OpenXAI MCP Server

Instructions

Implementation Reference

• index.js:525-562 (handler)

  The core handler function for the 'load_model' tool. Validates the model type, generates a Python code snippet using OpenXAI's LoadModel class, and returns a formatted response with model information and usage example.

    async loadModel(dataName, mlModel, pretrained = true) {
      const modelInfo = {
        ann: 'Artificial Neural Network',
        lr: 'Logistic Regression',
        rf: 'Random Forest',
        svm: 'Support Vector Machine',
        xgb: 'XGBoost'
      };
  
      const modelName = modelInfo[mlModel];
      if (!modelName) {
        throw new Error(`Model type '${mlModel}' not supported. Available models: ${Object.keys(modelInfo).join(', ')}`);
      }
  
      const codeExample = `
  # Example usage with OpenXAI:
  from openxai import LoadModel
  
  # Load the pre-trained model
  model = LoadModel(data_name='${dataName}', ml_model='${mlModel}', pretrained=${pretrained})
  
  # Use the model for predictions
  # predictions = model.predict(input_data)
  `;
  
      return {
        content: [
          {
            type: 'text',
            text: `Model loaded successfully!\n\n` +
                  `Dataset: ${dataName}\n` +
                  `Model type: ${modelName} (${mlModel})\n` +
                  `Pretrained: ${pretrained}\n\n` +
                  `Python code example:\n\`\`\`python${codeExample}\`\`\``
          }
        ]
      };
    }

• index.js:94-113 (schema)

  Input schema defining the parameters for the 'load_model' tool: data_name (required string), ml_model (required enum), pretrained (optional boolean).

  inputSchema: {
    type: 'object',
    properties: {
      data_name: {
        type: 'string',
        description: 'Name of the dataset the model was trained on'
      },
      ml_model: {
        type: 'string',
        description: 'Type of machine learning model (ann, lr, rf, svm, xgb)',
        enum: ['ann', 'lr', 'rf', 'svm', 'xgb']
      },
      pretrained: {
        type: 'boolean',
        description: 'Whether to load a pretrained model',
        default: true
      }
    },
    required: ['data_name', 'ml_model']
  }

• index.js:91-114 (registration)

  Tool registration in the list of available tools returned by ListToolsRequestSchema, including name, description, and input schema.

  {
    name: 'load_model',
    description: 'Load a pre-trained model from OpenXAI',
    inputSchema: {
      type: 'object',
      properties: {
        data_name: {
          type: 'string',
          description: 'Name of the dataset the model was trained on'
        },
        ml_model: {
          type: 'string',
          description: 'Type of machine learning model (ann, lr, rf, svm, xgb)',
          enum: ['ann', 'lr', 'rf', 'svm', 'xgb']
        },
        pretrained: {
          type: 'boolean',
          description: 'Whether to load a pretrained model',
          default: true
        }
      },
      required: ['data_name', 'ml_model']
    }
  },

• index.js:264-265 (handler)

  Dispatch case in the main CallToolRequestSchema handler that routes 'load_model' calls to the loadModel method.

  case 'load_model':
    return await this.loadModel(args.data_name, args.ml_model, args.pretrained);