logcatx

Local web dashboard for offline embedded dual-core log analysis. Feed it one log file, get an interactive browser dashboard. Deterministic analysis only — no LLM, no external log-mining dependency, no database, single-process, in-memory session.

Install & run

pip install logcatx
logcatx path/to/logfile.log          # parses, serves, opens browser
logcatx file.log --port 9000         # explicit port
logcatx file.log --no-browser        # headless start

The line format is fixed for v1: [MM/DD HH:MM:SS] [tid] [LEVEL] [core] [module] message. Lines that do not match are kept and shown as unparsed — never dropped.

Views

1. Overview — file info, level pie, top modules, top error templates.
2. Templates — searchable/sortable mined-template table; click a row to drill into its raw lines.
3. Timeline — bucketed stacked bars by level (or by top error templates), zoomable; drag-brush a range to jump to Raw Logs in that window; burst list marks templates whose count jumps beyond 5σ (leave-one-out).
4. Raw Logs — filter by level/core/module/keyword/template/time range, paged; click any line to expand ±5 lines of context (unparsed neighbors included).

A demo log ships at src/logcatx/static/synth-demo.log (16k synthetic lines with an injected error storm).

Manual verification checklist

• logcatx src/logcatx/static/synth-demo.log — Overview numbers match (16601 total, 600 ERROR).
• Templates: search stat wrong finds the storm template (300 hits); row click filters Raw Logs to it.
• Timeline: ERROR spike at 09:20 visible; brush it → Raw Logs scoped to that minute.
• Raw Logs: scroll 16k lines stays smooth (200/page paging); click a line → context block opens.
• Port fallback: start a second instance with the same --port; it binds port+1 automatically instead of failing.

Development

pip install -e '.[dev]'
pytest -q                                  # unit + e2e suite
ruff check src tests main.py               # lint
python tools/gen_large.py 1000000 /tmp/big.log   # perf fixture

Performance: 1,000,000 synthetic lines parse+mine in ~29s (single core).

License

Apache-2.0 — see LICENSE.