list_explainers

Instructions

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "list_explainers"
  }
}

Implementation Reference

• index.js:564-625 (handler)

  The main execution function for the list_explainers tool. Filters and returns available explanation methods (LIME, SHAP, etc.) with their details based on the method_type parameter.

  async listExplainers(methodType) {
    const explainers = {
      lime: {
        name: 'LIME (Local Interpretable Model-agnostic Explanations)',
        description: 'Local explanations by approximating the model locally with an interpretable model',
        supported_data_types: ['tabular', 'image', 'text'],
        explanation_type: 'local',
        model_agnostic: true
      },
      shap: {
        name: 'SHAP (SHapley Additive exPlanations)',
        description: 'Feature attribution based on cooperative game theory',
        supported_data_types: ['tabular', 'image', 'text'],
        explanation_type: 'local',
        model_agnostic: true
      },
      integrated_gradients: {
        name: 'Integrated Gradients',
        description: 'Attribution method based on gradients integrated along a path',
        supported_data_types: ['tabular', 'image', 'text'],
        explanation_type: 'local',
        model_agnostic: false,
        requires: 'PyTorch or TensorFlow model'
      },
      gradcam: {
        name: 'Grad-CAM (Gradient-weighted Class Activation Mapping)',
        description: 'Visual explanations for CNN models using gradients',
        supported_data_types: ['image'],
        explanation_type: 'local',
        model_agnostic: false,
        requires: 'CNN model'
      },
      guided_backprop: {
        name: 'Guided Backpropagation',
        description: 'Modified backpropagation for generating visual explanations',
        supported_data_types: ['image'],
        explanation_type: 'local',
        model_agnostic: false,
        requires: 'Neural network model'
      }
    };
  
    let result = [];
    if (methodType === 'all') {
      result = Object.entries(explainers).map(([key, value]) => ({
        method: key,
        ...value
      }));
    } else {
      result = explainers[methodType] ? [{ method: methodType, ...explainers[methodType] }] : [];
    }
  
    return {
      content: [
        {
          type: 'text',
          text: `Available OpenXAI explanation methods:\n\n` +
                JSON.stringify(result, null, 2)
        }
      ]
    };
  }

• index.js:118-128 (schema)

  JSON schema defining the input parameters for the list_explainers tool, including the optional method_type filter.

  inputSchema: {
    type: 'object',
    properties: {
      method_type: {
        type: 'string',
        description: 'Filter by explanation method type',
        enum: ['lime', 'shap', 'integrated_gradients', 'gradcam', 'all']
      }
    },
    required: []
  }

• index.js:115-129 (registration)

  Registration of the list_explainers tool in the ListToolsRequestHandler response, providing name, description, and schema.

  {
    name: 'list_explainers',
    description: 'List available explanation methods in OpenXAI',
    inputSchema: {
      type: 'object',
      properties: {
        method_type: {
          type: 'string',
          description: 'Filter by explanation method type',
          enum: ['lime', 'shap', 'integrated_gradients', 'gradcam', 'all']
        }
      },
      required: []
    }
  },

• index.js:267-268 (registration)

  Dispatch registration in the CallToolRequestHandler switch statement, calling the listExplainers handler with parameters.

  case 'list_explainers':
    return await this.listExplainers(args.method_type || 'all');