Vllm

Learn how to monitor LLM inference servers (vLLM, Hugging Face TGI, llama.cpp) using Prometheus and Grafana. This article covers what to measure, how to expose and scrape /metrics, example PromQL queries for p95 latency and tokens/sec, ready-to-use Docker Compose and Kubernetes manifests, Grafana dashboard provisioning, alerting, and real-world troubleshooting.

Strategic guide to hosting large language models locally with Ollama, llama.cpp, vLLM, or in the cloud. Compare tools, performance trade-offs, and cost considerations.

Запуск локальных языковых моделей (LLM) теперь практичен для разработчиков, стартапов и даже корпоративных команд. Но выбор правильного инструмента — Ollama, vLLM, LM Studio, LocalAI или других — зависит от ваших целей: