I built Metrik to answer a simple question for AI voice agents: which LLM is fastest right now?
If you use Vapi with multiple providers (OpenAI, Anthropic, Google, etc.), it’s hard to:
– Measure Time to First Token (TTFT) consistently
– See when a provider slows down
– Route calls to the fastest model without manual changes
What Metrik does
– Continuously pings multiple LLMs and logs TTFT/latency
– Shows a live dashboard by provider/model
– Can route Vapi agents to the currently fastest allowed model
– Exposes a free API so you can pull the metrics into your own tools
I’d love feedback on what other metrics you care about (besides TTFT), and whether you’d want this as a hosted service, self-hosted, or just a library.
mbouassa•14m ago
If you use Vapi with multiple providers (OpenAI, Anthropic, Google, etc.), it’s hard to: – Measure Time to First Token (TTFT) consistently – See when a provider slows down – Route calls to the fastest model without manual changes
What Metrik does – Continuously pings multiple LLMs and logs TTFT/latency – Shows a live dashboard by provider/model – Can route Vapi agents to the currently fastest allowed model – Exposes a free API so you can pull the metrics into your own tools
I’d love feedback on what other metrics you care about (besides TTFT), and whether you’d want this as a hosted service, self-hosted, or just a library.