hf_list_datasets

Instructions

Implementation Reference

• src/tools/datasets.ts:173-196 (handler)

  The MCP tool handler for 'hf_list_datasets': validates arguments using isDatasetSearchArgs, calls client.getDatasets(), and formats the CallToolResult.

  export async function handleListDatasets(client: HuggingFaceClient, args: unknown): Promise<CallToolResult> {
      try {
          if (!isDatasetSearchArgs(args)) {
              throw new Error("Invalid arguments for hf_list_datasets");
          }
  
          const results = await client.getDatasets(args as Record<string, any>);
          
          return {
              content: [{ type: "text", text: results }],
              isError: false,
          };
      } catch (error) {
          return {
              content: [
                  {
                      type: "text",
                      text: `Error: ${error instanceof Error ? error.message : String(error)}`,
                  },
              ],
              isError: true,
          };
      }
  }

• src/tools/datasets.ts:8-51 (schema)

  The tool definition for 'hf_list_datasets' including name, description, and detailed inputSchema for filtering and pagination parameters.

  export const listDatasetsToolDefinition: Tool = {
      name: "hf_list_datasets",
      description:
          "Get information from all datasets in the Hub. Supports filtering by search terms, authors, tags, and more. " +
          "Returns paginated results with dataset metadata including downloads, likes, and tags.",
      inputSchema: {
          type: "object", 
          properties: {
              search: {
                  type: "string",
                  description: "Filter based on substrings for repos and their usernames (e.g., 'pets', 'microsoft')"
              },
              author: {
                  type: "string",
                  description: "Filter datasets by author or organization (e.g., 'huggingface', 'microsoft')" 
              },
              filter: {
                  type: "string",
                  description: "Filter based on tags (e.g., 'task_categories:text-classification', 'languages:en')"
              },
              sort: {
                  type: "string",
                  description: "Property to use when sorting (e.g., 'downloads', 'author')"
              },
              direction: {
                  type: "string",
                  description: "Sort direction: '-1' for descending, anything else for ascending"
              },
              limit: {
                  type: "number", 
                  description: "Limit the number of datasets fetched"
              },
              full: {
                  type: "boolean",
                  description: "Whether to fetch most dataset data including all tags and files"
              },
              config: {
                  type: "boolean",
                  description: "Whether to also fetch the repo config"
              }
          },
          required: []
      }
  };

• src/client.ts:70-77 (helper)

  Core helper method in HuggingFaceClient: performs HTTP GET to Hugging Face Hub API '/api/datasets' endpoint with query params, returns pretty-printed JSON string of the response data.

  async getDatasets(params: Record<string, any> = {}): Promise<string> {
      try {
          const response: AxiosResponse = await this.httpClient.get('/api/datasets', { params });
          return JSON.stringify(response.data, null, 2);
      } catch (error) {
          throw new Error(`Failed to fetch datasets: ${error instanceof Error ? error.message : String(error)}`);
      }
  }

• src/server.ts:81-82 (registration)

  Registration and dispatch of 'hf_list_datasets' handler in the main HuggingFaceServer's CallToolRequestHandler switch statement.

  case 'hf_list_datasets':
      return handleListDatasets(this.client, args);

• src/server.ts:55-66 (registration)

  Registration of 'hf_list_datasets' tool definition (listDatasetsToolDefinition) in the ListToolsRequestHandler for tool discovery.

  this.server.setRequestHandler(ListToolsRequestSchema, async () => ({
      tools: [
          listModelsToolDefinition,
          getModelInfoToolDefinition,
          getModelTagsToolDefinition,
          listDatasetsToolDefinition,
          getDatasetInfoToolDefinition,
          getDatasetParquetToolDefinition,
          getCroissantToolDefinition,
          getDatasetTagsToolDefinition
      ],
  }));