.voice.LIVEKIT_NUM_CHANNELS,
+                frame_size_ms=settings.voice.LIVEKIT_FRAME_SIZE_MS,
+                pre_connect_audio=settings.voice.LIVEKIT_PRE_CONNECT_AUDIO,
+                pre_connect_audio_timeout=settings.voice.LIVEKIT_PRE_CONNECT_TIMEOUT,
                 noise_cancellation=lambda params: noise_cancellation.BVCTelephony()
                 if params.participant.kind == rtc.ParticipantKind.PARTICIPANT_KIND_SIP
                 else noise_cancellation.BVC(),

src/core/settings.py

@@ -67,10 +67,56 @@ class VoiceSettings(CoreSettings):
         description="LSD decoding steps (higher = better quality, slower)",
     )
 
+    # LiveKit Audio Input Settings
+    LIVEKIT_SAMPLE_RATE: int = Field(
+        default=24000,
+        description="Audio input sample rate (Hz)",
+    )
+    LIVEKIT_NUM_CHANNELS: int = Field(
+        default=1,
+        description="Number of audio input channels (1=mono)",
+    )
+    LIVEKIT_FRAME_SIZE_MS: int = Field(
+        default=20,
+        ge=10,
+        le=100,
+        description="Audio frame size in milliseconds (smaller = faster VAD response)",
+    )
+    LIVEKIT_PRE_CONNECT_AUDIO: bool = Field(
+        default=True,
+        description="Pre-connect audio before room join",
+    )
+    LIVEKIT_PRE_CONNECT_TIMEOUT: float = Field(
+        default=3.0,
+        ge=1.0,
+        le=10.0,
+        description="Timeout for pre-connect audio (seconds)",
+    )
+
+    # Voice Activity Detection Settings
+    VAD_MIN_SPEECH_DURATION: float = Field(
+        default=0.25,
+        ge=0.1,
+        le=1.0,
+        description="Minimum speech duration (seconds) before VAD activation",
+    )
+    VAD_MIN_SILENCE_DURATION: float = Field(
+        default=0.5,
+        ge=0.1,
+        le=2.0,
+        description="Minimum silence duration (seconds) before VAD deactivation",
+    )
+    VAD_THRESHOLD: float = Field(
+        default=0.6,
+        ge=0.0,
+        le=1.0,
+        description="VAD activation threshold (higher = less sensitive, 0.5 is Silero default)",
+    )
+
 
 class LLMSettings(CoreSettings):
     NVIDIA_API_KEY: Optional[str] = Field(default=None)
-    NVIDIA_MODEL: str = Field(default="meta/llama-3.1-8b-instruct")
+    NVIDIA_MODEL: str = Field(default="openai/gpt-oss-20b") # "meta/llama-3.1-8b-instruct"
 
     LLM_TEMPERATURE: float = Field(default=0.7, ge=0.0, le=2.0)
     LLM_MAX_TOKENS: int = Field(default=1024, gt=0)
@@ -81,7 +127,7 @@ class LiveKitSettings(CoreSettings):
     LIVEKIT_API_KEY: Optional[str] = Field(default=None)
     LIVEKIT_API_SECRET: Optional[str] = Field(default=None)
     LIVEKIT_AGENT_NAME: str = Field(default="open-voice-agent")
-    LIVEKIT_NUM_IDLE_PROCESSES: int = Field(default=2, ge=0)
+    LIVEKIT_NUM_IDLE_PROCESSES: int = Field(default=3, ge=0)
 
 
 class Settings(CoreSettings):

src/streamlit_app.py

@@ -34,21 +34,28 @@ def main() -> None:
 
     if not settings.livekit.LIVEKIT_URL:
         st.error("LIVEKIT_URL is not set in the environment.")
-        return
+        st.stop()
     if not settings.livekit.LIVEKIT_API_KEY or not settings.livekit.LIVEKIT_API_SECRET:
         st.error("LIVEKIT_API_KEY or LIVEKIT_API_SECRET is not set.")
-        return
+        st.stop()
 
-    default_room = f"voice-{uuid4().hex[:8]}"
-    room_name = st.text_input("Room name", value=st.session_state.get("room_name", default_room))
-    st.session_state["room_name"] = room_name
+    # Auto-generate room name once
+    if "room_name" not in st.session_state:
+        st.session_state["room_name"] = f"voice-{uuid4().hex[:8]}"
 
-    start = st.button("Start session", type="primary")
-    if start or "token" not in st.session_state:
-        token_data = create_room_token(room_name=room_name)
-        st.session_state["token"] = token_data.token
-        st.session_state["agent_dispatched"] = False
+    room_name = st.session_state["room_name"]
 
