activo 0.4.3 → 0.5.0

package/src/core/tools/standards.ts

@@ -1,889 +0,0 @@
-import fs from "fs";
-import path from "path";
-import crypto from "crypto";
-import pdfParse from "pdf-parse";
-import { toMarkdown as hwpToMarkdown } from "@ohah/hwpjs";
-import { Tool, ToolResult } from "./types.js";
-import { OllamaClient } from "../llm/ollama.js";
-import { loadConfig } from "../config.js";
-
-// RAG constants
-const STANDARDS_EMBEDDINGS_DIR = ".activo/standards-rag";
-const DEFAULT_EMBED_MODEL = "nomic-embed-text";
-
-// RAG interfaces
-interface StandardsChunk {
-  filepath: string;
-  section: string;
-  ruleId?: string;
-  content: string;
-}
-
-interface StandardsEmbedding {
-  chunk: StandardsChunk;
-  embedding: number[];
-  hash: string;
-}
-
-interface StandardsIndex {
-  version: string;
-  model: string;
-  createdAt: string;
-  updatedAt: string;
-  totalChunks: number;
-}
-
-// RAG helper functions
-function getStandardsRagDir(): string {
-  return path.resolve(process.cwd(), STANDARDS_EMBEDDINGS_DIR);
-}
-
-function ensureStandardsRagDir(): void {
-  const dir = getStandardsRagDir();
-  if (!fs.existsSync(dir)) {
-    fs.mkdirSync(dir, { recursive: true });
-  }
-}
-
-function getStandardsIndexPath(): string {
-  return path.join(getStandardsRagDir(), "index.json");
-}
-
-function getStandardsDataPath(): string {
-  return path.join(getStandardsRagDir(), "embeddings.json");
-}
-
-function loadStandardsIndex(): StandardsIndex | null {
-  const indexPath = getStandardsIndexPath();
-  if (fs.existsSync(indexPath)) {
-    try {
-      return JSON.parse(fs.readFileSync(indexPath, "utf-8"));
-    } catch {
-      return null;
-    }
-  }
-  return null;
-}
-
-function saveStandardsIndex(index: StandardsIndex): void {
-  ensureStandardsRagDir();
-  fs.writeFileSync(getStandardsIndexPath(), JSON.stringify(index, null, 2));
-}
-
-function loadStandardsEmbeddings(): StandardsEmbedding[] {
-  const dataPath = getStandardsDataPath();
-  if (fs.existsSync(dataPath)) {
-    try {
-      return JSON.parse(fs.readFileSync(dataPath, "utf-8"));
-    } catch {
-      return [];
-    }
-  }
-  return [];
-}
-
-function saveStandardsEmbeddings(data: StandardsEmbedding[]): void {
-  ensureStandardsRagDir();
-  fs.writeFileSync(getStandardsDataPath(), JSON.stringify(data));
-}
-
-function calculateHash(content: string): string {
-  return crypto.createHash("md5").update(content).digest("hex");
-}
-
-// Split markdown into semantic chunks (by sections/rules)
-function splitStandardsIntoChunks(content: string, filepath: string): StandardsChunk[] {
-  const chunks: StandardsChunk[] = [];
-  const lines = content.split("\n");
-
-  let currentSection = "";
-  let currentRuleId: string | undefined;
-  let currentContent: string[] = [];
-  let inRule = false;
-
-  for (const line of lines) {
-    // Detect rule pattern: ## RULE-XXX: Title
-    const ruleMatch = line.match(/^##\s+(RULE-\d+):\s*(.+)/i);
-    // Detect section headers
-    const sectionMatch = line.match(/^#+\s+(.+)/);
-
-    if (ruleMatch) {
-      // Save previous chunk
-      if (currentContent.length > 0) {
-        const text = currentContent.join("\n").trim();
-        if (text) {
-          chunks.push({
-            filepath,
-            section: currentSection,
-            ruleId: currentRuleId,
-            content: text,
-          });
-        }
-      }
-      currentRuleId = ruleMatch[1];
-      currentSection = ruleMatch[2];
-      currentContent = [line];
-      inRule = true;
-    } else if (sectionMatch && !inRule) {
-      // Save previous chunk
-      if (currentContent.length > 0) {
-        const text = currentContent.join("\n").trim();
-        if (text) {
-          chunks.push({
-            filepath,
-            section: currentSection,
-            ruleId: currentRuleId,
-            content: text,
-          });
-        }
-      }
-      currentSection = sectionMatch[1];
-      currentRuleId = undefined;
-      currentContent = [line];
-    } else if (line.match(/^##/) && inRule) {
-      // End of rule, start new section
