Nimbus Voice Agent · Playground

Configuration

LLM model Knowledge mode

Reranker Top-k chunks: 4

Response length Verbatim history turns: 3 Enable tools Streaming mode (off = batch)

🎙 Voice

Speech-to-text (ASR) Text-to-speech (TTS) Voice (OpenAI) Buffer audio before play Barge-in (interrupt on speech) Endpoint silence: 700 ms

Tool access (10/10)

System prompt (optional)

—end-to-end ms (last turn)

	Streaming	Batch
Total ms	—	—
Time to first token	—	n/a

ASR / TTS timings show here.

RAG chunks show here.

Tool calls & their latency show here.

Empty.