rtplot — real-time plotting over ZMQ

rtplot pushes live data from a Python script to a browser plot — locally or across a network — in a few lines of code. The plot page also hosts interactive widgets (buttons, sliders, dials, numeric / text displays) that feed values back to the sender in real time.

Typical use: a robot or data-acquisition script runs on a Raspberry Pi, and you watch signals and tweak gains from a laptop on the same Wi-Fi.

How it works

Two processes talking over ZMQ, plus a browser viewer:

flowchart TD
    script["Your Python script<br/>send_array() · poll_controls()"]
    subgraph server["rtplot-server"]
        browser["browser tab<br/>localhost:8050"]
    end
    script -- "data :5555" --> server
    server -- "controls :5556" --> script

Sender and server don't have to be on the same machine — see the networking guide.

Install

Server — grab the prebuilt binary from the Releases page:

No Python needed on the viewing machine. On Windows the binary opens a small Tk status window showing the listening URL.

Set columns per row in your sender's layout with PlotRow([Plot(names=["a"]), Plot(names=["b"])], columns=2). Start the server normally; each row can use a different column count. See the multi-column example for typed and dict forms, including a layout with two, one, and three columns.

Client — pip install in the env that runs your script:

pip install better-rtplot

(If you'd rather run the server from Python too, use pip install "better-rtplot[browser]" — see the API reference.)

Your first plot

flowchart TD
    t1["terminal 1: start the server"]
    t2["terminal 2: run your script"]
    br["browser: open localhost:8050"]
    t1 --> t2 --> br

Terminal 1 — start the server. Run the prebuilt rtplot-server binary you downloaded above. Open http://localhost:8050 in a browser — the page is blank until data arrives.

Terminal 2 — run your sender script. Save as my_plot.py:

from rtplot import client
import time

client.local_plot()
client.initialize_plots(["my signal"])

for i in range(1000):
    client.send_array(i * 0.01)
    time.sleep(0.01)

python my_plot.py

A rising line now draws itself in the browser tab.

Highlights

• Fast. Binary WebSocket deltas up to 1 kHz; the browser coalesces samples into one repaint per requestAnimationFrame, so rendering tracks your monitor refresh rate regardless of sample rate.
• Browser-based. aiohttp + uPlot, no desktop GUI toolkit, works over SSH port forwarding.
• Remote-friendly. Sender or plot host can bind. Live Bind / Connect buttons retarget without restart.
• Config lives with the data. The sender declares plot layout.
• Responsive presentation. Native phone sizing, essential actions, compact navigation, and explicit semantic styles. See the presentation guide and before/after screenshots.
• Sections and runtime state. Optional collapsible sections, persistent status controls, and presentation updates that preserve streaming history. See the section API and workflow example.
• Interactive controls. Buttons, sliders, dials, displays — polled from your loop, no threads, no callbacks.
• LaTeX labels. Use $...$ or \(...\) in plot and control text; the server bundles KaTeX and works offline.
• Static HTML snapshots. save_snapshot("out.html") writes a self-contained file with the current trace and math rendering embedded.

Where to go next

• API reference — every rtplot.client function, the plot-layout schema, interactive controls, snapshots, browser UI, and rtplot-server CLI flags.
• Networking guide — Mode A vs. Mode B, viewing from a phone or second laptop, the WSL2 wrinkle, Cloudflare Tunnel, Tailscale.
• Examples — runnable scripts with embedded HTML snapshots you can open offline.

Issues and feature requests: github.com/jmontp/rtplot/issues.

Dropdown controls: API and examples.