ContextSpatial
ContextSpatial will expand the ContextUnity agent pipeline into the voice domain, providing real-time audio interactions through WebRTC and native voice LLM APIs.
Planned Features
LiveKit Telephony
Use LiveKit as a Session Forwarding Unit (SFU) to stream browser/mobile microphone audio directly to active agent sessions without polling overhead.
Native Voice LLMs
Direct binding to OpenAI Realtime API and Gemini Live Bidi-streams — bypassing the traditional STT → LLM → TTS stack for significantly lower latency.
Semantic Interrupts
Voice agents maintain active conversational state with ContextRouter. During a voice session, the agent can fire tool operations — looking up memory from ContextBrain or executing database queries — mid-conversation.
Google ADK Bidi-streaming
Native support for Gemini Live API with multimodal data ingestion over WebSocket.
Architecture Concept
