agents
agents copied to clipboard
livekit
Build real-time multimodal AI applications 🤖🎙️📹
I'm trying to use Google Chirp voices for TTS, this is my agent configuration - ``` class SimpleAgent(Agent): def __init__(self) -> None: super().__init__( instructions=""" You are a translator. You translate...
On version 1.0.21 When I set the STT in the agent session azure STT does not work, however deepgram works ```python session = AgentSession[IntakeSessionState]( allow_interruptions=True, vad=ctx.proc.userdata["vad"], turn_detection=EnglishModel(), stt=azure.STT(), #stt=deepgram.STT(), llm=openai.LLM(model="gpt-4o"),...
# Add support for input_audio_noise_reduction in RealtimeModel ## Issue Resolves #2312 ## Summary OpenAI Realtime API supports `input_audio_noise_reduction` session property with `near_field` or `far_field` modes, but LiveKit's RealtimeModel was missing...
**Objective** I’m planning to implement end-to-end distributed tracing across the voice pipeline (STT → LLM → TTS → LiveKit) using OpenTelemetry and Jaeger. Question Is there any built-in support or...
Ergess is nice, it provides many features and possibilities. But there can be more simple and more resource thoughtful way of recording audio - recording right inside the agent. Agent...
### Summary The OpenAI Realtime API accepts an `input_audio_noise_reduction` session property (with `near_field` or `far_field` modes), exposed by the Python SDK through the `InputAudioNoiseReduction` helper type.[¹](https://github.com/openai/openai-python/blob/main/src/openai/types/beta/realtime/session_create_params.py?utm_source=chatgpt.com) LiveKit’s **RealtimeModel** currently omits...
When decorating a Agent method with a complex parameter my program crashes at the [line 162 in `livekit/agents/llm/_strict.py`](https://github.com/livekit/agents/blob/e57f6a0a4cf07190ca4b60aff6acdadc2cebce90/livekit-agents/livekit/agents/llm/_strict.py#L162): ```log t = non_null["type"] ~~~~~~~~^^^^^^^^ KeyError: 'type' ``` A toy example: ```py...
Introduce the Whisper plugin for LiveKit, enabling offline speech-to-text capabilities using local Whisper model inference. This includes: - Initial setup of the plugin structure with classes for Whisper model, speech-to-text...
to be reviewed: - dentist scheduler: a multi-agent example offering different functionalities integrated via Cal.com and Supabase APIs - conversation persistor (realtime and pipeline): an updated version for 1.0 events...
When using the `raw_schema` parameter of the `function_tool` decorator, the `prepare_function_arguments` method fails. This is because it is unable to map the `raw_fields` dict to the signature of the method...