-      if (currentContent.length > 0) {
-        const text = currentContent.join("\n").trim();
-        if (text) {
-          chunks.push({
-            filepath,
-            section: currentSection,
-            ruleId: currentRuleId,
-            content: text,
-          });
-        }
-      }
-      inRule = false;
-      currentRuleId = undefined;
-      const newSection = line.match(/^#+\s+(.+)/);
-      currentSection = newSection ? newSection[1] : "";
-      currentContent = [line];
-    } else {
-      currentContent.push(line);
-    }
-  }
-
-  // Save remaining content
-  if (currentContent.length > 0) {
-    const text = currentContent.join("\n").trim();
-    if (text) {
-      chunks.push({
-        filepath,
-        section: currentSection,
-        ruleId: currentRuleId,
-        content: text,
-      });
-    }
-  }
-
-  // Filter out small chunks (less than 50 chars)
-  return chunks.filter(c => c.content.length >= 50);
-}
-
-// Cosine similarity
-function cosineSimilarity(a: number[], b: number[]): number {
-  if (a.length !== b.length) return 0;
-  let dotProduct = 0;
-  let normA = 0;
-  let normB = 0;
-  for (let i = 0; i < a.length; i++) {
-    dotProduct += a[i] * b[i];
-    normA += a[i] * a[i];
-    normB += b[i] * b[i];
-  }
-  return dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));
-}
-
-// Resolve natural language directory paths
-function resolveOutputDir(outputDir: string | undefined): string {
-  if (!outputDir) {
-    return path.resolve(".activo/standards");
-  }
-
-  const normalized = outputDir.trim().toLowerCase();
-
-  // Natural language mappings
-  const currentDirPatterns = [
-    "현재 디렉토리", "현재 폴더", "현재디렉토리", "현재폴더",
-    "이 디렉토리", "이 폴더", "여기", "here",
-    "current directory", "current folder", "current dir",
-    ".", "./"
-  ];
-
-  for (const pattern of currentDirPatterns) {
-    if (normalized.includes(pattern)) {
-      return path.resolve(".");
-    }
-  }
-
-  // Handle ~ for home directory
-  if (outputDir.startsWith("~")) {
-    return path.resolve(outputDir.replace("~", process.env.HOME || ""));
-  }
-
-  return path.resolve(outputDir);
-}
-
-// Import PDF Tool
-export const importPdfTool: Tool = {
-  name: "import_pdf_standards",
-  description: "Import development standards from a PDF file and convert to markdown (PDF를 마크다운으로 변환). Use when user asks: 'PDF 변환', 'PDF를 md로', 'PDF 마크다운', 'convert PDF', '현재 디렉토리에 저장'.",
-  parameters: {
-    type: "object",
-    required: ["pdfPath"],
-    properties: {
-      pdfPath: {
-        type: "string",
-        description: "Path to the PDF file (PDF 파일 경로)",
-      },
-      outputDir: {
-        type: "string",
-        description: "Output directory. Use '.' or '현재 디렉토리' for current dir. Default: .activo/standards",
-      },
-      singleFile: {
-        type: "boolean",
-        description: "Save as single markdown file instead of chunks (default: false)",
-      },
-    },
-  },
-  handler: async (args): Promise<ToolResult> => {
-    try {
-      // Resolve PDF path (handle ~)
-      let pdfPath = args.pdfPath as string;
-      if (pdfPath.startsWith("~")) {
-        pdfPath = pdfPath.replace("~", process.env.HOME || "");
-      }
-      pdfPath = path.resolve(pdfPath);
-
-      const outputDir = resolveOutputDir(args.outputDir as string | undefined);
-      const singleFile = args.singleFile as boolean || false;
-
-      if (!fs.existsSync(pdfPath)) {
-        return { success: false, content: "", error: `PDF not found: ${pdfPath}` };
-      }
-
-      // Extract text from PDF using pdf-parse
-      const dataBuffer = fs.readFileSync(pdfPath);
-      const pdfData = await pdfParse(dataBuffer);
-      const text = pdfData.text;
-      const numpages = pdfData.numpages;
-
-      // Create output directory
-      if (!fs.existsSync(outputDir)) {
-        fs.mkdirSync(outputDir, { recursive: true });
-      }
-
-      const filename = path.basename(pdfPath, ".pdf");
-      const extractionDate = new Date().toISOString().split("T")[0];
-      const results: string[] = [];
-
-      if (singleFile) {
