---
name: log-digest
description: "Group repeated log events into a compact timeline with honest counts. Use for Operational log summaries, Batch-job diagnostics."
license: MIT
---

Condense supplied logs into a digest of distinct events and their sequence. Use this for many timestamped records; use an exception-chain method when one stack trace is the main evidence.

Group exact or meaningfully equivalent repeats by component, event, and correlation identifier. Keep first and last occurrence, count, severity, and changing values when the input supports them. Preserve transitions such as retry, recovery, and a new failure mode. Retain representative messages and identifiers exactly where they are needed to locate the event.

Remove routine success noise unless it establishes a boundary or recovery. Do not treat a sampled excerpt as the complete system history. Distinguish an explicit event count from a count inferred only from the supplied lines. Preserve timezones and ordering ambiguity.

Output a short timeline or grouped event list, followed by unresolved observations if supplied. Do not infer root cause from proximity, merge different request IDs, fabricate missing timestamps, or claim recovery without a success signal.