logdelve 0.3.0

A CLI/TUI tool for viewing and filtering log lines

logdelve

A terminal UI tool for viewing, filtering, and analyzing log lines. Built for outage investigation — find the needle in the haystack across thousands of log lines from multi-component applications.

Features

• Log level detection: Automatic extraction from JSON fields and text patterns, color-coded line backgrounds
• Multiple file input: Merge multiple log files chronologically with per-file component tagging
• Component detection: Kubernetes pods, Docker Compose services, JSON fields — with color-coded tags
• Anomaly detection: Baseline comparison to find log patterns that are new or changed (--baseline)
• Message analysis: Group log messages by event pattern, analyze JSON field value distributions
• Search: Multi-pattern search with up to 10 simultaneous patterns, each highlighted in a distinct color — with per-pattern n/N navigation control
• Interactive filtering: Filter by text, regex, JSON key-value, component, or log level — tabbed dialog with multi-select
• Filter management: Reorder, toggle, edit, delete, suspend/resume all filters with cursor preservation
• Bookmarks & Annotations: Mark lines, attach notes, navigate between bookmarks, persisted in sessions
• Sessions: Save, load, rename, delete named sessions — persists filters, bookmarks, and search patterns with auto-save
• Streaming large files: Chunked background loading for files up to 2-3GB with instant startup and progress display
• Live tailing: Follow growing log files in real-time with pause/resume
• Flexible time parsing: Natural language dates ("yesterday at 8am", "friday", "2 days ago")
• Themes: Choose from all built-in Textual themes with persistent preference
• AWS CloudWatch: Download and list CloudWatch log groups, streams, and events with stream names

📖 Full User Guide — detailed documentation for all features with examples

🎬 Demo Videos — step-by-step workflow walkthroughs: You Just Got Paged · What Changed? · Post-Mortem

Installation

Requires Python 3.13+.

# Recommended: install with uv
uv tool install logdelve

# With AWS CloudWatch support
uv tool install logdelve[aws]

# No uv? Install it first
curl -LsSf https://astral.sh/uv/install.sh | sh

# Alternative: pip
pip install logdelve

Quick Start

# View a log file
logdelve inspect app.log

# Merge multiple log files chronologically
logdelve inspect api.log worker.log scheduler.log

# Pipe logs from kubectl
kubectl logs deploy/my-app --since=1h | logdelve inspect

# Download and view CloudWatch logs
logdelve cloudwatch get /aws/ecs/my-service prefix -s 1h | logdelve inspect

# Filter by time range
logdelve inspect --start "14:30" --end "14:35" app.log

# Export filtered lines to file
logdelve inspect --session my-session --output export.log app.log

# Compare against a known-good baseline
logdelve inspect --baseline yesterday.log today.log

UI Layout

┌─────────────────────────────────────────────────────────────────────────────┐
│ f filter-in  F filter-out  x filters off  │  a analyze  │  / search         │ ← Toolbar
├──┬──┬────┬──────────┬───────────────────────────────────────────────────────┤
│▌ │E │·1  │ 10:30:01 │ {"event": "Connection refused", "host": "10.0.1.5"}   │ ← Anomaly + Level + Component + Time + Content
│  │I │·2  │ 10:30:02 │ {"event": "Request processed", "status": 200}         │
│  │I │·1  │ 10:30:03 │ {"event": "Health check passed"}                      │
│▌ │E │·1  │ 10:30:04 │ {"event": "Timeout after 5000ms", "path": "/api"}     │ ← Error background
│  │D │·3  │ 10:30:05 │ {"event": "Cache hit", "key": "user:42"}              │ ← Debug (dim)
│  │I │·2  │ 10:30:06 │ {"event": "Request processed", "status": 200}         │
├──┴──┴────┴──────────┴───────────────────────────────────────────────────────┤
│ 500 lines  E:12 W:3  A:2                                      app.log       │ ← Status bar
├─────────────────────────────────────────────────────────────────────────────┤
│ s Sessions  q Quit  h Help                                                  │ ← Footer
└─────────────────────────────────────────────────────────────────────────────┘
 ▲  ▲   ▲       ▲                            ▲
 │  │   │       │                            │
 │  │   │       │                            └─ Content (JSON or text)
 │  │   │       └─ Compact timestamp (HH:MM:SS)
 │  │   └─ Component tag (color-coded, c to cycle: tag/full/off)
 │  └─ Log level badge (E=error, W=warn, I=info, D=debug)
 └─ Anomaly marker (▌ = new pattern not in baseline)