-        // Save as single file
-        const outputFilename = `${sanitize(filename)}.md`;
-        const outputPath = path.join(outputDir, outputFilename);
-
-        let md = `# ${filename}\n\n`;
-        md += `> Source: ${path.basename(pdfPath)}\n`;
-        md += `> Extracted: ${extractionDate}\n`;
-        md += `> Pages: ${numpages}\n`;
-        md += `> Method: pdf-parse\n\n`;
-        md += `---\n\n`;
-        md += text;
-        md += `\n\n---\n`;
-        md += `[Edit this file to add structured rules]\n`;
-
-        fs.writeFileSync(outputPath, md, "utf-8");
-        results.push(outputFilename);
-      } else {
-        // Split into chunks
-        const chunks = splitIntoChunks(text, 3000);
-
-        for (let i = 0; i < chunks.length; i++) {
-          const chunkFilename = `${String(i + 1).padStart(2, "0")}_${sanitize(filename)}.md`;
-          const chunkPath = path.join(outputDir, chunkFilename);
-
-          let md = `# ${filename} - Part ${i + 1}\n\n`;
-          md += `> Source: ${path.basename(pdfPath)}\n`;
-          md += `> Extracted: ${extractionDate}\n`;
-          md += `> Pages: ${numpages}\n`;
-          md += `> Method: pdf-parse\n\n`;
-          md += `---\n\n`;
-          md += chunks[i];
-          md += `\n\n---\n`;
-          md += `[Edit this file to add structured rules]\n`;
-
-          fs.writeFileSync(chunkPath, md, "utf-8");
-          results.push(chunkFilename);
-        }
-
-        // Create index
-        const indexPath = path.join(outputDir, "_index.md");
-        let indexMd = `# Development Standards Index\n\n`;
-        indexMd += `> Source: ${path.basename(pdfPath)}\n`;
-        indexMd += `> Extracted: ${extractionDate}\n`;
-        indexMd += `> Files: ${results.length}\n`;
-        indexMd += `> Method: pdf-parse\n\n`;
-        indexMd += `## Files\n\n`;
-        for (const r of results) {
-          indexMd += `- [${r}](./${r})\n`;
-        }
-        fs.writeFileSync(indexPath, indexMd, "utf-8");
-      }
-
-      return {
-        success: true,
-        content: `PDF 변환 완료!\n\n` +
-          `📂 저장 위치: ${outputDir}\n` +
-          `📄 파일 수: ${results.length}\n` +
-          `📖 페이지: ${numpages}\n` +
-          `🔧 추출 방법: pdf-parse\n\n` +
-          `파일 목록:\n${results.map(r => `  - ${r}`).join("\n")}\n\n` +
-          `다음 단계: 파일을 편집하여 구조화된 규칙 추가\n` +
-          `형식: ## RULE-001: Title\n- 심각도: error|warning|info\n- 규칙: description`,
-      };
-    } catch (error) {
-      return { success: false, content: "", error: String(error) };
-    }
-  },
-};
-
-// List Standards Tool
-export const listStandardsTool: Tool = {
-  name: "list_standards",
-  description: "List all loaded development standards and rules.",
-  parameters: {
-    type: "object",
-    properties: {
-      directory: {
-        type: "string",
-        description: "Standards directory (default: .activo/standards)",
-      },
-    },
-  },
-  handler: async (args): Promise<ToolResult> => {
-    try {
-      const dir = resolveOutputDir(args.directory as string | undefined);
-
-      if (!fs.existsSync(dir)) {
-        return { success: true, content: "No standards directory found. Import a PDF first." };
-      }
-
-      const files = fs.readdirSync(dir).filter((f) => f.endsWith(".md") && f !== "_index.md");
-      if (files.length === 0) {
-        return { success: true, content: "No standard files found." };
-      }
-
-      let totalRules = 0;
-      const results: string[] = [];
-
-      for (const file of files) {
-        const content = fs.readFileSync(path.join(dir, file), "utf-8");
-        const rules = content.match(/^## [A-Z]+-\d+/gm) || [];
-        totalRules += rules.length;
-        results.push(`📄 ${file}: ${rules.length} rules`);
-      }
-
-      return {
-        success: true,
-        content: `Standards Directory: ${dir}\n\n${results.join("\n")}\n\nTotal: ${files.length} files, ${totalRules} rules`,
-      };
-    } catch (error) {
-      return { success: false, content: "", error: String(error) };
-    }
-  },
-};
-
-// Check Code Quality Tool
