@chatbot-packages/rag 0.1.0

package/dist/extractors/index.d.ts

@@ -0,0 +1,95 @@
+import { m as ExtractorOptions, n as ExtractorType, j as ExtractedDocument } from '../types-CjnplPJD.js';
+
+/**
+ * Base Extractor
+ */
+
+declare abstract class BaseExtractor {
+    protected options: ExtractorOptions;
+    constructor(options: ExtractorOptions);
+    /** Get the extractor type */
+    abstract getType(): ExtractorType;
+    /** Check if this extractor can handle the given path */
+    abstract canHandle(path: string): boolean;
+    /** Extract documents from the source */
+    abstract extract(): Promise<ExtractedDocument[]>;
+    /** Get file extension */
+    protected getExtension(path: string): string;
+    /** Normalize line endings */
+    protected normalizeLineEndings(text: string): string;
+    /** Clean excessive whitespace */
+    protected cleanWhitespace(text: string): string;
+}
+
+/**
+ * CHM (Compiled HTML Help) Extractor
+ *
+ * Extracts HTML content from .chm files using 7z.
+ * CHM files are Microsoft's compiled HTML help format.
+ */
+
+declare class CHMExtractor extends BaseExtractor {
+    private tempDir;
+    constructor(options: {
+        sourcePath: string;
+        outputDir?: string;
+    });
+    getType(): ExtractorType;
+    canHandle(path: string): boolean;
+    extract(): Promise<ExtractedDocument[]>;
+    private extractWithSevenZip;
+    private findHTMLFiles;
+}
+
+/**
+ * HTML Extractor
+ *
+ * Extracts text content from HTML files, preserving structure.
+ */
+
+interface HTMLExtractorOptions {
+    sourcePath: string;
+    /** Whether to preserve headings structure */
+    preserveHeadings?: boolean;
+    /** Whether to include links */
+    includeLinks?: boolean;
+    /** Tags to remove */
+    removeTags?: string[];
+}
+declare class HTMLExtractor extends BaseExtractor {
+    private htmlOptions;
+    constructor(options: HTMLExtractorOptions);
+    getType(): ExtractorType;
+    canHandle(path: string): boolean;
+    extract(): Promise<ExtractedDocument[]>;
+    extractFromString(html: string, filePath: string): Promise<ExtractedDocument>;
+    private extractFromDirectory;
+}
+
+/**
+ * Markdown Extractor
+ *
+ * Extracts content from Markdown files.
+ */
+
+declare class MarkdownExtractor extends BaseExtractor {
+    getType(): ExtractorType;
+    canHandle(path: string): boolean;
+    extract(): Promise<ExtractedDocument[]>;
+    private extractFromFile;
+    private extractTitle;
+    private extractFromDirectory;
+}
+
+/**
+ * Document Extractors
+ *
+ * Extract content from various file formats (CHM, HTML, Markdown)
+ */
+
+/**
+ * Auto-detect and extract documents from a path
+ */
+declare function extractDocuments(sourcePath: string): Promise<ExtractedDocument[]>;
+
+export { BaseExtractor, CHMExtractor, HTMLExtractor, type HTMLExtractorOptions, MarkdownExtractor, extractDocuments };

package/dist/extractors/index.js

@@ -0,0 +1,343 @@
+// src/extractors/base.ts
+var BaseExtractor = class {
+  options;
+  constructor(options) {
+    this.options = options;
+  }
+  /** Get file extension */
+  getExtension(path) {
+    const parts = path.split(".");
+    return parts.length > 1 ? parts[parts.length - 1].toLowerCase() : "";
+  }
+  /** Normalize line endings */
+  normalizeLineEndings(text) {
+    return text.replace(/\r\n/g, "\n").replace(/\r/g, "\n");
+  }
+  /** Clean excessive whitespace */
+  cleanWhitespace(text) {
+    return text.replace(/\n{3,}/g, "\n\n").replace(/[ \t]+/g, " ").trim();
+  }
+};
+
+// src/extractors/chm.ts
+import { exec } from "child_process";
+import { promisify } from "util";
+import { readdir as readdir2, readFile as readFile2, mkdir, rm } from "fs/promises";
+import { existsSync } from "fs";
+import { join as join2, basename as basename2, dirname } from "path";
+
+// src/extractors/html.ts
+import { readFile, readdir, stat } from "fs/promises";
+import { join, basename } from "path";
+import * as cheerio from "cheerio";
+import { convert } from "html-to-text";
+var HTMLExtractor = class extends BaseExtractor {
+  htmlOptions;
+  constructor(options) {
+    super(options);
+    this.htmlOptions = {
+      preserveHeadings: true,
+      includeLinks: false,
+      removeTags: ["script", "style", "nav", "footer", "header", "aside", "meta", "link"],
+      ...options
+    };
+  }
+  getType() {
+    return "html";
+  }
+  canHandle(path) {
+    const ext = this.getExtension(path);
+    return ext === "html" || ext === "htm";
+  }
+  async extract() {
+    const { sourcePath } = this.options;
+    const stats = await stat(sourcePath);
+    if (stats.isFile()) {
