Skip to content

Installation

FlowGraph runs locally so it can access your files and execute workflows where your data lives. The FlowGraph Web Editor is an authoring-only companion: use it to build and download versioned workflow JSON, then use the binary FlowGraph package to inspect or execute that workflow.

Requirements

  • Python 3.10 or newer.
  • uv for installing FlowGraph and creating an isolated Python environment.

Install uv if it is not already available:

curl -LsSf https://astral.sh/uv/install.sh | sh

On Windows, install uv using the installation instructions in the uv documentation, then open a new terminal so the command is available on your PATH.

Install the binary package from the package index

FlowGraph is distributed as a binary wheel through the package index. Create an isolated environment and install the package by name:

uv venv
uv pip install flowgraph

uv selects the compatible binary wheel for your Python interpreter and operating system. The current binary release targets CPython 3.11 through 3.14.

Run the installed command through the environment managed by uv:

uv run flowgraph --help

The binary package provides the headless flowgraph command for inspecting and executing workflow JSON.

Run temporarily with uvx

For a temporary installation without creating a persistent environment, run the package directly with uvx:

uvx flowgraph

The uvx environment is temporary. Use the binary-wheel installation above when you want a persistent FlowGraph environment.

Inspecting and running a workflow

FlowGraph includes a headless command for inspecting and executing workflow JSON without opening the graphical workspace. Check a workflow with:

uv run flowgraph inspect workflow.json

Inspection shows a compact interface summary by default. Add --full or -f to include the workflow graph.

The Web Editor and local application use the same versioned flowgraph-workflow JSON format. The editor does not execute workflows, read local files, perform mesh operations, or run Python source from a User Function node. Download the workflow and run it locally, for example:

uv run flowgraph run workflow.json

You can provide runtime inputs and parameter overrides without changing the workflow file:

uv run flowgraph run workflow.json \
  --input in_filename=input.stl \
  -i out_filename=output.geo

-i is an alias for --input. Use either form repeatedly to set multiple published workflow inputs.

If an input is omitted, the runner looks for its downstream node's saved parameter value and emits a warning to stderr when it uses that value. This includes nested workflow inputs and directly exported parameters. Inputs with no usable saved value, or with conflicting downstream values, must be supplied explicitly. See Omitted inputs and saved parameter values for details. Fallbacks do not modify the saved workflow file.

Continue with your first project.