Kibana-style Discover view for local JSON-lines (LDJSON / NDJSON / JSONL) log files, right inside VS Code.
KibanaLocal is an independent project. It is not affiliated with, endorsed by, or sponsored by Elastic N.V. or Elasticsearch B.V. Kibana and Elasticsearch are trademarks of Elasticsearch B.V.
Features
Indexes: add a log file, or a folder (every .log/.jsonl/.ndjson/.ldjson/.json/.txt/.out inside becomes part of that index). Query one, several (Ctrl/Cmd-click), or all.
Live tailing: byte-offset tailing; only new bytes are read. Handles truncation/rotation and partial lines.
Refresh modes: ● Live (real-time) · ⟳ Every N s · ❚❚ Manual (shows a +N badge for new lines).
message:"gateway timeout" (phrase) · level:(ERROR or WARN)
latency > 500 · @timestamp >= 2026-11-13T10:00
and / or / not / parentheses; bare words = free text over all values
Histogram (click a bar to zoom), time ranges, field sidebar with top values and ⊕/⊖ filters, column picker, expandable doc view (Table / JSON).
Auto-detects timestamp / level / message fields (override in settings). Non-JSON lines are kept; indented lines (stack traces) attach to the previous entry.
Mixed schemas: timestamp/level/message are detected per index, and the default level / message columns resolve per row — so level vs severity, message vs text line up in one view.
Big files: the engine runs in a worker thread, live updates are incremental (only new lines are matched), and the results grid is virtualized with on-demand paging.
Rotation-safe: follows logrotate (rename + create) via inode, detects in-place truncation and copytruncate.
Keyboard
/ focus search · ↑/↓PgUp/PgDnHome/End move through results · Enter expand/collapse · Tab reaches every control.
Usage
Click the KibanaLocal icon in the Activity Bar, or Command Palette → KibanaLocal: Open, then + Index and paste a path. Or right-click a log file/folder → Open in KibanaLocal. Press / to focus search.
Settings
Setting
Default
logscope.maxRows
200000
lines kept in memory per index (oldest by timestamp dropped)
logscope.maxMemoryMB
512
memory budget shared by all indexes (oldest by timestamp dropped); retained data is ~4× the raw file size
logscope.initialTailMB
50
only the last N MB of a large file are loaded at first
