app.py

import gradio as gr
import json
from graphviz import Digraph
import base64

def generate_concept_map(json_input: str) -> str:
    """
    Generate concept map from JSON and return as base64 image
    
    Args:
        json_input (str): JSON describing the concept map structure.
        
    Returns:
        str: Base64 data URL of the generated concept map
    """
    try:
        if not json_input.strip():
            return "Error: Empty input"
            
        data = json.loads(json_input)
        
        if 'central_node' not in data or 'nodes' not in data:
            raise ValueError("Missing required fields: central_node or nodes")

        # Create graph
        dot = Digraph(
            name='ConceptMap',
            format='png',
            graph_attr={
                'rankdir': 'TB',
                'splines': 'ortho',
                'bgcolor': 'transparent'
            }
        )
        
        # Central node (ellipse)
        dot.node(
            'central',
            data['central_node'],
            shape='ellipse',
            style='filled',
            fillcolor='#2196F3',
            fontcolor='white',
            fontsize='14'
        )
        
        # Process nodes (rectangles)
        for node in data['nodes']:
            node_id = node.get('id')
            label = node.get('label')
            relationship = node.get('relationship')
            
            # Validate node
            if not all([node_id, label, relationship]):
                raise ValueError(f"Invalid node: {node}")
                
            # Create main node (rectangle)
            dot.node(
                node_id,
                label,
                shape='box',
                style='filled',
                fillcolor='#4CAF50',
                fontcolor='white',
                fontsize='12'
            )
            
            # Connect to central node
            dot.edge(
                'central',
                node_id,
                label=relationship,
                color='#9C27B0',
                fontsize='10'
            )
            
            # Process subnodes (rectangles with lighter fill)
            for subnode in node.get('subnodes', []):
                sub_id = subnode.get('id')
                sub_label = subnode.get('label')
                sub_rel = subnode.get('relationship')
                
                if not all([sub_id, sub_label, sub_rel]):
                    raise ValueError(f"Invalid subnode: {subnode}")
                    
                dot.node(
                    sub_id,
                    sub_label,
                    shape='box',
                    style='filled',
                    fillcolor='#FFA726',
                    fontcolor='white',
                    fontsize='10'
                )
                
                dot.edge(
                    node_id,
                    sub_id,
                    label=sub_rel,
                    color='#E91E63',
                    fontsize='8'
                )

        # Convert to base64 image
        img_data = dot.pipe(format='png')
        img_base64 = base64.b64encode(img_data).decode()
        return f"data:image/png;base64,{img_base64}"

    except json.JSONDecodeError:
        return "Error: Invalid JSON format"
    except Exception as e:
        return f"Error: {str(e)}"

if __name__ == "__main__":
    # Sample JSON for placeholder
    sample_json = """
    {
      "central_node": "Artificial Intelligence (AI)",
      "nodes": [
        {
          "id": "ml",
          "label": "Machine Learning",
          "relationship": "core_component",
          "subnodes": [
            {
              "id": "sl",
              "label": "Supervised Learning",
              "relationship": "type_of",
              "subnodes": [
                {"id": "reg", "label": "Regression", "relationship": "technique"},
                {"id": "clf", "label": "Classification", "relationship": "technique"}
              ]
            },
            {
              "id": "ul",
              "label": "Unsupervised Learning",
              "relationship": "type_of",
              "subnodes": [
                {"id": "clus", "label": "Clustering", "relationship": "technique"},
                {"id": "dim", "label": "Dimensionality Reduction", "relationship": "technique"}
              ]
            }
          ]
        },
        {
          "id": "nlp",
          "label": "Natural Language Processing",
          "relationship": "application_area",
          "subnodes": [
            {
              "id": "sa",
              "label": "Sentiment Analysis",
              "relationship": "task",
              "subnodes": [
                {"id": "tc", "label": "Text Classification", "relationship": "method"},
                {"id": "absa", "label": "Aspect-Based Sentiment Analysis", "relationship": "method"}
              ]
            },
            {
              "id": "tr",
              "label": "Translation",
              "relationship": "task",
              "subnodes": [
                {"id": "nmt", "label": "Neural Machine Translation", "relationship": "method"},
                {"id": "rbt", "label": "Rule-Based Translation", "relationship": "method"}
              ]
            }
          ]
        },
        {
          "id": "cv",
          "label": "Computer Vision",
          "relationship": "application_area",
          "subnodes": [
            {
              "id": "od",
              "label": "Object Detection",
              "relationship": "task",
              "subnodes": [
                {"id": "yolo", "label": "YOLO", "relationship": "algorithm"},
                {"id": "rcnn", "label": "R-CNN", "relationship": "algorithm"}
              ]
            }
          ]
        }
      ]
    }
    """
    
    demo = gr.Interface(
        fn=generate_concept_map,
        inputs=gr.Textbox(
            value=sample_json,  # Pre-filled sample JSON
            placeholder="Paste structured JSON here...",
            label="JSON Input",
            lines=15
        ),
        outputs=gr.Image(
            label="Concept Map",
            type="filepath",
            interactive=False
        ),
        title="Advanced Concept Map Generator",
        description="Create complex concept maps from JSON with direct image output"
    )
    
    demo.launch(
        mcp_server=True,
        share=False,
        server_port=7860,
        server_name="0.0.0.0"
    )