ezmsg-tools

A namespace package for ezmsg to visualize running graphs and data.

The data visualization is highly fragile. Expect bugs.

Installation

Pre-requisites

• graphviz

On Mac, you should use brew:

• brew install graphviz
• export CFLAGS="-I $(brew --prefix graphviz)/include"
• export LDFLAGS="-L $(brew --prefix graphviz)/lib"

Release

Install the latest release from pypi with: pip install ezmsg-tools (or uv add ... or poetry add ...).

More than likely, you will want to include at least one of the extras when installing:

pip install "ezmsg-tools[all]"

Development Version

If you intend to edit ezmsg-tools then please refer to the Developers section below.

You can add the development version of ezmsg-tools directly from GitHub:

• Using pip: pip install git+https://github.com/ezmsg-org/ezmsg-tools.git@dev
• Using poetry: poetry add "git+https://github.com/ezmsg-org/ezmsg-tools.git@dev"
• Using uv: uv add git+https://github.com/ezmsg-org/ezmsg-tools --branch dev

You probably want to include the extras when installing the development version:

• pip install "ezmsg-tools[all] @ git+https://github.com/ezmsg-org/ezmsg-tools.git@dev"

Getting Started

This package includes some entrypoints with useful tools.

ezmsg-performance-monitor

This tool operates on logfiles created by ezmsg. Logfiles will automatically be created when running a pipeline containing nodes decorated with ezmsg.sigproc.util.profile.profile_subpub, and if the EZMSG_LOGLEVEL environment variable is set to DEBUG. The logfiles will be created in ~/.ezmsg/profile/ezprofiler.log by default but this can be changed with the EZMSG_PROFILE environment variable.

Most of the nodes provided by ezmsg.sigproc are already decorated to enable profiling, as is any custom nodes that inherit from ezmsg.sigproc.base.GenAxisArray. You can decorate other nodes with ezmsg.sigproc.util.profile.profile_subpub to enable profiling.

During a run with profiling enabled, the logfiles will be created in the specified location. You may wish to additionally create a graph file: (uv run) EZMSG_LOGLEVEL=WARN ezmsg mermaid > ~/.ezmsg/profile/ezprofiler.mermaid

During or after a pipeline run with profiling enabled, you can run (uv run ) performance-monitor to visualize the performance of the nodes in the pipeline.

Unlike signal-monitor, this tool does not require the pipeline to attach to an existing graph service because it relies exclusively on the logfile.

ezmsg-signal-monitor

The pipeline must be running on a graph service exposed on the network. For example, first, run the GraphService on an open port:

ezmsg --address 127.0.0.1:25978 start

Then run your usual pipeline but make sure it attaches to the graph address by passing graph_address=("127.0.0.1", 25978) as a kwarg to ez.run.

While the pipeline is running, you can run the signal-monitor tool with (uv run) signal-monitor --graph-addr 127.0.0.1:25978.

This launches a window with graph visualized on the left. Click on a node's output box to get a live visualization on the right side of the screen plotting the data as it leaves that node. Use a to toggle auto-scaling. With auto-scaling off, use -, and = to zoom out and in, respectively.

Currently only 2-D outputs are supported!

Don't forget to shutdown your graph service when you are done, e.g.: ezmsg --address 127.0.0.1:25978 shutdown

Developers

We use uv for development. It is not strictly required, but if you intend to contribute to ezmsg-tools then using uv will lead to the smoothest collaboration.

1. Install uv if not already installed.
2. Fork ezmsg-tools and clone your fork to your local computer.
3. Open a terminal and cd to the cloned folder.
4. Make sure pygraphviz pre-requisites are installed.
   • On mac: export CFLAGS="-I $(brew --prefix graphviz)/include" and export LDFLAGS="-L $(brew --prefix graphviz)/lib"
5. uv sync --all-extras --python 3.10 to create a .venv and install ezmsg-tools including dev and test dependencies.
6. After editing code and making commits, Run the test suite before making a PR: uv run pytest

Troubleshooting

Graphviz can be difficult to install on some systems. The simplest may be to use conda/mamba: conda install graphviz. If that fails, see here.