Caudata 🦎

Caudata is a collaborative, zero-config multi-server log streamer, real-time metrics dashboard, and high-performance TUI engine. It aggregates and streams real-time logs from multiple remote Linux servers securely over SSH config profiles, as well as the local machine, without installing any agents on remote hosts.

⭐ If you like this project, star it on GitHub — it helps a lot!

Features • Performance & Benchmarks • Installation • Quick Start • CLI Options • Keybindings & Mouse Controls • Configuration • Alternatives

---

Tired of SSH-ing into 5 different servers just to tail Docker logs? Caudata brings them all (and your local setup) into a single, responsive terminal dashboard.

• Zero-Agent SSH: Connects seamlessly via ~/.ssh/config or manual settings. No remote agents or daemons required.
• Local Machine Monitoring: Monitor local containers and system services via direct connection.
• Container Auto-Discovery: Auto-discovers Docker containers and reconnects log streams on container rebuilds/restarts.
• Docker Management: Control containers directly — Start, Stop, Restart, Kill, Inspect, and Remove.
• System Services & Custom Logs: Stream Systemd (Linux), Launchd (macOS), or custom file paths.
• Real-time Metrics: Live CPU, RAM, and disk monitoring for hosts and granular stats for containers.
• ⚡ Rust NIF Accelerated Engine: ~270k lines/s sustained ingest and 100k–530k lines/s across NIF parse/sanitize operations, with sub-element highlighting powered by Rustler.
• ⏱️ Zero-Latency ETS Direct Read: ETS-backed log buffers for non-blocking snapshot reads.
• 🖱️ Mouse & Visual Controls: SGR mouse support for scroll wheeling, drag-to-copy, sidebar clicking, footer action bar, and modal backdrop closing.
• Development Web View: Render the exact TUI inside a web browser via Phoenix LiveView.
• Tailscale VPN Integration: Connect to Tailscale hosts without a system daemon.
• Single-Binary Packaging: Self-contained Burrito executable (~13 MB binary, ~84 MB idle RAM).
• Self-Upgrades: Stay up to date with caudata upgrade.

---

🚀 Performance & Benchmarks

Caudata is engineered for ultra-low memory footprint and high 60 FPS TUI scrolling over heavy log streams.

Benchmark Matrix (Apple M2, 8GB RAM, Elixir 1.19.5 + Erlang 28.5 + Rust NIF)

Component / Task	Before Optimization	After Optimization	Improvement	RAM Footprint
ConfigStore.put_setting	91,340 ops/s (OS :emfile errors)	726,010 ops/s	🚀 8.0x Faster (Sạch 100% OS errors)	0.30 KB
LogSanitizer.process_chunk	39,490 chunks/s	123,610 chunks/s	🚀 3.1x Faster	⬇️ 1.50 KB (-90.1% RAM)
LogFormatter (JSON Parsing)	29,800 lines/s	111,970 lines/s	🚀 3.8x Faster	⬇️ 5.61 KB (-52.6% RAM)
LogFormatter (Text SubHighlight)	18,590 lines/s	49,050 lines/s	🚀 2.6x Faster	⬇️ 4.12 KB (-68.3% RAM)
LogStore.get_snapshot	1,050 ops/s (952 μs latency)	5,150 ops/s (121 μs latency)	🚀 4.9x Faster	0.71 KB
ViewHelper.get_displayed_logs	7,434.52 KB / filter	4.59 KB / filter	💣 1,620x RAM Reduction (-99.94%)	4.59 KB
Production Binary RAM Footprint	~98 MB (mix run dev)	~84 MB (Burrito binary, +S 2:2)	🧠 ~14% lower via release tuning (strip_beams, exclude_apps, +S 2:2)	~84 MB

---

Installation

Quick Install (Recommended)

Run the following command to download and install the pre-compiled binary:

curl -fsSL https://raw.githubusercontent.com/quaywin/caudata/main/install.sh | sh

Running/Building from Source

Caudata uses mise to manage development runtimes:

# Clone the repository
git clone https://github.com/quaywin/caudata.git
cd caudata

# Install required tools (Erlang, Elixir, Zig, and Rust)
mise install

# Install Elixir dependencies
mix deps.get

# Run development mode
mix run --no-halt

To build a standalone executable release using Burrito:

# Build for your current host architecture
mix release --overwrite

# Build for a specific target platform
BURRITO_TARGET=macos_aarch64 mix release --overwrite

---

Quick Start

1. Launch Caudata:

   caudata
   

2. Add Connections:
   • Remote Server (SSH): Caudata automatically scans your ~/.ssh/config. To add custom servers manually, press a in the TUI, select + Manual SSH Connection, and fill in the details.
   • Local Machine: Press a in the TUI, select + Local Machine Connection, and enter your local sudo password (optional, required if your local Docker or system logs require root permissions).
3. Connect & Stream: Use the TUI sidebar to select a host or local connection, press Enter to connect, and Caudata will auto-discover running Docker containers and services. Toggle active log streams using Space.

