Basic usage

kedro mermaid generate

Full example

kedro mermaid generate \
    --from-inputs=raw_data \          # Start from the `raw_data` dataset
    --to-outputs=report \             # End at the `report` dataset
    --nodes=clean_data,train_model,report_results \  # Include only these nodes
    --tags=model,reporting \          # Include only nodes tagged with `model` or `reporting`
    --namespaces=data,model \         # Include only nodes in the `data` or `model` namespaces
    --pipeline=data_science \         # Target the `data_science` pipeline
    --set-node-attr pattern="(?P<category>\w+)__(?P<node>\w+)" \  # Group nodes by prefix before `__`
    --set-node-attr params.shape=circle \  # Change node shape to circles
    --set-graph-attr declaration="flowchart TB" \  # Change layout to top-to-bottom
    --set-graph-attr config.layout=elk \  # Use the ELK layout engine
    --set-edge-attr params.arrow='---'  # Change edge arrows to dashed lines

Filter the diagram

Based on datasets, nodes, tags, or namespaces

kedro mermaid generate \
    --tags=model,reporting \
    --from-inputs=raw_data \
    --pipeline=data_science

Only show the nodes in a specific format

For instance, if your nodes are named like data__load_data, data__clean_data, model__train_model, and model__evaluate_model, you can use:

kedro mermaid generate --set-node-attr pattern="(?P<category>\w+)__(?P<node>\w+)"

This will group nodes into subgraphs based on the category prefix and label them with the node_name:

• data__load_data: appears in the data subgraph with the label Load Data
• data__clean_data: appears in the data subgraph with the label Clean Data
• model__train_model: appears in the model subgraph with the label Train Model
• model__evaluate_model: appears in the model subgraph with the label Evaluate Model

Limit the granularity of the diagram

If your nodes are named with an indicative pattern, such as level1.level2.level3.level4, you can create a regex pattern to capture only the top-level namespace:

kedro mermaid generate --set-node-attr pattern="(?P<category>[^.]+)(?:[.](?P<node>[^.]+)){2}"

Breakdown of the regex: - (?P<category>[^.]+): Captures the first segment before the first dot as the category. - (?:[.](?P<node>[^.]+)){2}: Non-capturing group that matches a dot followed by a segment, repeated twice to capture the next two segments as node.

This pattern will group nodes into subgraphs based on the top-level namespace and label them with the next two segments, effectively reducing the diagram's complexity while retaining meaningful structure. For example: - data.load.clean.transform: appears in the data subgraph with the label Load Clean - data.load.clean.aggregate: Will be merged with data.load.clean.transform as they share the same category and node labels. - model.train.evaluate.report: appears in the model subgraph with the label Train Evaluate

Style the diagram

Change the layout

kedro mermaid generate --set-graph-attr declaration="flowchart TB"

Use the ELK layout engine

kedro mermaid generate --set-graph-attr config.layout=elk