Acherion

Acherion is an embeddable flow-graph editor and runtime workbench for Python. It provides a NiceGUI-based visual designer, a typed graph/runtime API, and a standalone host that compiles graphs into executable Python code.

What Acherion provides

• A standalone workbench you can launch with acherion or python -m acherion
• A reusable NiceGUI workbench component for embedding inside your own app
• A typed graph model for saving, restoring, compiling, and executing graphs
• Preview execution hooks with transient runtime bindings and value summaries
• Built-in NumPy and Plotly support for array- and figure-oriented nodes
• Extension points for custom hosts, catalog modules, preview adapters, and validation helpers

Installation

Install the published package:

pip install acherion

Install from source for local development:

pip install -e ".[dev]"

Python 3.11 or newer is required.

Standalone quick start

Run the packaged standalone workbench:

acherion

Equivalent entry points:

python -m acherion

from acherion.app import main

main(host='127.0.0.1', port=8081, reload=False)

The standalone app mounts a full-screen Acherion workbench with:

• the graph designer
• the generated-code pane
• live preview execution for the default run event
• graph-to-Python validation through the standalone host

Embedding quick start

The core API lives in acherion. The NiceGUI embedding layer lives in acherion.embed.

from nicegui import ui

from acherion import StandaloneAcherionHost
from acherion.embed import AcherionWorkbench


persisted_graph: dict[str, object] = {}


@ui.page('/')
def index() -> None:
		workbench: AcherionWorkbench | None = None

		def handle_change() -> None:
				if workbench is None:
						return
				persisted_graph.clear()
				persisted_graph.update(workbench.designer.to_dict())

		workbench = AcherionWorkbench(
				host=StandaloneAcherionHost(),
				on_change=handle_change,
				refresh_code_after_build=True,
			theme_overrides={
				'surface': '#111827',
				'text': '#f9fafb',
				'primary': '#22c55e',
				'sidebar_panel_bg': 'rgba(255,255,255,0.05)',
			},
		)
		workbench.build()
		workbench.designer.set_graph_from_dict(persisted_graph)


ui.run()

This uses the generic standalone host, but the workbench can also run against your own host implementation through AcherionHost or compose_acherion_host.

Passing theme_overrides opts the embedded workbench into Acherion's scoped theme CSS without restyling the rest of the host application. Keys can be the semantic aliases shown above or raw CSS variable names like --oe-bg and --ach-sidebar-panel-bg. Pass an empty dict if you want the default Acherion embedded theme with no overrides.

You can also switch themes after build():

workbench.set_theme_overrides({
	'surface': '#ffffff',
	'text': '#0f172a',
	'primary': '#2563eb',
})

workbench.set_theme_overrides({
	'surface': '#0f172a',
	'text': '#e2e8f0',
	'primary': '#38bdf8',
})

workbench.clear_theme_overrides()

clear_theme_overrides() disables the scoped embedded Acherion theme and lets the surrounding host application styling take over again.

Core concepts

Graph model

• AcherionGraph is the serializable top-level graph object.
• AcherionNode stores one node's kind, title, id, and params.
• AcherionDesigner.to_dict() returns a JSON-safe representation.
• AcherionDesigner.set_graph_from_dict() restores persisted state.

Workbench layer

• AcherionWorkbench wraps AcherionDesigner with code-view and validation behavior.
• AcherionWorkbench.generated_user_code() returns the current compiled code.
• AcherionWorkbench supports callbacks for change handling, validation, preview execution, and custom code views.

Host layer

An AcherionHost is responsible for integrating the designer with your app. Hosts provide:

• external event definitions
• code generation
• runtime bindings for previews or generated code
• graph normalization/system-node sync
• host-owned configuration UI for special nodes

Runtime helpers

The standalone runtime helpers in acherion are useful even outside the UI:

from acherion import compile_acherion_graph, execute_acherion_graph

They compile a graph into Python source and execute the generated event handler with optional scoped bindings.

Extension points

Acherion exposes several host-facing registration hooks:

• register_catalog_module_specs to expose Python modules in the function catalog
• register_preview_value_adapter to customize preview summaries/type tags
• register_acherion_validation_extension to extend validation globals
• register_acherion_node_definition to add or replace node metadata

Custom node definitions are only part of the story: if you introduce a new node kind, your host/compiler also needs to know how to emit runtime code for it.

Documentation set

Repository documentation is split into focused guides:

• docs/overview.md for architecture and feature overview
• docs/standalone.md for standalone usage and persistence patterns
• docs/embedding.md for NiceGUI embedding and host composition
• docs/extending.md for catalogs, validation, preview, and node extension hooks
• docs/releasing.md for local packaging checks and GitHub/PyPI release flow

Contributing

Contributions are welcome, especially changes that improve the public runtime API, embedded workbench behavior, packaging flow, and documentation.

For local development:

pip install -e ".[dev]"
python -m pytest
python -m build --sdist --wheel
python -m twine check dist/*

When adding regression coverage, prefer behavior-focused tests over brittle UI snapshots. The most useful tests in this repository usually validate:

• public API behavior
• graph compile and execution paths
• theme override normalization and generated CSS contracts
• extension and registration hooks

This repository uses conventional commits and semantic versioning. Pull request titles and merge commits should follow the same convention, for example:

• feat: add preview regression coverage
• fix: preserve theme override aliases
• docs: clarify embedded host setup

In general:

• use feat: for user-visible capabilities or meaningful public API expansion
• use fix: or perf: for patch-level runtime changes
• add or update focused regression tests when changing behavior that users rely on

Packaging and release

Build distributions locally with:

python -m build --sdist --wheel
python -m twine check dist/*

This repository includes GitHub Actions workflows for CI, conventional-commit enforcement, semantic versioned GitHub releases, and trusted publisher releases to PyPI. See docs/releasing.md for the operational flow.