Dashboards

vLLM 추론 서버 모니터링

Ollama dashboard for the ollama-exporter written in go (https://github.com/maravexa/ollama-exporter)

Claude Code usage analytics on Prometheus: cost, token consumption, sessions, active coding time, lines of code, commits, pull requests, tool decisions, cache hit ratio, and leaderboards by user, model, and session. Consumes Anthropic's OpenTelemetry metrics via OTLP. Works with Prometheus, VictoriaMetrics, Mimir, and Thanos. Requires Grafana 11+.

Comprehensive monitoring dashboard for Claude Code CLI usage. Track sessions, tokens, costs, commits, pull requests, lines of code, and developer productivity metrics. Compatible with Prometheus, VictoriaMetrics, Mimir, and Thanos. Metrics required: - claude_code_session_count_total - claude_code_token_usage_tokens_total - claude_code_cost_usage_USD_total - claude_code_active_time_seconds_total - claude_code_lines_of_code_count_total - claude_code_commit_count_total - claude_code_pull_

基于原版 #24755 的完整中文汉化版本。覆盖 vLLM 推理核心指标：调度器效率、KV Cache 使用率、TTFT/TPOT 延迟、Prefix Cache 命中率、请求完成原因分布等。已适配 Grafana 10.4+，开箱即用。

Pulze.ai Dashboard around LLM application performance, costs, and usage associated with various providers and models.

同时支持 SGLang 和 vLLM 的统一 LLM 推理监控 Grafana Dashboard，数据源为 Prometheus。该模板将两类推理后端的可比指标放在同一视图中展示，覆盖请求量、Token 吞吐、延迟分位数、队列状态、缓存行为以及 Engine 或 TP Rank 维度的分布情况。适用于混合推理集群、后端迁移、压测对比和统一运维场景，帮助团队用一致的视角观察不同推理框架的服务表现。

vLLM 추론 서버 모니터링 대시보드

llama.cpp LLM server metrics

Ollama LLM runtime metrics — models, VRAM, GPU telemetry, process stats. Requires Neurix ollama_exporter.

GPU monitoring for self-hosted AI: are your tensor cores actually working, what does each run cost in kWh, and is your VRAM degrading? Built and battle-tested on a local LLM rig (100B+ parameter models). Requires dcgm-exporter with DCP/profiling metrics enabled. Multi-GPU via UUID variable.

LiteLLM proxy panels Fully compatible with LiteLLM 1.92.0

This is a dashboard for Langraph API System metrics

vLLM serving overview: first-token and inter-token latency, throughput, queue depth, KV cache, prefix-cache reuse and a TTFT error budget, every panel by (model_name). V1 metric names; several panels need the llm:* recording rules.

Usage, cost, and quality metrics for Open WebUI — users, chats, tokens, spend by model and by user, and per-model satisfaction. Requires the Open WebUI Prometheus exporter (polls the REST API, no database access).

Metrics for Local AI

GodAgent v2 — agentic pipeline observability. Tracks LLM costs, token usage, Motor Memory hit rate, circuit breakers, pipeline CLEAR score.