feat(mistralai): add streaming STT support for Voxtral Realtime

examples/voice_agents/README.md

@@ -23,6 +23,7 @@ session = AgentSession(
 ### 🚀 Getting Started
 
 - [`basic_agent.py`](./basic_agent.py) - A fundamental voice agent using LiveKit Inference with metrics collection
+- [`mistral_realtime_stt_agent.py`](./mistral_realtime_stt_agent.py) - Comparison agent using Mistral realtime STT, OpenAI LLM, and ElevenLabs Flash v2.5 TTS
 
 ### 🛠️ Tool Integration & Function Calling
 

examples/voice_agents/mistral_realtime_stt_agent.py

@@ -0,0 +1,106 @@
+"""Comparison voice agent using Mistral STT and ElevenLabs TTS.
+
+Requires:
+    MISTRAL_API_KEY for Mistral STT
+    OPENAI_API_KEY for the LLM
+    ELEVEN_API_KEY for ElevenLabs TTS
+
+This example uses Mistral's realtime streaming STT model, OpenAI for the LLM,
+and ElevenLabs Flash v2.5 for TTS using an explicit voice ID already referenced
+elsewhere in the repo. The Mistral STT settings are tuned to the best plugin-only
+configuration we observed in comparison runs against the Deepgram example.
+
+For the cleanest console-mode testing, use headphones to avoid the agent hearing
+its own playback through the microphone.
+"""
+
+import logging
+
+from dotenv import load_dotenv
+
+from livekit.agents import (
+    Agent,
+    AgentServer,
+    AgentSession,
+    JobContext,
+    JobProcess,
+    TurnHandlingOptions,
+    cli,
+    room_io,
+)
+from livekit.plugins import elevenlabs, mistralai, openai, silero
+from livekit.plugins.turn_detector.multilingual import MultilingualModel
+
+logger = logging.getLogger("mistral-realtime-stt-agent")
+
+load_dotenv()
+
+
+class MyAgent(Agent):
+    def __init__(self) -> None:
+        super().__init__(
+            instructions=(
+                "Your name is Kelly, built by LiveKit. "
+                "Keep responses concise, natural, and voice-friendly. "
+                "Reply in one or two short sentences."
+            )
+        )
+
+    async def on_enter(self) -> None:
+        self.session.generate_reply(
+            instructions=(
+                "Greet the user and mention that you are using Mistral for speech-to-text "
+                "and ElevenLabs for text-to-speech. Keep the greeting short."
+            )
+        )
+
+
+server = AgentServer()
+
+
+def prewarm(proc: JobProcess) -> None:
+    proc.userdata["vad"] = silero.VAD.load()
+
+
+server.setup_fnc = prewarm
+
+
+@server.rtc_session()
+async def entrypoint(ctx: JobContext) -> None:
+    ctx.log_context_fields = {"room": ctx.room.name}
+
+    session = AgentSession(
+        stt=mistralai.STT(
+            model="voxtral-mini-transcribe-realtime-2602",
+            target_streaming_delay_ms=240,
+            chunk_duration_ms=10,
+            finalize_delay_ms=650,
+        ),
+        llm=openai.LLM(model="gpt-4.1-mini"),
+        tts=elevenlabs.TTS(
+            model="eleven_flash_v2_5",
+            voice_id="hpp4J3VqNfWAUOO0d1Us",
+        ),
+        vad=ctx.proc.userdata["vad"],
+        turn_handling=TurnHandlingOptions(
+            turn_detection=MultilingualModel(),
+            interruption={
+                "resume_false_interruption": True,
+                "false_interruption_timeout": 1.0,
+            },
+        ),
+        preemptive_generation=False,
+        aec_warmup_duration=3.0,
+    )
+
+    await session.start(
+        agent=MyAgent(),
+        room=ctx.room,
+        room_options=room_io.RoomOptions(
+            audio_input=room_io.AudioInputOptions(),
+        ),
+    )
+
+
+if __name__ == "__main__":
+    cli.run_app(server)

livekit-plugins/livekit-plugins-mistralai/livekit/plugins/mistralai/__init__.py

@@ -4,11 +4,11 @@
 
 from .llm import LLM
 from .log import logger
-from .stt import STT
+from .stt import STT, SpeechStream
 from .tts import TTS
 from .version import __version__
 
-__all__ = ["LLM", "STT", "TTS", "__version__"]
+__all__ = ["LLM", "STT", "SpeechStream", "TTS", "__version__"]
 
 
 class MistralAIPlugin(Plugin):

livekit-plugins/livekit-plugins-mistralai/livekit/plugins/mistralai/models.py

@@ -17,7 +17,11 @@
 ]
 
 STTModels = Literal[
-    "voxtral-small-2507", "voxtral-mini-2507", "voxtral-mini-latest", "voxtral-small-latest"
+    "voxtral-small-2507",
+    "voxtral-mini-2507",
+    "voxtral-mini-latest",
+    "voxtral-small-latest",
+    "voxtral-mini-transcribe-realtime-2602",
 ]
 
 TTSModels = Literal["voxtral-mini-tts-2603", "voxtral-mini-tts-latest"]