-export const checkQualityTool: Tool = {
-  name: "check_code_quality",
-  description: "Check code against loaded development standards.",
-  parameters: {
-    type: "object",
-    required: ["filepath"],
-    properties: {
-      filepath: {
-        type: "string",
-        description: "File or directory to check",
-      },
-      standardsDir: {
-        type: "string",
-        description: "Standards directory (default: .activo/standards)",
-      },
-    },
-  },
-  handler: async (args): Promise<ToolResult> => {
-    try {
-      const filepath = path.resolve(args.filepath as string);
-      const standardsDir = resolveOutputDir(args.standardsDir as string | undefined);
-
-      if (!fs.existsSync(filepath)) {
-        return { success: false, content: "", error: `Path not found: ${filepath}` };
-      }
-
-      // Load standards
-      let standards = "";
-      if (fs.existsSync(standardsDir)) {
-        const files = fs.readdirSync(standardsDir).filter((f) => f.endsWith(".md") && f !== "_index.md");
-        for (const file of files) {
-          standards += fs.readFileSync(path.join(standardsDir, file), "utf-8") + "\n\n";
-        }
-      }
-
-      // Get code
-      let code = "";
-      const stat = fs.statSync(filepath);
-      if (stat.isFile()) {
-        code = fs.readFileSync(filepath, "utf-8");
-      } else {
-        return { success: false, content: "", error: "Directory check not yet supported. Specify a file." };
-      }
-
-      // Build analysis prompt
-      const prompt = buildAnalysisPrompt(code, filepath, standards);
-
-      // Call Ollama
-      const config = loadConfig();
-      const client = new OllamaClient(config.ollama);
-
-      const response = await client.chat([{ role: "user", content: prompt }]);
-
-      return { success: true, content: response.content };
-    } catch (error) {
-      return { success: false, content: "", error: String(error) };
-    }
-  },
-};
-
-// Helper functions
-function splitIntoChunks(text: string, maxSize: number): string[] {
-  const chunks: string[] = [];
-  const paragraphs = text.split(/\n\n+/);
-  let current = "";
-
-  for (const para of paragraphs) {
-    if (current.length + para.length > maxSize && current.length > 0) {
-      chunks.push(current.trim());
-      current = "";
-    }
-    current += para + "\n\n";
-  }
-
-  if (current.trim()) {
-    chunks.push(current.trim());
-  }
-
-  return chunks.length > 0 ? chunks : [text];
-}
-
-function sanitize(name: string): string {
-  return name.replace(/[^a-zA-Z0-9가-힣\s-]/g, "").replace(/\s+/g, "_").slice(0, 50);
-}
-
-function buildAnalysisPrompt(code: string, filepath: string, standards: string): string {
-  const ext = path.extname(filepath);
-  const lang = { ".ts": "typescript", ".js": "javascript", ".java": "java", ".py": "python" }[ext] || "text";
-
-  let prompt = `당신은 코드 품질 전문가입니다. 아래 코드를 분석하세요.\n\n`;
-
-  if (standards) {
-    prompt += `[개발 표준 규칙]\n${standards.slice(0, 4000)}\n\n`;
-  }
-
-  prompt += `[분석 대상 코드]\n파일: ${filepath}\n\`\`\`${lang}\n${code.slice(0, 8000)}\n\`\`\`\n\n`;
-  prompt += `[점검 요청]\n`;
-  prompt += `1. 규칙 위반 사항 (있다면)\n`;
-  prompt += `2. 개선 제안\n`;
-  prompt += `3. 전반적인 코드 품질 평가\n`;
-
-  return prompt;
-}
-
-// Import HWP Tool
-export const importHwpTool: Tool = {
-  name: "import_hwp_standards",
-  description: "Import development standards from a HWP file and convert to markdown (HWP를 마크다운으로 변환). Use when user asks: 'HWP 변환', 'HWP를 md로', '한글 파일 변환', 'convert HWP'.",
-  parameters: {
-    type: "object",
-    required: ["hwpPath"],
-    properties: {
-      hwpPath: {
-        type: "string",
-        description: "Path to the HWP file (HWP 파일 경로)",
-      },
-      outputDir: {
-        type: "string",
-        description: "Output directory. Use '.' or '현재 디렉토리' for current dir. Default: .activo/standards",
-      },
-      singleFile: {
-        type: "boolean",
-        description: "Save as single markdown file instead of chunks (default: false)",
-      },
-    },
-  },
-  handler: async (args): Promise<ToolResult> => {