+      const content = await readFile(sourcePath, "utf-8");
+      return [await this.extractFromString(content, sourcePath)];
+    }
+    if (stats.isDirectory()) {
+      return this.extractFromDirectory(sourcePath);
+    }
+    throw new Error(`Invalid path: ${sourcePath}`);
+  }
+  async extractFromString(html, filePath) {
+    const $ = cheerio.load(html);
+    for (const tag of this.htmlOptions.removeTags || []) {
+      $(tag).remove();
+    }
+    let title = $("title").text().trim();
+    if (!title) {
+      title = $("h1").first().text().trim();
+    }
+    if (!title) {
+      title = basename(filePath, ".html").replace(/-|_/g, " ");
+    }
+    const text = convert($.html(), {
+      wordwrap: false,
+      preserveNewlines: true,
+      selectors: [
+        { selector: "h1", options: { uppercase: false, prefix: "\n# " } },
+        { selector: "h2", options: { uppercase: false, prefix: "\n## " } },
+        { selector: "h3", options: { uppercase: false, prefix: "\n### " } },
+        { selector: "h4", options: { uppercase: false, prefix: "\n#### " } },
+        { selector: "h5", options: { uppercase: false, prefix: "\n##### " } },
+        { selector: "h6", options: { uppercase: false, prefix: "\n###### " } },
+        { selector: "ul", options: { itemPrefix: "  - " } },
+        { selector: "ol", options: { itemPrefix: "  1. " } },
+        { selector: "table", format: "dataTable" },
+        { selector: "a", options: { ignoreHref: !this.htmlOptions.includeLinks } },
+        { selector: "img", format: "skip" }
+      ]
+    });
+    const content = this.cleanWhitespace(this.normalizeLineEndings(text));
+    return {
+      path: filePath,
+      title,
+      content,
+      format: "html",
+      metadata: {
+        originalLength: html.length,
+        extractedLength: content.length
+      }
+    };
+  }
+  async extractFromDirectory(dir) {
+    const documents = [];
+    const scanDir = async (currentDir) => {
+      const entries = await readdir(currentDir, { withFileTypes: true });
+      for (const entry of entries) {
+        const fullPath = join(currentDir, entry.name);
+        if (entry.isDirectory()) {
+          await scanDir(fullPath);
+        } else if (entry.isFile() && this.canHandle(entry.name)) {
+          try {
+            const content = await readFile(fullPath, "utf-8");
+            const doc = await this.extractFromString(content, fullPath);
+            if (doc.content.trim()) {
+              documents.push(doc);
+            }
+          } catch (error) {
+            console.warn(`Failed to extract ${fullPath}:`, error);
+          }
+        }
+      }
+    };
+    await scanDir(dir);
+    return documents;
+  }
+};
+
+// src/extractors/chm.ts
+var execAsync = promisify(exec);
+var CHMExtractor = class extends BaseExtractor {
+  tempDir;
+  constructor(options) {
+    super(options);
+    this.tempDir = options.outputDir || join2(dirname(options.sourcePath), ".chm-extract-temp");
+  }
+  getType() {
+    return "chm";
+  }
+  canHandle(path) {
+    return this.getExtension(path) === "chm";
+  }
+  async extract() {
+    const { sourcePath } = this.options;
+    if (!existsSync(sourcePath)) {
+      throw new Error(`CHM file not found: ${sourcePath}`);
+    }
+    await mkdir(this.tempDir, { recursive: true });
+    try {
+      await this.extractWithSevenZip(sourcePath, this.tempDir);
+      const htmlFiles = await this.findHTMLFiles(this.tempDir);
+      const documents = [];
+      for (const htmlFile of htmlFiles) {
+        try {
+          const content = await readFile2(htmlFile, "utf-8");
+          const htmlExtractor = new HTMLExtractor({ sourcePath: htmlFile });
+          const extracted = await htmlExtractor.extractFromString(content, htmlFile);
+          if (extracted.content.trim()) {
+            documents.push({
+              ...extracted,
+              format: "chm",
+              metadata: {
+                ...extracted.metadata,
+                sourceChm: basename2(sourcePath),
+                originalPath: htmlFile.replace(this.tempDir, "")
+              }
+            });
+          }
+        } catch (error) {
+          console.warn(`Failed to parse HTML file: ${htmlFile}`, error);
+        }
+      }
+      return documents;
+    } finally {
+      if (existsSync(this.tempDir)) {
+        await rm(this.tempDir, { recursive: true, force: true });
+      }
+    }
+  }
+  async extractWithSevenZip(chmPath, outputDir) {
+    try {
+      await execAsync(`7z x "${chmPath}" -o"${outputDir}" -y`, {
+        maxBuffer: 50 * 1024 * 1024
+        // 50MB buffer
