flograph

A visual node-based Python programming environment: flow-based dataflow on an infinite Blueprint-style canvas, where every node is real, editable Python.

Install

flograph is a standard pip-installable package (hatchling build backend):

pip install .            # or: uv pip install .   /   pip install -e .

This puts a flograph command on your PATH and makes python -m flograph work. To build a distributable wheel/sdist instead:

python -m build          # or: uv build   ->   dist/flograph-*.whl

The project was renamed from flopy to flograph because flopy is already taken on PyPI (USGS MODFLOW).

One-file script (no install)

If you can't pip install on a given machine (locked-down work laptop, no package-index access, etc.) but PySide6, pandas, jedi, and psutil are already available there, build a single self-contained .py file instead:

python scripts/build_onefile.py          # writes dist/flograph_onefile_<version>.py

Copy (or email yourself) that one file and run it directly — no install step, no flograph package needed on the target machine:

python flograph_onefile_<version>.py     # or: python flograph_onefile_<version>.py project.flograph

It embeds flograph's own source as a base64 zip and unpacks it to a temp dir at startup; it does not bundle the third-party dependencies themselves.

Run it

flograph                 # console entry point (after install)
python -m flograph                          # equivalent module entry point
python main.py project.flograph             # open a project
python -m flograph.engine.headless project.flograph   # run without GUI

The idea

• Nodes are Python scripts. Every node — including the shipped library — is a small module: a NODE dict declaring typed ports, an optional PARAMS list that auto-generates its properties form, and a run(ctx, **inputs) function. Double-click any node to read or fork its code in the built-in editor (syntax highlighting, jedi completion, error markers on the failing line).
• Dataflow semantics. Data flows through typed ports; execution is a topological walk of the dirty subgraph; every node's outputs are cached, so re-runs only recompute what changed. Status LEDs: gray idle, yellow queued, pulsing blue running, green done, red error.
• Inspect everything. Click any node or wire to see the data on it — paged table view for DataFrames (millions of rows are fine), matplotlib figures with a toolbar, pretty-printed objects. Per-node stdout/logs in the console dock.

Canvas

Projects are plain JSON (.flograph); caches are never saved, so a reopened project is fully reproducible with one F5.

Node library

The shipped library covers the essentials:

• IO — read/write CSV, Excel, Parquet, JSON (incl. JSONL), SQLite (query in, table out), inline Table.
• Transform — Select Columns, Filter Rows, Sort, Join, Group By, Expression, Concatenate, Missing Values, Duplicate Row Filter, Rename Columns, Pivot, Unpivot, Row Sampling, Convert Types, String Manipulation, Statistics.
• Viz — Show Table, Show Plot (live on-canvas cards), Show Plotly (fully interactive plotly.js chart embedded on the canvas — hover, zoom and pan in place; needs pip install plotly, e.g. via Tools > Manage Packages).
• Scripting / Util — Python Script, Constant, Reroute, Note, Action Button.

Packages

Tools > Manage Packages installs, upgrades and uninstalls pip packages in flograph's own environment (the venv running the app). Nodes execute in-process, so anything installed there is immediately importable from a node's run() — no restart needed for new installs; upgrades of modules the app has already imported take effect on the next launch. The dialog uses pip when the interpreter has it and falls back to uv pip (uv-made venvs ship without pip); flograph's own core dependencies are protected from uninstall.

Writing a node

"""My Node

Docstring first paragraph shows in the properties panel.
"""
NODE = {
    "label": "My Node",
    "category": "Transform",
    "inputs":  [("table", "dataframe")],
    "outputs": [("result", "dataframe")],
}
PARAMS = [
    {"name": "factor", "type": "float", "default": 1.0},
]

def run(ctx, table):
    ctx.log(f"scaling by {ctx.params['factor']}")
    ctx.check_cancelled()          # cooperative cancellation
    return {"result": table * ctx.params["factor"]}

Port types: any, dataframe, series, number, string, bool, object, figure. columns-typed params render with a ▾ picker listing the columns of the DataFrames cached on the node's inputs (run upstream once to populate it); add "multi": False for single-column params so picking replaces instead of toggling a comma list. Rules: treat inputs as read-only (outputs are cached by reference); heavy imports go inside run(); matplotlib figures must use the OO API (matplotlib.figure.Figure()), never pyplot.

Drop new .py files under src/flograph/nodes/<category>/ and they appear in the library on next launch.

Development

uv pip install -p .venv/bin/python -e ".[dev]"
QT_QPA_PLATFORM=offscreen .venv/bin/python -m pytest tests/

Architecture (src layout):

• flograph/core — Qt-free model: graph, typed ports, script contract, registry, JSON serialization. Fully unit-testable; a poison test keeps Qt and pandas out of its import graph.
• flograph/engine — background execution: plan builder, single-thread pool worker, output cache, cancellation, per-node stdout capture, tracebacks mapped to node script lines.
• flograph/nodes — the standard library; each node is a script file loaded as text through the same contract as user code.
• flograph/ui — canvas (QGraphicsView from scratch), code editor, inspector, properties, console. One rule everywhere: QUndoCommands are the only writers to the graph; items react to graph events.

Changelog

See CHANGELOG.md for what's new in each version.

License

MIT — free for commercial and private use, modification, and redistribution; just keep the copyright and license notice.