import type { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import type {
CallToolResult,
ToolAnnotations,
} from "@modelcontextprotocol/sdk/types.js";
import state from "@/config.js";
import { createToolError } from "@/utils/createToolError.js";
import { getGlobalVaultManager } from "@/utils/getVaultManager.js";
import { execute as writePropertyExecute } from "../write_property/index.js";
import {
type CreateDocumentWithPropertiesParams,
createDocumentWithPropertiesParamsSchema,
} from "./params.js";
export const name = "create_document_with_properties";
export const annotations: ToolAnnotations = {
title: "Create Document with Properties",
openWorldHint: true,
};
export const description = `
Starts and completes a two-step workflow for AI-generated frontmatter properties.
Step 1: Call this tool with sourcePath (and optional outputPath). It returns a structured instruction payload and a content preview for AI analysis.
Step 2: Call this same tool again with aiGeneratedProperties. The tool then writes those properties by executing the same write logic used by the 'write_property' tool.
Use this tool when an AI agent should orchestrate analysis and write in a consistent workflow.
`;
export const register = (mcpServer: McpServer) => {
mcpServer.registerTool(
name,
{
title: annotations.title || name,
description: description,
inputSchema: createDocumentWithPropertiesParamsSchema.shape,
annotations: annotations,
},
execute,
);
};
export const execute = async (
params: CreateDocumentWithPropertiesParams,
): Promise<CallToolResult> => {
const vaultDirPath = state.vaultPath;
if (!vaultDirPath) {
return createToolError(
"VAULT_DIR_PATH environment variable is not set",
"Set VAULT_DIR_PATH to your Obsidian vault directory",
);
}
let vaultManager = null;
try {
vaultManager = getGlobalVaultManager();
} catch (e) {
return createToolError((e as Error).message);
}
try {
const targetPath = params.outputPath || params.sourcePath;
if (params.aiGeneratedProperties) {
const writeResult = await writePropertyExecute({
filePath: targetPath,
properties: params.aiGeneratedProperties,
quiet: params.quiet || false,
});
return writeResult;
}
const document = await vaultManager.getDocumentInfo(params.sourcePath);
if (document === null) {
return createToolError(
`Source document not found: ${params.sourcePath}`,
"Verify the file path and ensure the file exists. Use the vault tool with 'list_all' action to see all available files.",
);
}
const instructionForAI = {
ai_prompt:
"Your task is to analyze the provided text content and create a JSON object of document properties based on the requested schema. After creating the JSON object, you MUST call the 'create_document_with_properties' tool again, passing back the original parameters along with the new 'aiGeneratedProperties' parameter containing your generated JSON.",
purpose:
"Analyze document content and return a structured JSON object of properties.",
content_to_analyze: {
source_path: params.sourcePath,
content_preview:
document.content.substring(0, 2000) +
(document.content.length > 2000 ? "..." : ""),
},
// AI가 다시 호출해야 할 다음 작업에 대한 명확한 명세
next_action_required: {
tool_to_call: name, // 바로 이 도구('create_document_with_properties')를 다시 호출
parameters_to_add: {
sourcePath: params.sourcePath,
outputPath: params.outputPath,
overwrite: params.overwrite,
quiet: params.quiet,
aiGeneratedProperties: "<- Insert your generated JSON object here.",
},
example_of_next_call: {
tool_name: name,
parameters: {
sourcePath: params.sourcePath,
aiGeneratedProperties: {
title: "Serverless 환경에서 I/O 처리 최적화 경험기",
date: "2025-04-03",
tags: ["serverless", "optimization"],
summary:
"Promise.all, Worker를 벤치마크하며 서버리스 환경에서의 I/O 처리 최적화 경험기를 공유합니다.",
slug: "serverless-io-optimization",
category: "code",
completed: true,
},
},
},
},
};
return {
isError: false,
content: [
{
type: "text",
text: JSON.stringify(instructionForAI, null, 2),
},
],
};
} catch (error) {
return createToolError(
`Failed to create workflow instruction: ${error instanceof Error ? error.message : String(error)}`,
);
}
};
export default {
name,
description,
annotations,
inputSchema: createDocumentWithPropertiesParamsSchema.shape,
execute,
register,
};
