What to build
A voice agent example that profiles and visualizes end-to-end latency by measuring Time-to-First-Byte (TTFB) for each component in the pipeline: STT recognition time, LLM response time, and TTS generation time — displayed in a real-time dashboard.
Why this matters
Voice agent developers need to understand where latency originates in their STT → LLM → TTS pipeline to optimize the conversation experience. Deepgram's STT and TTS are among the fastest in the industry (247ms median TTFB on independent benchmarks), but developers have no easy way to measure and visualize this advantage. A latency profiling example helps developers optimize their pipeline and makes Deepgram's speed advantage tangible and measurable.
Suggested scope
- Python backend: Deepgram STT + Voice Agent API with per-component timing instrumentation
- TypeScript frontend: React dashboard showing per-turn latency waterfall chart
- Measure: STT TTFB, LLM TTFB, TTS TTFB, and total round-trip time
- Display: real-time waterfall chart, rolling average, percentile distribution
- Deepgram APIs: streaming STT and Aura TTS
Acceptance criteria
Raised by the DX intelligence system.
What to build
A voice agent example that profiles and visualizes end-to-end latency by measuring Time-to-First-Byte (TTFB) for each component in the pipeline: STT recognition time, LLM response time, and TTS generation time — displayed in a real-time dashboard.
Why this matters
Voice agent developers need to understand where latency originates in their STT → LLM → TTS pipeline to optimize the conversation experience. Deepgram's STT and TTS are among the fastest in the industry (247ms median TTFB on independent benchmarks), but developers have no easy way to measure and visualize this advantage. A latency profiling example helps developers optimize their pipeline and makes Deepgram's speed advantage tangible and measurable.
Suggested scope
Acceptance criteria
Raised by the DX intelligence system.