---

CLI Options

# Show help menu
caudata --help        # or: caudata -h

# Show current version
caudata --version     # or: caudata -v

# Run the web UI server on port 4000
caudata web

# Run the web UI server on a custom port and address
caudata web --port 8080   # or: caudata web -p 8080
CAUDATA_IP=0.0.0.0 caudata web

# Run the web UI server with user-space Tailscale VPN tunneling
caudata web --port 8080 --tailscale --authkey <your-tailscale-key> --tailscale_port 80

# Automatically upgrade to the latest release binary
caudata upgrade

---

Keybindings & Mouse Controls

Press ? inside the TUI to open the full interactive keybinding reference.

🖱️ Mouse Controls

Gesture	Location	Action
Scroll Wheel	Logs Pane	Scroll log stream (auto-fetches history at top)
Drag & Select	Logs Pane	Highlight log range & auto-copy to OS clipboard on release
Left Click	Sidebar	Select server/container & start stream
Left Click	Footer Bar	Trigger action shortcut ([a] Add, [s] Settings, [?] Help, etc.)
Left Click	Modals	Switch tabs, pick actions, or click backdrop to close (Esc)

Global

Key	Action
1 / h / ←	Jump focus to Sidebar panel
2 / l / →	Jump focus to Logs panel
Tab	Toggle active panel focus / switch sidebar tabs
a / A	Open the Add Connection modal
s / S	Open the Global Settings modal
f / F	Toggle log panel full-screen view
t / T	Toggle displaying logs with timestamps
?	Toggle the Help modal
q / Q / Ctrl+C	Exit Caudata

Sidebar (Panel 1: Servers & Containers)

Key	Action
↑ / ↓ (or k / j)	Navigate server & container tree
Enter	Connect to server / select container to stream
Space	Toggle log stream on/off
m / M	Open Docker Container Actions (Start/Stop/Restart/Kill/Inspect/Remove)

Logs Pane (Panel 2: Log Stream)

Key	Action
↑ / ↓ (or k / j)	Scroll logs up/down (auto-accelerates on hold)
g / G	Jump to top (g) or bottom (G) of logs
PageUp / PageDown	Scroll logs page-by-page
/	Enter live regex filter search mode
v / V	Enter Visual Select Mode (extend selection with j/k)
y / Y	Copy all displayed logs (or selected lines in Visual Mode) to clipboard
o	Swap anchor/cursor ends (in Visual Select Mode)
Esc	Clear active filter regex / exit visual mode

---

Configuration

Settings and connection profiles are stored in an ETS-backed database at ~/.caudata/config.db. Caudata automatically defaults the log buffer capacity to 10,000 lines per stream with zero-lag virtualized rendering to maintain low memory footprint and high 60 FPS performance.

Environment Variables

Variable	Description	Default
CAUDATA_CONFIG_PATH	Directory or file path for the configuration database.	~/.caudata/config.db
CAUDATA_TUI	Explicitly enable or disable the Terminal UI (true/false).	Auto-detected
CAUDATA_IP	Bind address for the web server (e.g. 127.0.0.1 or 0.0.0.0).	127.0.0.1
PORT	Port to run the web server on when running caudata web.	4000
TAILSCALE_AUTHKEY	Tailscale authorization key for VPN integration.	None

---

Alternatives

If you are looking for other tools to monitor or view logs, here is how Caudata compares to existing open-source options:

Project	Interface	Primary Focus	Multi-Server (SSH)	Docker Container Auto-Discovery	Peak Throughput / RAM
Caudata	TUI & Web	Agentless Multi-Server Log/Metrics	Yes (native SSH config)	Yes (remote/local)	~270k lines/s ingest (~84 MB RAM)
MultiTail	TUI	Aggregated file/command logs	Yes (manual pipes)	No	~50k lines/s (~80 MB RAM)
Dozzle	Web	Docker container logs	Yes (requires agents)	Yes (needs setup)	~50k lines/s (~70 MB RAM)
lnav	TUI	Log parsing & SQL query	No (mainly local)	No	~300k lines/s (~80 MB RAM)
Lazydocker	TUI	Local Docker management	No	Yes (local only)	~40k lines/s (~100 MB RAM)