Making AI serving profitable starts with knowing where you're losing. Profile measures your hardware ceiling against live telemetry, finds the exact gap, and gives you the fix, one change at a time.123
Value in a minute
download
curl -L https://github.com/jungledesh/profile/releases/latest/download/profile -o profile
install
chmod +x profile && mv profile /usr/local/bin/
run
profile diagnose --url http://localhost:8000/metrics --duration 2m
Requirements:
— vLLM instance exporting standard Prometheus metrics
— NVIDIA Linux runtime with native NVML (
— vLLM instance exporting standard Prometheus metrics
— NVIDIA Linux runtime with native NVML (
libnvidia-ml.so) access
Less Words. Less Noise. More Signal. More Value.
Demo video coming soon