What problem does it solve? Investigating latency spikes, error rates, and resource saturation requires writing correct PromQL and running it against a Prometheus server, which is error-prone without guidance on metric types, label filtering, and query structure. ## Core Features & Use Cases - Instant and Range Queries: Run PromQL queries against Prometheus, Thanos, Cortex, VictoriaMetrics, or Grafana Mimir with table, CSV, JSON, or ASCII graph output. - Metric Discovery: List metric names, labels, and metadata with the metrics, labels, and meta subcommands before writing queries. - Debugging Methodologies: Apply USE (Utilization, Saturation, Errors) and RED (Rate, Errors, Duration) methods to isolate slow instances and diagnose error spikes. - Use Case: A pod in your Kubernetes cluster is slower than others. Isolate it with a per-instance histogram_quantile query, graph p99 latency over the last two hours, then correlate with CPU and database metrics on that host. ## Quick Start Ask the assistant to graph the HTTP request rate for your API over the last hour using promql-cli.
