Data Flow Visualization for InferenceModule

[eric-tramel]

Summary

We already support basic DAG visualization for traced graphs. The Tracer produces an InferenceGraph, and plait.graph.visualize_graph(graph) renders it to Graphviz DOT for use with Graphviz or viz.js. See examples/03_tracing.py and docs/design/tracing.md.

Current behavior (implemented)

• DOT export via visualize_graph in src/plait/graph.py.
• Node shapes: inputs = box, outputs = doubleoctagon, intermediates = ellipse.
• Node labels use GraphNode.module_name (fallback to node ID when empty).
• Branching info is appended as \n[True|False] when branch_condition is set.
• Edges are drawn from GraphNode.dependencies.
• Example shows rendering via dot -Tpng graph.dot -o graph.png or viz-js.

Not implemented / potential follow-ups

• Parameter/value previews in node labels.
• CLI/helper to render DOT to PNG/SVG.
• Mermaid output.
• Execution-state highlighting in the graph.

Acceptance Criteria (updated)

• DOT export for InferenceGraph via visualize_graph.
• Shapes differentiate inputs/outputs/intermediates.
• Dependencies rendered as edges.
• Branch condition labels included.
• Example in examples/03_tracing.py with rendering guidance.
• Optional: parameter/value previews in node labels.
• Optional: CLI/helper to render DOT to PNG/SVG.
• Optional: Mermaid/interactive visualization.