Sankey visualisation

plot_lineage_sankey renders any dependency graph as an interactive Plotly Sankey diagram — orange flow lines on a navy background, matching the Teradata visual identity.

Basic usage

from tdfs4ds.lineage import build_teradata_dependency_graph, plot_lineage_sankey

graph = build_teradata_dependency_graph(sql_query=sql)
fig   = plot_lineage_sankey(graph, title='Customer Dataset Lineage')
fig.show()                       # inline in a Jupyter notebook

Save to a self-contained HTML

from tdfs4ds.lineage import show_plotly_robust

show_plotly_robust(fig, html_path='lineage_customer_dataset.html')

show_plotly_robust writes a standalone HTML file that requires no Python or Jupyter — share it with stakeholders directly.

Customise the figure

plot_lineage_sankey is a thin Plotly wrapper, so you can post-process the figure:

fig.update_layout(
    title_text='Customer Dataset Lineage',
    font=dict(family='Inter', size=12),
    paper_bgcolor='#FFFFFF',
)
fig.write_html('lineage.html')

Reading the diagram

Visual	Meaning
Left → Right	Upstream → Downstream
Node width	Cumulative weight (number of incoming edges)
Edge colour	Constant orange (Teradata brand)
Hover text	Object name + type (table / view / dataset)

The graph is directed but Sankey rendering is symmetric — read it as a flow of derived columns rather than as physical I/O.

Combining with the consumer agent

The agent's LINEAGE intent calls build_teradata_dependency_graph under the hood and synthesises a plain-language summary alongside a depth-ordered dependency list:

from tdfs4ds.agent import consumer_agent
print(consumer_agent("Where does total_amount come from?", thread_id="x"))

For a richer multi-process view, build the graph yourself and use plot_lineage_sankey — it scales to dozens of nodes without losing readability.