mermaiden

mermaiden generates deterministic Mermaid source from typed Python diagram models. It supports the Mermaid syntax families registered by the package and validates each diagram before rendering its text.

The package produces Mermaid text only. Render it with Mermaid in a browser, Markdown viewer, or your own CLI workflow.

Installation

Python 3.12 or later is required.

pip install mermaiden

The wheel declares and installs its Python dependencies. The compat CLI command needs no additional tools. Full SVG validation through Application.validate_render() shells out to the mmdc executable and requires a compatible browser. Development installs the exact Mermaid CLI version from package-lock.json; set PUPPETEER_EXECUTABLE_PATH when using a system browser.

Quick start

from mermaiden import Application

with Application.create() as application:
    diagrams = application.available_diagrams()
    sequence = application.diagram_info("sequenceDiagram")
    description = application.diagram_description(sequence.id)
    payload_type = application.command_payload(sequence.id, "add_participant")
    payload = payload_type.validate({"id": "api", "label": "API", "kind": "control"})

    assert sequence in diagrams
    assert "add_participant" in description.commands
    assert payload.values["kind"] == "control"

Application.available_diagrams() returns the supported diagram catalog. Application.diagram_info(diagram_id) returns the typed diagram API for an individual syntax. CLI workflows are available through python -m mermaiden.cli.

The architecture and ownership map identifies the layers, entry points, lifetimes, and authoritative files behind that catalog.

Application API

Application is the boundary for API and persistence adapters. Each handle uses the process-owned scope created once by the bootstrap module; use it as a context manager or call close() to invalidate that handle. Create a diagram by Mermaid syntax id, apply a named domain command, and persist the JSON-safe snapshot returned by the application.

from mermaiden import Application

with Application.create() as application:
    diagram = application.create_diagram("sequenceDiagram")
    change = application.apply_batch(
        diagram,
        (
            {"operation": "add_participant", "arguments": {"id": "client", "label": "Client"}},
            {"operation": "add_participant", "arguments": {"id": "api", "label": "API"}},
            {
                "operation": "add_message",
                "arguments": {
                    "id": "request",
                    "source_id": "client",
                    "target_id": "api",
                    "label": "Request",
                },
            },
        ),
    )
    assert change.can_commit

    payload = application.snapshot(diagram).to_dict()
    restored = application.restore(payload)
    source = application.render(restored)
    report = application.validate_render(restored)
    if not report.success:
        raise RuntimeError(report.diagnostics)
    svg = report.svg

apply_batch() accepts an ordered sequence of {"operation": str, "arguments": mapping} objects. It validates command payloads as usual, defers whole-diagram constraint inspection until the batch is complete, and restores the exact pre-batch snapshot if any command or the final validation fails. An empty batch leaves the diagram unchanged and returns its current validation report.

validate_render() is non-mutating and runs Mermaid's complete rendering and layout phase through mmdc; its report contains the compatible Mermaid version, SVG output, and structured diagnostics.

Snapshots have a versioned envelope and may be stored as JSON. Version 6 uses registry-owned discriminators such as mermaiden/element/classDiagram/class; snapshots never contain importable Python module paths. Its closed envelope schema is published at src/mermaiden/runtime/snapshot/schema.v6.json. Earlier and unknown versions are rejected; there is no implicit migration or compatibility reader. Newly created and incomplete diagrams are marked as drafts: callers may snapshot and restore them between accepted commands, but Application.render() rejects them until their blocking constraints are resolved. Snapshot parsing and typed hydration reject malformed persisted data, and restoration verifies snapshots that were recorded as valid. Command argument values use the diagram operation names; JSON string values are accepted for enum arguments.

The caller can discover the REST contract without maintaining a manifest. diagram_description() returns JSON Schema for the diagram's elements, relations, annotations, and commands. command_payload() returns the generated Pydantic request model for one command. The generated mutation contract records the supported update, move, reorder, and retarget behavior for every registered diagram.

Element removal is conservative by default: remove_element rejects an element that still has descendants, relations, or annotations. Passing cascade: true removes the complete diagram-defined subtree and every dependent relation and annotation atomically. In Tree View diagrams, branches define that subtree. Removing the final element returns the diagram to an empty, persistable draft; drafts have no Mermaid source until they become valid again.

Updating and moving elements

Mutation arguments are JSON-shaped and validated before the diagram changes. Updates preserve identity, moves preserve the complete subtree, and reorders require the exact current members of one collection. Rejected mutations leave the complete snapshot unchanged.

from mermaiden import Application

with Application.create() as application:
    diagram = application.create_diagram("block")
    application.execute(diagram, "add_group", {"id": "source_example", "label": "Source Example"})
    application.execute(diagram, "add_group", {"id": "target_example", "label": "Target Example"})
    for id, label in (
        ("first_example", "First Example"),
        ("second_example", "Second Example"),
        ("third_example", "Third Example"),
    ):
        application.execute(
            diagram,
            "add_block",
            {"id": id, "label": label, "parent_id": "source_example"},
        )

    application.execute(
        diagram,
        "update_element",
        {"id": "first_example", "kind": "block_node", "changes": {"label": "Updated First Example"}},
    )
    application.execute(
        diagram,
        "move_element",
        {"id": "first_example", "kind": "block_node", "parent_id": "target_example", "position": 0},
    )
    application.execute(
        diagram,
        "reorder_elements",
        {"parent_id": "source_example", "element_ids": ["third_example", "second_example"]},
    )

    for operation, arguments in (
        (
            "update_element",
            {"id": "first_example", "kind": "block_node", "changes": {"id": "renamed_example"}},
        ),
        (
            "move_element",
            {"id": "second_example", "kind": "block_node", "parent_id": "third_example"},
        ),
        (
            "reorder_elements",
            {"parent_id": "source_example", "element_ids": ["second_example"]},
        ),
    ):
        before = application.snapshot(diagram).to_dict()
        try:
            application.execute(diagram, operation, arguments)
        except RuntimeError:
            pass
        else:
            raise AssertionError(f"{operation} unexpectedly succeeded")
        assert application.snapshot(diagram).to_dict() == before

    snapshot = application.snapshot(diagram).to_dict()
    restored = application.restore(snapshot)
    assert application.snapshot(restored).to_dict() == snapshot
    assert application.render(restored) == application.render(diagram)
    report = application.validate_render(restored)
    assert report.success and report.svg

Development

Create the development environment from the committed lock before running repository commands:

uv sync --locked --group dev

The fast tier performs no npm, browser, network, or external Mermaid work:

make fast-check

Full SVG compatibility is an explicit integration tier. It installs the lock-pinned Mermaid CLI once and requires a compatible browser:

make integration

Generate the complete fixture preview without opening a browser; CI runs this non-interactive prerequisite:

make diagrams-preview

The interactive command is local-only and opens the generated preview with the platform open command:

make diagrams-test

The complete host-mode CI target runs quality, pytest, Mermaid compatibility, and package verification concurrently:

make ci

GitHub Actions runs quality, pytest, Mermaid compatibility, and package verification concurrently; the ci job is their stable aggregate result for branch protection.

License

Mermaiden is proprietary software. No permission to use, copy, modify, or distribute it is granted without prior express written permission. See LICENSE for the complete terms.

Release

Releases are versioned by annotated vX.Y.Z tags. make ci first runs uv lock --check and uv sync --locked --group dev, so release verification uses the same committed development lock. After it passes, create and push the tag, then publish the GitHub release:

git tag -a v2.0.0 -m "v2.0.0"
git push origin v2.0.0
gh release create v2.0.0 --verify-tag --generate-notes --title v2.0.0