-    try {
-      // Resolve HWP path (handle ~)
-      let hwpPath = args.hwpPath as string;
-      if (hwpPath.startsWith("~")) {
-        hwpPath = hwpPath.replace("~", process.env.HOME || "");
-      }
-      hwpPath = path.resolve(hwpPath);
-
-      const outputDir = resolveOutputDir(args.outputDir as string | undefined);
-      const singleFile = args.singleFile as boolean || false;
-
-      if (!fs.existsSync(hwpPath)) {
-        return { success: false, content: "", error: `HWP not found: ${hwpPath}` };
-      }
-
-      // Parse HWP file using @ohah/hwpjs
-      const dataBuffer = fs.readFileSync(hwpPath);
-      const { markdown: text } = hwpToMarkdown(dataBuffer, {
-        image: "base64",
-        useHtml: false,
-        includeVersion: false,
-        includePageInfo: false,
-      });
-
-      // Create output directory
-      if (!fs.existsSync(outputDir)) {
-        fs.mkdirSync(outputDir, { recursive: true });
-      }
-
-      const filename = path.basename(hwpPath, ".hwp");
-      const extractionDate = new Date().toISOString().split("T")[0];
-      const results: string[] = [];
-
-      if (singleFile) {
-        // Save as single file
-        const outputFilename = `${sanitize(filename)}.md`;
-        const outputPath = path.join(outputDir, outputFilename);
-
-        let md = `# ${filename}\n\n`;
-        md += `> Source: ${path.basename(hwpPath)}\n`;
-        md += `> Extracted: ${extractionDate}\n`;
-        md += `> Format: HWP\n`;
-        md += `> Method: hwp.js\n\n`;
-        md += `---\n\n`;
-        md += text;
-        md += `\n\n---\n`;
-        md += `[Edit this file to add structured rules]\n`;
-
-        fs.writeFileSync(outputPath, md, "utf-8");
-        results.push(outputFilename);
-      } else {
-        // Split into chunks
-        const chunks = splitIntoChunks(text, 3000);
-
-        for (let i = 0; i < chunks.length; i++) {
-          const chunkFilename = `${String(i + 1).padStart(2, "0")}_${sanitize(filename)}.md`;
-          const chunkPath = path.join(outputDir, chunkFilename);
-
-          let md = `# ${filename} - Part ${i + 1}\n\n`;
-          md += `> Source: ${path.basename(hwpPath)}\n`;
-          md += `> Extracted: ${extractionDate}\n`;
-          md += `> Format: HWP\n`;
-          md += `> Method: hwp.js\n\n`;
-          md += `---\n\n`;
-          md += chunks[i];
-          md += `\n\n---\n`;
-          md += `[Edit this file to add structured rules]\n`;
-
-          fs.writeFileSync(chunkPath, md, "utf-8");
-          results.push(chunkFilename);
-        }
-
-        // Create index
-        const indexPath = path.join(outputDir, "_index.md");
-        let indexMd = `# Development Standards Index\n\n`;
-        indexMd += `> Source: ${path.basename(hwpPath)}\n`;
-        indexMd += `> Extracted: ${extractionDate}\n`;
-        indexMd += `> Files: ${results.length}\n`;
-        indexMd += `> Method: hwp.js\n\n`;
-        indexMd += `## Files\n\n`;
-        for (const r of results) {
-          indexMd += `- [${r}](./${r})\n`;
-        }
-        fs.writeFileSync(indexPath, indexMd, "utf-8");
-      }
-
-      return {
-        success: true,
-        content: `HWP 변환 완료!\n\n` +
-          `📂 저장 위치: ${outputDir}\n` +
-          `📄 파일 수: ${results.length}\n` +
-          `📖 형식: HWP\n` +
-          `🔧 추출 방법: hwp.js\n\n` +
-          `파일 목록:\n${results.map(r => `  - ${r}`).join("\n")}\n\n` +
-          `다음 단계: 파일을 편집하여 구조화된 규칙 추가\n` +
-          `형식: ## RULE-001: Title\n- 심각도: error|warning|info\n- 규칙: description`,
-      };
-    } catch (error) {
-      return { success: false, content: "", error: String(error) };
-    }
-  },
-};
-
-// Index Standards for RAG
-export const indexStandardsTool: Tool = {
-  name: "index_standards",
-  description: "Index development standards for RAG search. Run this after importing PDF/HWP files to enable semantic search.",
-  parameters: {
-    type: "object",
-    properties: {
-      directory: {
-        type: "string",
-        description: "Standards directory (default: .activo/standards)",