+      });
+    } catch (error) {
+      try {
+        await execAsync(`7za x "${chmPath}" -o"${outputDir}" -y`, {
+          maxBuffer: 50 * 1024 * 1024
+        });
+      } catch {
+        throw new Error(
+          `Failed to extract CHM file. Please ensure 7z is installed.
+  - On Ubuntu/Debian: sudo apt-get install p7zip-full
+  - On macOS: brew install p7zip
+  - On Windows: Install 7-Zip from https://7-zip.org/
+Original error: ${error}`
+        );
+      }
+    }
+  }
+  async findHTMLFiles(dir) {
+    const htmlFiles = [];
+    const scanDir = async (currentDir) => {
+      const entries = await readdir2(currentDir, { withFileTypes: true });
+      for (const entry of entries) {
+        const fullPath = join2(currentDir, entry.name);
+        if (entry.isDirectory()) {
+          if (!entry.name.startsWith("$") && !entry.name.startsWith("#")) {
+            await scanDir(fullPath);
+          }
+        } else if (entry.isFile()) {
+          const ext = this.getExtension(entry.name);
+          if (ext === "html" || ext === "htm") {
+            htmlFiles.push(fullPath);
+          }
+        }
+      }
+    };
+    await scanDir(dir);
+    return htmlFiles.sort();
+  }
+};
+
+// src/extractors/markdown.ts
+import { readFile as readFile3, readdir as readdir3, stat as stat2 } from "fs/promises";
+import { join as join3, basename as basename3 } from "path";
+var MarkdownExtractor = class extends BaseExtractor {
+  getType() {
+    return "markdown";
+  }
+  canHandle(path) {
+    const ext = this.getExtension(path);
+    return ext === "md" || ext === "markdown";
+  }
+  async extract() {
+    const { sourcePath } = this.options;
+    const stats = await stat2(sourcePath);
+    if (stats.isFile()) {
+      return [await this.extractFromFile(sourcePath)];
+    }
+    if (stats.isDirectory()) {
+      return this.extractFromDirectory(sourcePath);
+    }
+    throw new Error(`Invalid path: ${sourcePath}`);
+  }
+  async extractFromFile(filePath) {
+    const content = await readFile3(filePath, "utf-8");
+    const normalizedContent = this.normalizeLineEndings(content);
+    let title = this.extractTitle(normalizedContent);
+    if (!title) {
+      title = basename3(filePath, ".md").replace(/-|_/g, " ");
+    }
+    return {
+      path: filePath,
+      title,
+      content: this.cleanWhitespace(normalizedContent),
+      format: "markdown",
+      metadata: {
+        originalLength: content.length
+      }
+    };
+  }
+  extractTitle(content) {
+    const h1Match = content.match(/^#\s+(.+)$/m);
+    if (h1Match) {
+      return h1Match[1].trim();
+    }
+    const underlineMatch = content.match(/^(.+)\n=+\s*$/m);
+    if (underlineMatch) {
+      return underlineMatch[1].trim();
+    }
+    return void 0;
+  }
+  async extractFromDirectory(dir) {
+    const documents = [];
+    const scanDir = async (currentDir) => {
+      const entries = await readdir3(currentDir, { withFileTypes: true });
+      for (const entry of entries) {
+        const fullPath = join3(currentDir, entry.name);
+        if (entry.isDirectory()) {
+          if (!entry.name.startsWith(".") && entry.name !== "node_modules") {
+            await scanDir(fullPath);
+          }
+        } else if (entry.isFile() && this.canHandle(entry.name)) {
+          try {
+            const doc = await this.extractFromFile(fullPath);
+            if (doc.content.trim()) {
+              documents.push(doc);
+            }
+          } catch (error) {
+            console.warn(`Failed to extract ${fullPath}:`, error);
+          }
+        }
+      }
+    };
+    await scanDir(dir);
+    return documents;
+  }
+};
+
+// src/extractors/index.ts
+import { stat as stat3 } from "fs/promises";
+async function extractDocuments(sourcePath) {
+  const stats = await stat3(sourcePath);
+  const ext = sourcePath.split(".").pop()?.toLowerCase() || "";
+  if (stats.isFile()) {
+    if (ext === "chm") {
+      const extractor = new CHMExtractor({ sourcePath });
+      return extractor.extract();
+    }
+    if (ext === "html" || ext === "htm") {
+      const extractor = new HTMLExtractor({ sourcePath });
+      return extractor.extract();
+    }
+    if (ext === "md" || ext === "markdown") {
+      const extractor = new MarkdownExtractor({ sourcePath });
+      return extractor.extract();
+    }
+    throw new Error(`Unsupported file format: ${ext}`);
+  }
+  if (stats.isDirectory()) {
+    const allDocuments = [];
+    const htmlExtractor = new HTMLExtractor({ sourcePath });
+    const htmlDocs = await htmlExtractor.extract();
+    allDocuments.push(...htmlDocs);
