{"asOf":"2026-10-28","storesIpAddresses":false,"archetype":"voice-support-agent","requestedArchetype":"voice-support-agent","budgetAppliedUsd":null,"title":"Sub-500ms Real-Time Voice & Telephony Agent Stack","summary":"Low-latency WebRTC/SIP voice pipeline pairing streaming STT, low-TTFT LLM routing, expressive TTS, and multi-turn persona simulation.","estimatedMonthlyUsd":{"hobby":10,"startup":140,"scale":1120},"oneLineInstallCmd":"npm install livekit @deepgram/sdk groq-sdk @elevenlabs/elevenlabs-js braintrust","requiredEnvVars":["LIVEKIT_API_KEY","DEEPGRAM_API_KEY","GROQ_API_KEY","ELEVENLABS_API_KEY","BRAINTRUST_API_KEY"],"wiringNotes":"LiveKit Agents SDK orchestrates Deepgram STT -> Groq LLM -> ElevenLabs/Cartesia TTS over persistent WebSockets.","slots":[{"role":"WebRTC & SIP Voice Transport","jtbd":"voice-transport","providerId":"livekit","providerName":"LiveKit Cloud","editorialOverall":8.48,"activeSku":"livekit-standard","envVars":["LIVEKIT_API_KEY"],"smokeTestCmd":"test -n \"$LIVEKIT_API_KEY\" && echo \"LIVEKIT_API_KEY configured\"","day2Gotcha":"You still compose STT/LLM/TTS (more assembly than Vapi)","whyItFits":"Sub-100ms global WebRTC mesh with native turn-detection and interruption handling.","decideUrl":"/api/decide?jtbd=voice-transport"},{"role":"Streaming Speech-to-Text (STT)","jtbd":"speech-to-text","providerId":"deepgram","providerName":"Deepgram","editorialOverall":8.5,"activeSku":"deepgram-nova-3-mono-prerecorded","envVars":["DEEPGRAM_API_KEY"],"smokeTestCmd":"curl -sS https://api.deepgram.com/v1/projects -H \"Authorization: Token $DEEPGRAM_API_KEY\"","day2Gotcha":"TTS less 'celebrity voice' than ElevenLabs","whyItFits":"Nova-3 streaming transcription under 250ms p95 latency.","decideUrl":"/api/decide?jtbd=speech-to-text"},{"role":"Low-TTFT Conversation LLM","jtbd":"fast-inference","providerId":"groq","providerName":"Groq","editorialOverall":7.95,"activeSku":"openai/gpt-oss-120b","envVars":["GROQ_API_KEY"],"smokeTestCmd":"curl -sS https://api.groq.com/openai/v1/models -H \"Authorization: Bearer $GROQ_API_KEY\"","day2Gotcha":"Limited model variety","whyItFits":"Sub-150ms time-to-first-token keeps total voice turn latency under 500ms.","decideUrl":"/api/decide?jtbd=ultra-low-latency"},{"role":"Streaming Text-to-Speech (TTS)","jtbd":"text-to-speech","providerId":"elevenlabs","providerName":"ElevenLabs","editorialOverall":8.27,"activeSku":"elevenlabs-flash-v2-5","envVars":["ELEVENLABS_API_KEY"],"smokeTestCmd":"curl -sS https://api.elevenlabs.io/v1/user -H \"xi-api-key: $ELEVENLABS_API_KEY\"","day2Gotcha":"Voice cloning ethics/ToS care required","whyItFits":"Expressive low-latency streaming audio with WebSocket chunking.","decideUrl":"/api/decide?jtbd=speech-synthesis"},{"role":"Voice & Persona Eval Harness","jtbd":"multi-turn-user-simulation-evals","providerId":"braintrust","providerName":"Braintrust","editorialOverall":8.23,"activeSku":"braintrust-pro","envVars":["BRAINTRUST_API_KEY"],"smokeTestCmd":"curl -sS https://api.braintrust.dev/v1/project -H \"Authorization: Bearer $BRAINTRUST_API_KEY\"","day2Gotcha":"Broader tracing competitors exist","whyItFits":"Automated adversarial caller simulation before production deploy.","decideUrl":"/api/decide?jtbd=multi-turn-user-simulation-evals"}],"cite":"Agent Provider Depot Blueprint (voice-support-agent, as of 2026-10-28). https://getagentdepot.com/blueprints"}