-      },
-    },
-  },
-  handler: async (args): Promise<ToolResult> => {
-    try {
-      const dir = resolveOutputDir(args.directory as string | undefined);
-
-      if (!fs.existsSync(dir)) {
-        return { success: false, content: "", error: `Standards directory not found: ${dir}` };
-      }
-
-      const files = fs.readdirSync(dir).filter((f) => f.endsWith(".md") && f !== "_index.md");
-      if (files.length === 0) {
-        return { success: false, content: "", error: "No markdown files found in standards directory" };
-      }
-
-      const config = loadConfig();
-      const client = new OllamaClient(config.ollama);
-
-      const allChunks: StandardsChunk[] = [];
-
-      // Collect all chunks from all files
-      for (const file of files) {
-        const content = fs.readFileSync(path.join(dir, file), "utf-8");
-        const chunks = splitStandardsIntoChunks(content, file);
-        allChunks.push(...chunks);
-      }
-
-      if (allChunks.length === 0) {
-        return { success: false, content: "", error: "No valid chunks found in standards files" };
-      }
-
-      // Generate embeddings
-      const embeddings: StandardsEmbedding[] = [];
-      let processed = 0;
-
-      for (const chunk of allChunks) {
-        const embedding = await client.embed(chunk.content, DEFAULT_EMBED_MODEL);
-        embeddings.push({
-          chunk,
-          embedding,
-          hash: calculateHash(chunk.content),
-        });
-        processed++;
-      }
-
-      // Save embeddings
-      saveStandardsEmbeddings(embeddings);
-
-      // Save index
-      const index: StandardsIndex = {
-        version: "1.0",
-        model: DEFAULT_EMBED_MODEL,
-        createdAt: new Date().toISOString(),
-        updatedAt: new Date().toISOString(),
-        totalChunks: embeddings.length,
-      };
-      saveStandardsIndex(index);
-
-      return {
-        success: true,
-        content: `Standards indexed successfully!\n\n` +
-          `📂 Directory: ${dir}\n` +
-          `📄 Files: ${files.length}\n` +
-          `🔖 Chunks: ${embeddings.length}\n` +
-          `🧠 Model: ${DEFAULT_EMBED_MODEL}\n\n` +
-          `Use 'search_standards' to find relevant rules.`,
-      };
-    } catch (error) {
-      return { success: false, content: "", error: String(error) };
-    }
-  },
-};
-
-// Search Standards using RAG
-export const searchStandardsTool: Tool = {
-  name: "search_standards",
-  description: "Search development standards using semantic search (RAG). Returns relevant rules/sections based on query.",
-  parameters: {
-    type: "object",
-    required: ["query"],
-    properties: {
-      query: {
-        type: "string",
-        description: "Search query (e.g., 'variable naming', 'error handling', 'SQL injection')",
-      },
-      topK: {
-        type: "number",
-        description: "Number of results to return (default: 5)",
-      },
-    },
-  },
-  handler: async (args): Promise<ToolResult> => {
-    try {
-      const query = args.query as string;
-      const topK = (args.topK as number) || 5;
-
-      const embeddings = loadStandardsEmbeddings();
-      if (embeddings.length === 0) {
-        return {
-          success: false,
-          content: "",
-          error: "No standards indexed. Run 'index_standards' first.",
-        };
-      }
-
-      const config = loadConfig();
-      const client = new OllamaClient(config.ollama);
-
-      // Get query embedding
-      const queryEmbedding = await client.embed(query, DEFAULT_EMBED_MODEL);
-
-      // Calculate similarities
-      const results = embeddings.map((e) => ({
-        ...e,
-        similarity: cosineSimilarity(queryEmbedding, e.embedding),
-      }));
-
-      // Sort by similarity and get top K
-      results.sort((a, b) => b.similarity - a.similarity);
-      const topResults = results.slice(0, topK);
-
-      // Format results
-      let output = `## Search Results for: "${query}"\n\n`;
-      output += `Found ${topResults.length} relevant standards:\n\n`;
-
-      for (let i = 0; i < topResults.length; i++) {
