Show HN: ML condenses billions of logs into a tiny snapshot your LLM can debug
TL;DR
Rocketgraph is an open-source, self-hosted log analysis tool. It sits beside existing observability stacks such as Datadog, Loki, CloudWatch, Sentry or ClickHouse and pulls log windows instead of creating a second ingest pipeline. Its ML layer condenses raw logs into structural templates, scores anomalies and can evaluate new log lines in real time against a trained model. The project says it uses Drain3, Isolation Forest and Half-Space-Trees, not an LLM directly.
Nauti's Take
The interesting part is not the AI triage wrapper, but the compression step before it. If the ML layer turns messy logs into reproducible patterns, an LLM finally gets a usable debugging window instead of millions of noisy lines.
Still, the pitch is PR-heavy: the big claims need proof in messy, real production systems.
Briefingshow
Observability is still largely built for humans: dashboards, alerts and manual queries. As more code is written and debugged by AI, debugging context needs to become easier for machines to inspect. Rocketgraph targets that gap by turning raw logs into stable patterns so an LLM is not asked to search through noise.
Sources