+    const mdExtractor = new MarkdownExtractor({ sourcePath });
+    const mdDocs = await mdExtractor.extract();
+    allDocuments.push(...mdDocs);
+    return allDocuments;
+  }
+  throw new Error(`Invalid path: ${sourcePath}`);
+}
+export {
+  BaseExtractor,
+  CHMExtractor,
+  HTMLExtractor,
+  MarkdownExtractor,
+  extractDocuments
+};
+//# sourceMappingURL=index.js.map

package/dist/extractors/index.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/extractors/base.ts","../../src/extractors/chm.ts","../../src/extractors/html.ts","../../src/extractors/markdown.ts","../../src/extractors/index.ts"],"sourcesContent":["/**\n * Base Extractor\n */\n\nimport type { ExtractedDocument, ExtractorOptions, ExtractorType } from '../types.js';\n\nexport abstract class BaseExtractor {\n  protected options: ExtractorOptions;\n\n  constructor(options: ExtractorOptions) {\n    this.options = options;\n  }\n\n  /** Get the extractor type */\n  abstract getType(): ExtractorType;\n\n  /** Check if this extractor can handle the given path */\n  abstract canHandle(path: string): boolean;\n\n  /** Extract documents from the source */\n  abstract extract(): Promise<ExtractedDocument[]>;\n\n  /** Get file extension */\n  protected getExtension(path: string): string {\n    const parts = path.split('.');\n    return parts.length > 1 ? parts[parts.length - 1].toLowerCase() : '';\n  }\n\n  /** Normalize line endings */\n  protected normalizeLineEndings(text: string): string {\n    return text.replace(/\\r\\n/g, '\\n').replace(/\\r/g, '\\n');\n  }\n\n  /** Clean excessive whitespace */\n  protected cleanWhitespace(text: string): string {\n    return text\n      .replace(/\\n{3,}/g, '\\n\\n')\n      .replace(/[ \\t]+/g, ' ')\n      .trim();\n  }\n}\n","/**\n * CHM (Compiled HTML Help) Extractor\n *\n * Extracts HTML content from .chm files using 7z.\n * CHM files are Microsoft's compiled HTML help format.\n */\n\nimport { exec } from 'child_process';\nimport { promisify } from 'util';\nimport { readdir, readFile, mkdir, rm } from 'fs/promises';\nimport { existsSync } from 'fs';\nimport { join, basename, dirname } from 'path';\nimport { BaseExtractor } from './base.js';\nimport { HTMLExtractor } from './html.js';\nimport type { ExtractedDocument, ExtractorType } from '../types.js';\n\nconst execAsync = promisify(exec);\n\nexport class CHMExtractor extends BaseExtractor {\n  private tempDir: string;\n\n  constructor(options: { sourcePath: string; outputDir?: string }) {\n    super(options);\n    this.tempDir = options.outputDir || join(dirname(options.sourcePath), '.chm-extract-temp');\n  }\n\n  getType(): ExtractorType {\n    return 'chm';\n  }\n\n  canHandle(path: string): boolean {\n    return this.getExtension(path) === 'chm';\n  }\n\n  async extract(): Promise<ExtractedDocument[]> {\n    const { sourcePath } = this.options;\n\n    if (!existsSync(sourcePath)) {\n      throw new Error(`CHM file not found: ${sourcePath}`);\n    }\n\n    // Create temp directory\n    await mkdir(this.tempDir, { recursive: true });\n\n    try {\n      // Extract CHM using 7z\n      await this.extractWithSevenZip(sourcePath, this.tempDir);\n\n      // Find and parse all HTML files\n      const htmlFiles = await this.findHTMLFiles(this.tempDir);\n      const documents: ExtractedDocument[] = [];\n\n      for (const htmlFile of htmlFiles) {\n        try {\n          const content = await readFile(htmlFile, 'utf-8');\n          const htmlExtractor = new HTMLExtractor({ sourcePath: htmlFile });\n          const extracted = await htmlExtractor.extractFromString(content, htmlFile);\n\n          if (extracted.content.trim()) {\n            documents.push({\n              ...extracted,\n              format: 'chm',\n              metadata: {\n                ...extracted.metadata,\n                sourceChm: basename(sourcePath),\n                originalPath: htmlFile.replace(this.tempDir, ''),\n              },\n            });\n          }\n        } catch (error) {\n          // Skip files that can't be parsed\n          console.warn(`Failed to parse HTML file: ${htmlFile}`, error);\n        }\n      }\n\n      return documents;\n    } finally {\n      // Cleanup temp directory\n      if (existsSync(this.tempDir)) {\n        await rm(this.tempDir, { recursive: true, force: true });\n      }\n    }\n  }\n\n  private async extractWithSevenZip(chmPath: string, outputDir: string): Promise<void> {\n    try {\n      // Try 7z first\n      await execAsync(`7z x \"${chmPath}\" -o\"${outputDir}\" -y`, {\n        maxBuffer: 50 * 1024 * 1024, // 50MB buffer\n      });\n    } catch (error) {\n      // Try 7za as fallback (common on some systems)\n      try {\n        await execAsync(`7za x \"${chmPath}\" -o\"${outputDir}\" -y`, {\n          maxBuffer: 50 * 1024 * 1024,\n        });\n      } catch {\n        throw new Error(\n          `Failed to extract CHM file. Please ensure 7z is installed.