+    # Auto-create token once
+    if "token" not in st.session_state:
+        try:
+            token_data = create_room_token(room_name=room_name)
+            st.session_state["token"] = token_data.token
+            st.session_state["agent_dispatched"] = False
+        except Exception as exc:
+            st.error(f"Failed to create room token: {exc}")
+            st.stop()
+
+    # Auto-dispatch agent once
     if not st.session_state.get("agent_dispatched"):
         try:
             dispatch_agent_sync(
@@ -58,14 +65,9 @@ def main() -> None:
             st.session_state["agent_dispatched"] = True
         except Exception as exc:
             st.error(f"Failed to dispatch agent: {exc}")
-            return
-
-    token = st.session_state.get("token")
-    if not token:
-        st.info("Click Start session to generate a LiveKit token.")
-        return
+            st.stop()
 
-    render_client(token=token, livekit_url=settings.livekit.LIVEKIT_URL)
+    render_client(token=st.session_state["token"], livekit_url=settings.livekit.LIVEKIT_URL)
 
 
 if __name__ == "__main__":

src/ui/index.html

@@ -6,7 +6,7 @@
       name="viewport"
       content="width=device-width, initial-scale=1, maximum-scale=1"
     />
-    <title>LiveKit Voice UI</title>
+    <title>Open Voice Agent</title>
     <style>
       :root {
         color-scheme: light dark;
@@ -17,22 +17,31 @@
         padding: 16px;
         background: #0f1115;
         color: #f2f2f2;
+        display: flex;
+        justify-content: center;
+        align-items: center;
+        min-height: 100vh;
       }
-      .panel {
-        border: 1px solid #2a2f3a;
-        border-radius: 12px;
-        padding: 16px;
-        background: #151923;
+      .container {
+        max-width: 800px;
+        width: 100%;
       }
       .status {
         margin-bottom: 12px;
         font-size: 14px;
         color: #b9c0d4;
       }
+      canvas {
+        width: 100%;
+        height: 140px;
+        background: #0f1115;
+        border-radius: 8px;
+        margin-bottom: 12px;
+      }
       .controls {
         display: flex;
         gap: 8px;
-        margin-top: 12px;
+        margin-bottom: 12px;
       }
       button {
         padding: 8px 12px;
@@ -46,12 +55,6 @@
         opacity: 0.6;
         cursor: not-allowed;
       }
-      canvas {
-        width: 100%;
-        height: 140px;
-        background: #0b0e14;
-        border-radius: 8px;
-      }
       .note {
         font-size: 12px;
         color: #8b93a7;
@@ -60,7 +63,7 @@
     </style>
   </head>
   <body>
-    <div class="panel">
+    <div class="container">
       <div id="status" class="status">Idle</div>
       <canvas id="wave" width="800" height="140"></canvas>
       <div class="controls">
@@ -69,7 +72,7 @@
         <button id="mute" disabled>Mute mic</button>
       </div>
       <div class="note">
-        Mic audio is streamed to LiveKit. Remote agent audio plays automatically.
+        Powered by open source models (Moonshine STT, NVIDIA LLM, Pocket TTS) running on LiveKit infrastructure.
       </div>
       <audio id="remote-audio" autoplay></audio>
     </div>

src/ui/main.js

@@ -43,8 +43,10 @@ function drawWave() {
     const value = dataArray[i * step] || 0;
     const normalized = value / 255;
     const barHeight = Math.max(6, normalized * canvas.height);
+    const halfHeight = barHeight / 2;
+    const centerY = canvas.height / 2;
+    const y = centerY - halfHeight;
     const x = i * barWidth;
-    const y = canvas.height - barHeight;
     ctx.fillStyle = "#5b8cff";
     ctx.fillRect(x + 2, y, barWidth - 4, barHeight);
   }

uv.lock

@@ -549,6 +549,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/fa/5e/f8e9a1d23b9c20a551a8a02ea3637b4642e22c2626e3a13a9a29cdea99eb/importlib_metadata-8.7.1-py3-none-any.whl", hash = "sha256:5a1f80bf1daa489495071efbb095d75a634cf28a8bc299581244063b53176151", size = 27865, upload-time = "2025-12-21T10:00:18.329Z" },
 ]
 
+[[package]]
+name = "iniconfig"
+version = "2.3.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/72/34/14ca021ce8e5dfedc35312d08ba8bf51fdd999c576889fc2c24cb97f4f10/iniconfig-2.3.0.tar.gz", hash = "sha256:c76315c77db068650d49c5b56314774a7804df16fee4402c1f19d6d15d8c4730", size = 20503, upload-time = "2025-10-18T21:55:43.219Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/cb/b1/3846dd7f199d53cb17f49cba7e651e9ce294d8497c8c150530ed11865bb8/iniconfig-2.3.0-py3-none-any.whl", hash = "sha256:f631c04d2c48c52b84d0d0549c99ff3859c98df65b3101406327ecc7d53fbf12", size = 7484, upload-time = "2025-10-18T21:55:41.639Z" },
+]
+
 [[package]]
 name = "jinja2"
 version = "3.1.6"
@@ -1201,6 +1210,7 @@ dependencies = [
     { name = "opentelemetry-sdk" },
     { name = "pocket-tts" },
     { name = "pydantic-settings" },
+    { name = "pytest" },
     { name = "python-dotenv" },
     { name = "scipy" },
     { name = "streamlit" },
@@ -1209,20 +1219,21 @@ dependencies = [
 
