Pytter

A terminal-based internet speed test with a live TUI dashboard — measures ping, download, and upload throughput using multiple parallel connections against globally distributed servers.

Preview

Features

• Live dashboard — animated ping/download/upload readout rendered with curses, updated in real time.
• Accurate high-speed measurement — uses parallel TCP streams and a duration-based measurement window (not a single-file timer), so results stay accurate on fast connections (100+ Mbps) instead of being skewed by TCP slow-start and connection setup overhead.
• Global, provider-agnostic servers — download traffic is spread across Hetzner and OVH endpoints in Europe, North America, Asia, and Oceania; upload traffic rotates across multiple independent echo services. If one provider throttles or blocks a request, the affected connection automatically rotates to the next server — no manual configuration required.
• No external dependencies — built entirely on the Python standard library (except on Windows, see below).
• Cancelable at any time — press q to stop a running test early.
• Optional IP privacy — hide your public IP from the display with a single flag.

Requirements

• Python 3.8 or newer

• Linux / macOS: no extra packages needed — curses ships with the standard library.

• Windows: the standard library does not include curses. Install:

  pip install windows-curses

Installation

git clone https://github.com/Ars-byte/pytter.git
cd pytter

No further setup is required — there is no requirements.txt because the project has no third-party dependencies (aside from windows-curses on Windows).

Usage

Run the test:

python3 main.py

Hide your IP address from the results screen:

python3 main.py -hideip

Controls

Key	Action
q / Q	Cancel the running test
Ctrl+C	Force-quit immediately
any key	Dismiss the results screen

How the measurement works

Naively timing a single download of a small file produces misleading results on fast connections: most of the elapsed time is spent in TCP connection setup and slow-start rather than steady-state transfer. Pytter avoids this the same way tools like fast.com and speedtest.net do:

1. Parallel streams — several TCP connections run concurrently (6 for download, 4 for upload) instead of one.
2. Warm-up phase — the first ~2 seconds of traffic are discarded so TCP windows have time to ramp up.
3. Fixed measurement window — bytes transferred across all streams are summed over a steady ~8 second window.
4. Throughput = total bytes transferred × 8 ÷ measured elapsed time.

Servers used

Phase	Providers
Download	Hetzner (Falkenstein, Nuremberg, Helsinki, Ashburn, Hillsboro), OVH (Gravelines, Strasbourg, Roubaix, Singapore, Sydney, Beauharnois)
Upload	httpbin.org, postman-echo.com, httpbingo.org
Ping	Public DNS resolvers (Google, Cloudflare, OpenDNS) via TCP handshake timing

Each worker thread iterates over its provider list independently; if a request to one server fails, that thread moves on to the next entry without interrupting the rest of the test.

Project structure

.
├── main.py                    # Entry point, curses event loop
└── src/
    ├── backend/
    │   └── speedtest.py       # SpeedTester: ping/download/upload measurement
    └── frontend/
        ├── ascii_art.py       # Logo, progress bars, ping labels
        └── tui.py             # TUI: dashboard and results rendering

License

MIT