\\n` +\n            `  - On Ubuntu/Debian: sudo apt-get install p7zip-full\\n` +\n            `  - On macOS: brew install p7zip\\n` +\n            `  - On Windows: Install 7-Zip from https://7-zip.org/\\n` +\n            `Original error: ${error}`\n        );\n      }\n    }\n  }\n\n  private async findHTMLFiles(dir: string): Promise<string[]> {\n    const htmlFiles: string[] = [];\n\n    const scanDir = async (currentDir: string) => {\n      const entries = await readdir(currentDir, { withFileTypes: true });\n\n      for (const entry of entries) {\n        const fullPath = join(currentDir, entry.name);\n\n        if (entry.isDirectory()) {\n          // Skip system directories\n          if (!entry.name.startsWith('$') && !entry.name.startsWith('#')) {\n            await scanDir(fullPath);\n          }\n        } else if (entry.isFile()) {\n          const ext = this.getExtension(entry.name);\n          if (ext === 'html' || ext === 'htm') {\n            htmlFiles.push(fullPath);\n          }\n        }\n      }\n    };\n\n    await scanDir(dir);\n    return htmlFiles.sort();\n  }\n}\n","/**\n * HTML Extractor\n *\n * Extracts text content from HTML files, preserving structure.\n */\n\nimport { readFile, readdir, stat } from 'fs/promises';\nimport { join, basename } from 'path';\nimport * as cheerio from 'cheerio';\nimport { convert } from 'html-to-text';\nimport { BaseExtractor } from './base.js';\nimport type { ExtractedDocument, ExtractorType } from '../types.js';\n\nexport interface HTMLExtractorOptions {\n  sourcePath: string;\n  /** Whether to preserve headings structure */\n  preserveHeadings?: boolean;\n  /** Whether to include links */\n  includeLinks?: boolean;\n  /** Tags to remove */\n  removeTags?: string[];\n}\n\nexport class HTMLExtractor extends BaseExtractor {\n  private htmlOptions: HTMLExtractorOptions;\n\n  constructor(options: HTMLExtractorOptions) {\n    super(options);\n    this.htmlOptions = {\n      preserveHeadings: true,\n      includeLinks: false,\n      removeTags: ['script', 'style', 'nav', 'footer', 'header', 'aside', 'meta', 'link'],\n      ...options,\n    };\n  }\n\n  getType(): ExtractorType {\n    return 'html';\n  }\n\n  canHandle(path: string): boolean {\n    const ext = this.getExtension(path);\n    return ext === 'html' || ext === 'htm';\n  }\n\n  async extract(): Promise<ExtractedDocument[]> {\n    const { sourcePath } = this.options;\n    const stats = await stat(sourcePath);\n\n    if (stats.isFile()) {\n      const content = await readFile(sourcePath, 'utf-8');\n      return [await this.extractFromString(content, sourcePath)];\n    }\n\n    if (stats.isDirectory()) {\n      return this.extractFromDirectory(sourcePath);\n    }\n\n    throw new Error(`Invalid path: ${sourcePath}`);\n  }\n\n  async extractFromString(html: string, filePath: string): Promise<ExtractedDocument> {\n    const $ = cheerio.load(html);\n\n    // Remove unwanted tags\n    for (const tag of this.htmlOptions.removeTags || []) {\n      $(tag).remove();\n    }\n\n    // Extract title\n    let title = $('title').text().trim();\n    if (!title) {\n      title = $('h1').first().text().trim();\n    }\n    if (!title) {\n      title = basename(filePath, '.html').replace(/-|_/g, ' ');\n    }\n\n    // Convert to text with structure preservation\n    const text = convert($.html(), {\n      wordwrap: false,\n      preserveNewlines: true,\n      selectors: [\n        { selector: 'h1', options: { uppercase: false, prefix: '\\n# ' } },\n        { selector: 'h2', options: { uppercase: false, prefix: '\\n## ' } },\n        { selector: 'h3', options: { uppercase: false, prefix: '\\n### ' } },\n        { selector: 'h4', options: { uppercase: false, prefix: '\\n#### ' } },\n        { selector: 'h5', options: { uppercase: false, prefix: '\\n##### ' } },\n        { selector: 'h6', options: { uppercase: false, prefix: '\\n###### ' } },\n        { selector: 'ul', options: { itemPrefix: '  - ' } },\n        { selector: 'ol', options: { itemPrefix: '  1. ' } },\n        { selector: 'table', format: 'dataTable' },\n        { selector: 'a', options: { ignoreHref: !this.htmlOptions.includeLinks } },\n        { selector: 'img', format: 'skip' },\n      ],\n    });\n\n    const content = this.cleanWhitespace(this.normalizeLineEndings(text));\n\n    return {\n      path: