victorialogs-analysis

What problem does it solve? Investigating incidents by dumping raw logs floods the context window and hides the real signal. This Skill enforces a statistics-first workflow for VictoriaLogs so you get error rates, top streams, and error patterns before ever looking at individual log lines. ## Core Features & Use Cases - Statistics-First Investigation: The get_statistics.py script returns total volume, error rate, logs per minute, top streams, and normalized error patterns as the mandatory first step. - Strategic Log Sampling: sample_logs.py supports errors_only, warnings_up, around_time, and all strategies, hard-capped at 50 entries to protect context. - Field Discovery and Raw LogsQL: list_fields.py reveals available fields and values, while query_logs.py executes arbitrary LogsQL with automatic safety limits. - Use Case: During a production incident, run get_statistics.py to see a 12% error rate concentrated in one stream, then sample errors_only logs around the spike timestamp to identify the failing pattern without reading thousands of lines. ## Quick Start Ask the agent to investigate recent errors in VictoriaLogs by starting with log statistics for the last hour.