Use Cases

Outage Investigation with CloudWatch

# 1. Download baseline (yesterday, everything was fine)
logdelve cloudwatch get /aws/ecs/my-service "" \
  -s "yesterday 6:00" -e "yesterday 8:00" > baseline.log

# 2. Download current logs (outage happening now)
logdelve cloudwatch get /aws/ecs/my-service "" -s 1h > current.log

# 3. Compare — automatically shows only anomalous lines
logdelve inspect --baseline baseline.log current.log

# 4. In the TUI:
#    - Anomaly filter is auto-enabled (! to toggle off)
#    - Press 'a' to analyze message groups
#    - Press 'e' to filter by ERROR level
#    - Press 'x' to suspend all filters and see context around a line

CloudWatch Log Download

# Flexible time formats
logdelve cloudwatch get /aws/ecs/my-service prefix -s "2 days ago"
logdelve cloudwatch get /aws/ecs/my-service prefix -s "friday" -e "saturday"
logdelve cloudwatch get /aws/ecs/my-service prefix -s "yesterday at 8am"
logdelve cloudwatch get /aws/ecs/my-service prefix -s 1h          # shorthand
logdelve cloudwatch get /aws/ecs/my-service prefix -s "Feb 13 2026 7:58"

# List available log groups and streams
logdelve cloudwatch groups /aws/ecs/
logdelve cloudwatch streams /aws/ecs/my-service

# Live tail CloudWatch logs
logdelve cloudwatch get /aws/ecs/my-service prefix --tail | logdelve inspect

# Each line includes [stream-name] prefix for component detection

Analyzing Multi-Component Logs

# Pipe mixed logs from multiple pods
kubectl logs -l app=my-service --prefix --since=30m | logdelve inspect

# In the TUI:
#    - Component tags (·1, ·2) identify pods — press 'c' to see full names
#    - Press 'f'/'F' and switch to Component tab to filter by pod
#    - Press 'e' to filter by log level (ERROR → WARN → INFO)
#    - Press 'a' then 'm' to switch to field analysis mode
#    - Select a field value to filter (e.g., http_status: 500)

Keybindings

Key	Action		Key	Action
f / F	Filter in / out		/ / ?	Search forward / backward
e	Cycle level filter		n / N	Next / previous match
Ctrl+D	Clear all search patterns
!	Toggle anomaly filter		j	Toggle JSON pretty-print
x	Suspend / resume filters		c	Cycle component display
m	Manage filters		a	Analyze messages
1-9	Toggle individual filter		s	Session manager
p	Pause / resume tailing		h	Help screen
r	Show related (trace ID)		:	Go to line number
b / B	Bookmark / list		A	Annotate bookmark
[ / ]	Prev / next bookmark		@	Jump to timestamp
Ctrl+E	Export filtered lines		t	Select theme

📖 Full keyboard reference in the User Guide

Development

# Clone the repo
git clone https://github.com/chassing/logdelve.git
cd logdelve

# Install dependencies
uv sync

# Run from source
uv run logdelve inspect sample.log

# Run all checks (lint, format, typecheck, tests)
make test

# Individual targets
make lint        # ruff check
make format      # ruff format
make typecheck   # mypy
make clean       # remove caches

# Performance benchmark
uv run python scripts/perf_test.py

Contributing

1. Fork the repo and create a feature branch
2. Install dependencies: uv sync
3. Make your changes
4. Run checks: make test (runs lint, format check, type check, and tests)
5. Submit a pull request

Code style

• Python 3.13+ with type hints (strict mypy)
• Formatting and linting via Ruff
• Pydantic models for data structures
• Textual framework for the TUI

Releasing

See AGENT.md for the full release checklist.

# After updating version + changelog + README:
git push
gh release create vX.Y.Z --title "vX.Y.Z" --notes "See CHANGELOG.md"

The publish.yml workflow builds and publishes to PyPI automatically. Requires PYPI_TOKEN secret in the GitHub repo settings.

Alternatives

• lnav — Feature-rich log file navigator with SQL queries, automatic format detection, and timeline view. C++-based.
• tailspin — Log file highlighter with automatic pattern detection. Focused on making logs readable, no filtering or analysis.
• GoAccess — Real-time web log analyzer with terminal and HTML dashboards. Specialized for access logs (Apache, Nginx), not general-purpose.
• Textualog — Textual-based log viewer. Minimal feature set, no filtering or anomaly detection.
• jq — JSON processor. Powerful for one-off queries but no TUI, no live tailing, no session management.

License

MIT