filePath,\n      title,\n      content,\n      format: 'html',\n      metadata: {\n        originalLength: html.length,\n        extractedLength: content.length,\n      },\n    };\n  }\n\n  private async extractFromDirectory(dir: string): Promise<ExtractedDocument[]> {\n    const documents: ExtractedDocument[] = [];\n\n    const scanDir = async (currentDir: string) => {\n      const entries = await readdir(currentDir, { withFileTypes: true });\n\n      for (const entry of entries) {\n        const fullPath = join(currentDir, entry.name);\n\n        if (entry.isDirectory()) {\n          await scanDir(fullPath);\n        } else if (entry.isFile() && this.canHandle(entry.name)) {\n          try {\n            const content = await readFile(fullPath, 'utf-8');\n            const doc = await this.extractFromString(content, fullPath);\n            if (doc.content.trim()) {\n              documents.push(doc);\n            }\n          } catch (error) {\n            console.warn(`Failed to extract ${fullPath}:`, error);\n          }\n        }\n      }\n    };\n\n    await scanDir(dir);\n    return documents;\n  }\n}\n","/**\n * Markdown Extractor\n *\n * Extracts content from Markdown files.\n */\n\nimport { readFile, readdir, stat } from 'fs/promises';\nimport { join, basename } from 'path';\nimport { BaseExtractor } from './base.js';\nimport type { ExtractedDocument, ExtractorType } from '../types.js';\n\nexport class MarkdownExtractor extends BaseExtractor {\n  getType(): ExtractorType {\n    return 'markdown';\n  }\n\n  canHandle(path: string): boolean {\n    const ext = this.getExtension(path);\n    return ext === 'md' || ext === 'markdown';\n  }\n\n  async extract(): Promise<ExtractedDocument[]> {\n    const { sourcePath } = this.options;\n    const stats = await stat(sourcePath);\n\n    if (stats.isFile()) {\n      return [await this.extractFromFile(sourcePath)];\n    }\n\n    if (stats.isDirectory()) {\n      return this.extractFromDirectory(sourcePath);\n    }\n\n    throw new Error(`Invalid path: ${sourcePath}`);\n  }\n\n  private async extractFromFile(filePath: string): Promise<ExtractedDocument> {\n    const content = await readFile(filePath, 'utf-8');\n    const normalizedContent = this.normalizeLineEndings(content);\n\n    // Extract title from first H1 or filename\n    let title = this.extractTitle(normalizedContent);\n    if (!title) {\n      title = basename(filePath, '.md').replace(/-|_/g, ' ');\n    }\n\n    return {\n      path: filePath,\n      title,\n      content: this.cleanWhitespace(normalizedContent),\n      format: 'markdown',\n      metadata: {\n        originalLength: content.length,\n      },\n    };\n  }\n\n  private extractTitle(content: string): string | undefined {\n    // Look for first H1 heading\n    const h1Match = content.match(/^#\\s+(.+)$/m);\n    if (h1Match) {\n      return h1Match[1].trim();\n    }\n\n    // Look for underline-style H1\n    const underlineMatch = content.match(/^(.+)\\n=+\\s*$/m);\n    if (underlineMatch) {\n      return underlineMatch[1].trim();\n    }\n\n    return undefined;\n  }\n\n  private async extractFromDirectory(dir: string): Promise<ExtractedDocument[]> {\n    const documents: ExtractedDocument[] = [];\n\n    const scanDir = async (currentDir: string) => {\n      const entries = await readdir(currentDir, { withFileTypes: true });\n\n      for (const entry of entries) {\n        const fullPath = join(currentDir, entry.name);\n\n        if (entry.isDirectory()) {\n          // Skip hidden and common non-content directories\n          if (!entry.name.startsWith('.') && entry.name !== 'node_modules') {\n            await scanDir(fullPath);\n          }\n        } else if (entry.isFile() && this.canHandle(entry.name)) {\n          try {\n            const doc = await this.extractFromFile(fullPath);\n            if (doc.content.trim()) {\n              documents.push(doc);\n            }\n          } catch (error) {\n            console.warn(`Failed to extract ${fullPath}:`, error);\n          }\n        }\n      }\n    };\n\n    await scanDir(dir);\n    return documents;\n  }\n}\n","/**\n * Document Extractors\n *\n * Extract content from various file formats (CHM, HTML, Markdown)\n */\n\nexport { BaseExtractor } from './base.js';\nexport { CHMExtractor } from './chm.js';\nexport { HTMLExtractor, type HTMLExtractorOptions } from './html.js';\nexport { MarkdownExtractor } from './markdown.js';\n\nimport { stat } from 'fs/promises';\nimport { CHMExtractor } from './chm.js';\nimport { HTMLExtractor } from './html.js';\nimport { MarkdownExtractor } from './markdown.js';\nimport