-        const r = topResults[i];
-        output += `### ${i + 1}. ${r.chunk.ruleId || r.chunk.section || "Section"}\n`;
-        output += `📄 File: ${r.chunk.filepath}\n`;
-        output += `📊 Relevance: ${(r.similarity * 100).toFixed(1)}%\n\n`;
-        output += `${r.chunk.content.slice(0, 500)}${r.chunk.content.length > 500 ? "..." : ""}\n\n`;
-        output += `---\n\n`;
-      }
-
-      return { success: true, content: output };
-    } catch (error) {
-      return { success: false, content: "", error: String(error) };
-    }
-  },
-};
-
-// Check Code Quality with RAG
-export const checkQualityWithRagTool: Tool = {
-  name: "check_quality_rag",
-  description: "Check code quality using RAG to find relevant standards automatically.",
-  parameters: {
-    type: "object",
-    required: ["filepath"],
-    properties: {
-      filepath: {
-        type: "string",
-        description: "File to check",
-      },
-      topK: {
-        type: "number",
-        description: "Number of relevant standards to use (default: 5)",
-      },
-    },
-  },
-  handler: async (args): Promise<ToolResult> => {
-    try {
-      const filepath = path.resolve(args.filepath as string);
-      const topK = (args.topK as number) || 5;
-
-      if (!fs.existsSync(filepath)) {
-        return { success: false, content: "", error: `File not found: ${filepath}` };
-      }
-
-      const embeddings = loadStandardsEmbeddings();
-      if (embeddings.length === 0) {
-        return {
-          success: false,
-          content: "",
-          error: "No standards indexed. Run 'index_standards' first after importing PDF/HWP.",
-        };
-      }
-
-      // Read code
-      const code = fs.readFileSync(filepath, "utf-8");
-      const config = loadConfig();
-      const client = new OllamaClient(config.ollama);
-
-      // Create query from code (first 1000 chars + filename)
-      const queryText = `Code analysis for ${path.basename(filepath)}:\n${code.slice(0, 1000)}`;
-      const queryEmbedding = await client.embed(queryText, DEFAULT_EMBED_MODEL);
-
-      // Find relevant standards
-      const results = embeddings.map((e) => ({
-        ...e,
-        similarity: cosineSimilarity(queryEmbedding, e.embedding),
-      }));
-      results.sort((a, b) => b.similarity - a.similarity);
-      const relevantStandards = results.slice(0, topK);
-
-      // Build standards context
-      let standardsContext = "## Relevant Development Standards\n\n";
-      for (const r of relevantStandards) {
-        if (r.chunk.ruleId) {
-          standardsContext += `### ${r.chunk.ruleId}: ${r.chunk.section}\n`;
-        } else {
-          standardsContext += `### ${r.chunk.section}\n`;
-        }
-        standardsContext += `${r.chunk.content}\n\n`;
-      }
-
-      // Build analysis prompt
-      const ext = path.extname(filepath);
-      const lang = { ".ts": "typescript", ".js": "javascript", ".java": "java", ".py": "python" }[ext] || "text";
-
-      const prompt = `You are a code quality expert. Analyze the code based on the provided development standards.
-
-${standardsContext}
-
-## Code to Analyze
-File: ${filepath}
-\`\`\`${lang}
-${code.slice(0, 6000)}
-\`\`\`
-
-## Analysis Request
-1. Check for violations of the above standards
-2. Provide specific line numbers if possible
-3. Suggest improvements
-4. Rate overall compliance (1-10)
-
-Respond in Korean.`;
-
-      const response = await client.chat([{ role: "user", content: prompt }]);
-
-      return {
-        success: true,
-        content: `## Code Quality Analysis (RAG)\n\n` +
-          `📄 File: ${filepath}\n` +
-          `🔖 Standards used: ${relevantStandards.length}\n\n` +
-          `---\n\n${response.content}`,
-      };
-    } catch (error) {
-      return { success: false, content: "", error: String(error) };
-    }
-  },
-};
-
-// All standards tools
-export const standardsTools: Tool[] = [
-  importPdfTool,
-  importHwpTool,
-  listStandardsTool,
-  checkQualityTool,
-  indexStandardsTool,
-  searchStandardsTool,
-  checkQualityWithRagTool,
-];