 [package.metadata]
 requires-dist = [
-    { name = "langchain-nvidia-ai-endpoints" },
-    { name = "langgraph" },
+    { name = "langchain-nvidia-ai-endpoints", specifier = ">=1.0.3" },
+    { name = "langgraph", specifier = ">=1.0.8" },
     { name = "livekit-agents", extras = ["silero", "turn-detector"], specifier = ">=1.3.3,<1.4" },
     { name = "livekit-api", specifier = "==1.1.0" },
-    { name = "livekit-plugins-langchain" },
+    { name = "livekit-plugins-langchain", specifier = ">=1.3.3" },
     { name = "livekit-plugins-noise-cancellation", specifier = "~=0.2" },
     { name = "opentelemetry-api", specifier = ">=1.34.0,<1.35.0" },
     { name = "opentelemetry-exporter-otlp", specifier = ">=1.34.0,<1.35.0" },
     { name = "opentelemetry-sdk", specifier = ">=1.34.0,<1.35.0" },
-    { name = "pocket-tts" },
-    { name = "pydantic-settings" },
-    { name = "python-dotenv" },
-    { name = "scipy" },
-    { name = "streamlit" },
+    { name = "pocket-tts", specifier = ">=1.0.3" },
+    { name = "pydantic-settings", specifier = ">=2.12.0" },
+    { name = "pytest", specifier = ">=8.0.0" },
+    { name = "python-dotenv", specifier = ">=1.2.1" },
+    { name = "scipy", specifier = ">=1.17.0" },
+    { name = "streamlit", specifier = ">=1.54.0" },
     { name = "transformers", git = "https://github.com/huggingface/transformers.git?rev=7769f660935b5d48b73bf6711d0a78b6f8f98739" },
 ]
 
@@ -1447,6 +1458,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/5d/cf/881b457eccacac9e5b2ddd97d5071fb6d668307c57cbf4e3b5278e06e536/pillow-12.1.0-cp312-cp312-win_arm64.whl", hash = "sha256:65b80c1ee7e14a87d6a068dd3b0aea268ffcabfe0498d38661b00c5b4b22e74c", size = 2452612, upload-time = "2026-01-02T09:11:29.309Z" },
 ]
 
+[[package]]
+name = "pluggy"
+version = "1.6.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/f9/e2/3e91f31a7d2b083fe6ef3fa267035b518369d9511ffab804f839851d2779/pluggy-1.6.0.tar.gz", hash = "sha256:7dcc130b76258d33b90f61b658791dede3486c3e6bfb003ee5c9bfb396dd22f3", size = 69412, upload-time = "2025-05-15T12:30:07.975Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/54/20/4d324d65cc6d9205fabedc306948156824eb9f0ee1633355a8f7ec5c66bf/pluggy-1.6.0-py3-none-any.whl", hash = "sha256:e920276dd6813095e9377c0bc5566d94c932c33b27a3e3945d8389c374dd4746", size = 20538, upload-time = "2025-05-15T12:30:06.134Z" },
+]
+
 [[package]]
 name = "pocket-tts"
 version = "1.0.3"
@@ -1658,6 +1678,22 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/5a/dc/491b7661614ab97483abf2056be1deee4dc2490ecbf7bff9ab5cdbac86e1/pyreadline3-3.5.4-py3-none-any.whl", hash = "sha256:eaf8e6cc3c49bcccf145fc6067ba8643d1df34d604a1ec0eccbf7a18e6d3fae6", size = 83178, upload-time = "2024-09-19T02:40:08.598Z" },
 ]
 
+[[package]]
+name = "pytest"
+version = "9.0.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+    { name = "iniconfig" },
+    { name = "packaging" },
+    { name = "pluggy" },
+    { name = "pygments" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/d1/db/7ef3487e0fb0049ddb5ce41d3a49c235bf9ad299b6a25d5780a89f19230f/pytest-9.0.2.tar.gz", hash = "sha256:75186651a92bd89611d1d9fc20f0b4345fd827c41ccd5c299a868a05d70edf11", size = 1568901, upload-time = "2025-12-06T21:30:51.014Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/3b/ab/b3226f0bd7cdcf710fbede2b3548584366da3b19b5021e74f5bde2a8fa3f/pytest-9.0.2-py3-none-any.whl", hash = "sha256:711ffd45bf766d5264d487b917733b453d917afd2b0ad65223959f59089f875b", size = 374801, upload-time = "2025-12-06T21:30:49.154Z" },
+]
+
 [[package]]
 name = "python-dateutil"
 version = "2.9.0.post0"