type { ExtractedDocument } from '../types.js';\n\n/**\n * Auto-detect and extract documents from a path\n */\nexport async function extractDocuments(sourcePath: string): Promise<ExtractedDocument[]> {\n  const stats = await stat(sourcePath);\n  const ext = sourcePath.split('.').pop()?.toLowerCase() || '';\n\n  // Single file\n  if (stats.isFile()) {\n    if (ext === 'chm') {\n      const extractor = new CHMExtractor({ sourcePath });\n      return extractor.extract();\n    }\n    if (ext === 'html' || ext === 'htm') {\n      const extractor = new HTMLExtractor({ sourcePath });\n      return extractor.extract();\n    }\n    if (ext === 'md' || ext === 'markdown') {\n      const extractor = new MarkdownExtractor({ sourcePath });\n      return extractor.extract();\n    }\n    throw new Error(`Unsupported file format: ${ext}`);\n  }\n\n  // Directory - extract all supported files\n  if (stats.isDirectory()) {\n    const allDocuments: ExtractedDocument[] = [];\n\n    // Extract HTML files\n    const htmlExtractor = new HTMLExtractor({ sourcePath });\n    const htmlDocs = await htmlExtractor.extract();\n    allDocuments.push(...htmlDocs);\n\n    // Extract Markdown files\n    const mdExtractor = new MarkdownExtractor({ sourcePath });\n    const mdDocs = await mdExtractor.extract();\n    allDocuments.push(...mdDocs);\n\n    return allDocuments;\n  }\n\n  throw new Error(`Invalid path: ${sourcePath}`);\n}\n"],"mappings":";AAMO,IAAe,gBAAf,MAA6B;AAAA,EACxB;AAAA,EAEV,YAAY,SAA2B;AACrC,SAAK,UAAU;AAAA,EACjB;AAAA;AAAA,EAYU,aAAa,MAAsB;AAC3C,UAAM,QAAQ,KAAK,MAAM,GAAG;AAC5B,WAAO,MAAM,SAAS,IAAI,MAAM,MAAM,SAAS,CAAC,EAAE,YAAY,IAAI;AAAA,EACpE;AAAA;AAAA,EAGU,qBAAqB,MAAsB;AACnD,WAAO,KAAK,QAAQ,SAAS,IAAI,EAAE,QAAQ,OAAO,IAAI;AAAA,EACxD;AAAA;AAAA,EAGU,gBAAgB,MAAsB;AAC9C,WAAO,KACJ,QAAQ,WAAW,MAAM,EACzB,QAAQ,WAAW,GAAG,EACtB,KAAK;AAAA,EACV;AACF;;;ACjCA,SAAS,YAAY;AACrB,SAAS,iBAAiB;AAC1B,SAAS,WAAAA,UAAS,YAAAC,WAAU,OAAO,UAAU;AAC7C,SAAS,kBAAkB;AAC3B,SAAS,QAAAC,OAAM,YAAAC,WAAU,eAAe;;;ACLxC,SAAS,UAAU,SAAS,YAAY;AACxC,SAAS,MAAM,gBAAgB;AAC/B,YAAY,aAAa;AACzB,SAAS,eAAe;AAcjB,IAAM,gBAAN,cAA4B,cAAc;AAAA,EACvC;AAAA,EAER,YAAY,SAA+B;AACzC,UAAM,OAAO;AACb,SAAK,cAAc;AAAA,MACjB,kBAAkB;AAAA,MAClB,cAAc;AAAA,MACd,YAAY,CAAC,UAAU,SAAS,OAAO,UAAU,UAAU,SAAS,QAAQ,MAAM;AAAA,MAClF,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,UAAyB;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,UAAU,MAAuB;AAC/B,UAAM,MAAM,KAAK,aAAa,IAAI;AAClC,WAAO,QAAQ,UAAU,QAAQ;AAAA,EACnC;AAAA,EAEA,MAAM,UAAwC;AAC5C,UAAM,EAAE,WAAW,IAAI,KAAK;AAC5B,UAAM,QAAQ,MAAM,KAAK,UAAU;AAEnC,QAAI,MAAM,OAAO,GAAG;AAClB,YAAM,UAAU,MAAM,SAAS,YAAY,OAAO;AAClD,aAAO,CAAC,MAAM,KAAK,kBAAkB,SAAS,UAAU,CAAC;AAAA,IAC3D;AAEA,QAAI,MAAM,YAAY,GAAG;AACvB,aAAO,KAAK,qBAAqB,UAAU;AAAA,IAC7C;AAEA,UAAM,IAAI,MAAM,iBAAiB,UAAU,EAAE;AAAA,EAC/C;AAAA,EAEA,MAAM,kBAAkB,MAAc,UAA8C;AAClF,UAAM,IAAY,aAAK,IAAI;AAG3B,eAAW,OAAO,KAAK,YAAY,cAAc,CAAC,GAAG;AACnD,QAAE,GAAG,EAAE,OAAO;AAAA,IAChB;AAGA,QAAI,QAAQ,EAAE,OAAO,EAAE,KAAK,EAAE,KAAK;AACnC,QAAI,CAAC,OAAO;AACV,cAAQ,EAAE,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,KAAK;AAAA,IACtC;AACA,QAAI,CAAC,OAAO;AACV,cAAQ,SAAS,UAAU,OAAO,EAAE,QAAQ,QAAQ,GAAG;AAAA,IACzD;AAGA,UAAM,OAAO,QAAQ,EAAE,KAAK,GAAG;AAAA,MAC7B,UAAU;AAAA,MACV,kBAAkB;AAAA,MAClB,WAAW;AAAA,QACT,EAAE,UAAU,MAAM,SAAS,EAAE,WAAW,OAAO,QAAQ,OAAO,EAAE;AAAA,QAChE,EAAE,UAAU,MAAM,SAAS,EAAE,WAAW,OAAO,QAAQ,QAAQ,EAAE;AAAA,QACjE,EAAE,UAAU,MAAM,SAAS,EAAE,WAAW,OAAO,QAAQ,SAAS,EAAE;AAAA,QAClE,EAAE,UAAU,MAAM,SAAS,EAAE,WAAW,OAAO,QAAQ,UAAU,EAAE;AAAA,QACnE,EAAE,UAAU,MAAM,SAAS,EAAE,WAAW,OAAO,QAAQ,WAAW,EAAE;AAAA,QACpE,EAAE,UAAU,MAAM,SAAS,EAAE,WAAW,OAAO,QAAQ,YAAY,EAAE;AAAA,QACrE,EAAE,UAAU,MAAM,SAAS,EAAE,YAAY,OAAO,EAAE;AAAA,QAClD,EAAE,UAAU,MAAM,SAAS,EAAE,YAAY,QAAQ,EAAE;AAAA,QACnD,EAAE,UAAU,SAAS,QAAQ,YAAY;AAAA,QACzC,EAAE,UAAU,KAAK,SAAS,EAAE,YAAY,CAAC,KAAK,YAAY,aAAa,EAAE;AAAA,QACzE,EAAE,UAAU,OAAO,QAAQ,OAAO;AAAA,MACpC;AAAA,IACF,CAAC;AAED,UAAM,UAAU,KAAK,gBAAgB,KAAK,qBAAqB,IAAI,CAAC;AAEpE,WAAO;AAAA,MACL,MAAM;AAAA,MACN;AAAA,MACA;AAAA,MACA,QAAQ;AAAA,MACR,UAAU;AAAA,QACR,gBAAgB,KAAK;AAAA,QACrB,iBAAiB,QAAQ;AAAA,MAC3B;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,qBAAqB,KAA2C;AAC5E,UAAM,YAAiC,CAAC;AAExC,UAAM,UAAU,OAAO,eAAuB;AAC5C,YAAM,UAAU,MAAM,QAAQ,YAAY,EAAE,eAAe,KAAK,CAAC;AAEjE,iBAAW,SAAS,SAAS;AAC3B,cAAM,WAAW,KAAK,YAAY,MAAM,IAAI;AAE5C,YAAI,MAAM,YAAY,GAAG;AACvB,gBAAM,QAAQ,QAAQ;AAAA,QACxB,WAAW,MAAM,OAAO,KAAK,KAAK,UAAU,MAAM,IAAI,GAAG;AACvD,cAAI;AACF,kBAAM,UAAU,MAAM,SAAS,UAAU,OAAO;AAChD,kBAAM,MAAM,MAAM,KAAK,kBAAkB,SAAS,QAAQ;AAC1D,gBAAI,IAAI,QAAQ,KAAK,GAAG;AACtB,wBAAU,KAAK,GAAG;AAAA,YACpB;AAAA,UACF,SAAS,OAAO;AACd,oBAAQ,KAAK,qBAAqB,QAAQ,KAAK,KAAK;AAAA,UACtD;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