vLLM Diagnostic Tool vllm-doctor

·2026.06.08 18:10

Key point

It's a CLI tool that analyzes vLLM server metrics to diagnose queue pressure, TTFT, KV cache status, and more.

vllm-doctor is a CLI tool that reads metrics from a vLLM server's /metrics endpoint or a Prometheus instance and performs rule-based checks.

Its main features are as follows:

Detects key indicators such as Queue pressure, TTFT/TPOT, and KV cache pressure
When an issue occurs, provides the relevant indicator, confidence level, likely cause, and specific recommendations
Supports text or JSON format output and offers a --watch mode for real-time monitoring

It is currently an early-stage project available as open source.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.