,GAAG;AACjB,WAAO;AAAA,EACT;AACF;;;AD3HA,IAAM,YAAY,UAAU,IAAI;AAEzB,IAAM,eAAN,cAA2B,cAAc;AAAA,EACtC;AAAA,EAER,YAAY,SAAqD;AAC/D,UAAM,OAAO;AACb,SAAK,UAAU,QAAQ,aAAaC,MAAK,QAAQ,QAAQ,UAAU,GAAG,mBAAmB;AAAA,EAC3F;AAAA,EAEA,UAAyB;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,UAAU,MAAuB;AAC/B,WAAO,KAAK,aAAa,IAAI,MAAM;AAAA,EACrC;AAAA,EAEA,MAAM,UAAwC;AAC5C,UAAM,EAAE,WAAW,IAAI,KAAK;AAE5B,QAAI,CAAC,WAAW,UAAU,GAAG;AAC3B,YAAM,IAAI,MAAM,uBAAuB,UAAU,EAAE;AAAA,IACrD;AAGA,UAAM,MAAM,KAAK,SAAS,EAAE,WAAW,KAAK,CAAC;AAE7C,QAAI;AAEF,YAAM,KAAK,oBAAoB,YAAY,KAAK,OAAO;AAGvD,YAAM,YAAY,MAAM,KAAK,cAAc,KAAK,OAAO;AACvD,YAAM,YAAiC,CAAC;AAExC,iBAAW,YAAY,WAAW;AAChC,YAAI;AACF,gBAAM,UAAU,MAAMC,UAAS,UAAU,OAAO;AAChD,gBAAM,gBAAgB,IAAI,cAAc,EAAE,YAAY,SAAS,CAAC;AAChE,gBAAM,YAAY,MAAM,cAAc,kBAAkB,SAAS,QAAQ;AAEzE,cAAI,UAAU,QAAQ,KAAK,GAAG;AAC5B,sBAAU,KAAK;AAAA,cACb,GAAG;AAAA,cACH,QAAQ;AAAA,cACR,UAAU;AAAA,gBACR,GAAG,UAAU;AAAA,gBACb,WAAWC,UAAS,UAAU;AAAA,gBAC9B,cAAc,SAAS,QAAQ,KAAK,SAAS,EAAE;AAAA,cACjD;AAAA,YACF,CAAC;AAAA,UACH;AAAA,QACF,SAAS,OAAO;AAEd,kBAAQ,KAAK,8BAA8B,QAAQ,IAAI,KAAK;AAAA,QAC9D;AAAA,MACF;AAEA,aAAO;AAAA,IACT,UAAE;AAEA,UAAI,WAAW,KAAK,OAAO,GAAG;AAC5B,cAAM,GAAG,KAAK,SAAS,EAAE,WAAW,MAAM,OAAO,KAAK,CAAC;AAAA,MACzD;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,oBAAoB,SAAiB,WAAkC;AACnF,QAAI;AAEF,YAAM,UAAU,SAAS,OAAO,QAAQ,SAAS,QAAQ;AAAA,QACvD,WAAW,KAAK,OAAO;AAAA;AAAA,MACzB,CAAC;AAAA,IACH,SAAS,OAAO;AAEd,UAAI;AACF,cAAM,UAAU,UAAU,OAAO,QAAQ,SAAS,QAAQ;AAAA,UACxD,WAAW,KAAK,OAAO;AAAA,QACzB,CAAC;AAAA,MACH,QAAQ;AACN,cAAM,IAAI;AAAA,UACR;AAAA;AAAA;AAAA;AAAA,kBAIqB,KAAK;AAAA,QAC5B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,cAAc,KAAgC;AAC1D,UAAM,YAAsB,CAAC;AAE7B,UAAM,UAAU,OAAO,eAAuB;AAC5C,YAAM,UAAU,MAAMC,SAAQ,YAAY,EAAE,eAAe,KAAK,CAAC;AAEjE,iBAAW,SAAS,SAAS;AAC3B,cAAM,WAAWH,MAAK,YAAY,MAAM,IAAI;AAE5C,YAAI,MAAM,YAAY,GAAG;AAEvB,cAAI,CAAC,MAAM,KAAK,WAAW,GAAG,KAAK,CAAC,MAAM,KAAK,WAAW,GAAG,GAAG;AAC9D,kBAAM,QAAQ,QAAQ;AAAA,UACxB;AAAA,QACF,WAAW,MAAM,OAAO,GAAG;AACzB,gBAAM,MAAM,KAAK,aAAa,MAAM,IAAI;AACxC,cAAI,QAAQ,UAAU,QAAQ,OAAO;AACnC,sBAAU,KAAK,QAAQ;AAAA,UACzB;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,GAAG;AACjB,WAAO,UAAU,KAAK;AAAA,EACxB;AACF;;;AEhIA,SAAS,YAAAI,WAAU,WAAAC,UAAS,QAAAC,aAAY;AACxC,SAAS,QAAAC,OAAM,YAAAC,iBAAgB;AAIxB,IAAM,oBAAN,cAAgC,cAAc;AAAA,EACnD,UAAyB;AACvB,WAAO;AAAA,EACT;AAAA,EAEA,UAAU,MAAuB;AAC/B,UAAM,MAAM,KAAK,aAAa,IAAI;AAClC,WAAO,QAAQ,QAAQ,QAAQ;AAAA,EACjC;AAAA,EAEA,MAAM,UAAwC;AAC5C,UAAM,EAAE,WAAW,IAAI,KAAK;AAC5B,UAAM,QAAQ,MAAMC,MAAK,UAAU;AAEnC,QAAI,MAAM,OAAO,GAAG;AAClB,aAAO,CAAC,MAAM,KAAK,gBAAgB,UAAU,CAAC;AAAA,IAChD;AAEA,QAAI,MAAM,YAAY,GAAG;AACvB,aAAO,KAAK,qBAAqB,UAAU;AAAA,IAC7C;AAEA,UAAM,IAAI,MAAM,iBAAiB,UAAU,EAAE;AAAA,EAC/C;AAAA,EAEA,MAAc,gBAAgB,UAA8C;AAC1E,UAAM,UAAU,MAAMC,UAAS,UAAU,OAAO;AAChD,UAAM,oBAAoB,KAAK,qBAAqB,OAAO;AAG3D,QAAI,QAAQ,KAAK,aAAa,iBAAiB;AAC/C,QAAI,CAAC,OAAO;AACV,cAAQC,UAAS,UAAU,KAAK,EAAE,QAAQ,QAAQ,GAAG;AAAA,IACvD;AAEA,WAAO;AAAA,MACL,MAAM;AAAA,MACN;AAAA,MACA,SAAS,KAAK,gBAAgB,iBAAiB;AAAA,MAC/C,QAAQ;AAAA,MACR,UAAU;AAAA,QACR,gBAAgB,QAAQ;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,aAAa,SAAqC;AAExD,UAAM,UAAU,QAAQ,MAAM,aAAa;AAC3C,QAAI,SAAS;AACX,aAAO,QAAQ,CAAC,EAAE,KAAK;AAAA,IACzB;AAGA,UAAM,iBAAiB,QAAQ,MAAM,gBAAgB;AACrD,QAAI,gBAAgB;AAClB,aAAO,eAAe,CAAC,EAAE,KAAK;AAAA,IAChC;AAEA,WAAO;AAAA,EACT;AAAA,EAEA,MAAc,qBAAqB,KAA2C;AAC5E,UAAM,YAAiC,CAAC;AAExC,UAAM,UAAU,OAAO,eAAuB;AAC5C,YAAM,UAAU,MAAMC,SAAQ,YAAY,EAAE,eAAe,KAAK,CAAC;AAEjE,iBAAW,SAAS,SAAS;AAC3B,cAAM,WAAWC,MAAK,YAAY,MAAM,IAAI;AAE5C,YAAI,MAAM,YAAY,GAAG;AAEvB,cAAI,CAAC,MAAM,KAAK,WAAW,GAAG,KAAK,MAAM,SAAS,gBAAgB;AAChE,kBAAM,QAAQ,QAAQ;AAAA,UACxB;AAAA,QACF,WAAW,MAAM,OAAO,KAAK,KAAK,UAAU,MAAM,IAAI,GAAG;AACvD,cAAI;AACF,kBAAM,MAAM,MAAM,KAAK,gBAAgB,QAAQ;AAC/C,gBAAI,IAAI,QAAQ,KAAK,GAAG;AACtB,wBAAU,KAAK,GAAG;AAAA,YACpB;AAAA,UACF,SAAS,OAAO;AACd,oBAAQ,KAAK,qBAAqB,QAAQ,KAAK,KAAK;AAAA,UACtD;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,GAAG;AACjB,WAAO;AAAA,EACT;AACF;;;AC5FA,SAAS,QAAAC,aAAY;AASrB,eAAsB,iBAAiB,YAAkD;AACvF,QAAM,QAAQ,MAAMC,MAAK,UAAU;AACnC,QAAM,MAAM,WAAW,MAAM,GAAG,EAAE,IAAI,GAAG,YAAY,KAAK;AAG1D,MAAI,MAAM,OAAO,GAAG;AAClB,QAAI,QAAQ,OAAO;AACjB,YAAM,YAAY,IAAI,aAAa,EAAE,WAAW,CAAC;AACjD,aAAO,UAAU,QAAQ;AAAA,IAC3B;AACA,QAAI,QAAQ,UAAU,QAAQ,OAAO;AACnC,YAAM,YAAY,IAAI,cAAc,EAAE,WAAW,CAAC;AAClD,aAAO,UAAU,QAAQ;AAAA,IAC3B;AACA,QAAI,QAAQ,QAAQ,QAAQ,YAAY;AACtC,YAAM,YAAY,IAAI,kBAAkB,EAAE,WAAW,CAAC;AACtD,aAAO,UAAU,QAAQ;AAAA,IAC3B;AACA,UAAM,IAAI,MAAM,4BAA4B,GAAG,EAAE;AAAA,EACnD;AAGA,MAAI,MAAM,YAAY,GAAG;AACvB,UAAM,eAAoC,CAAC;AAG3C,UAAM,gBAAgB,IAAI,cAAc,EAAE,WAAW,CAAC;AACtD,UAAM,WAAW,MAAM,cAAc,QAAQ;AAC7C,iBAAa,KAAK,GAAG,QAAQ;AAG7B,UAAM,cAAc,IAAI,kBAAkB,EAAE,WAAW,CAAC;AACxD,UAAM,SAAS,MAAM,YAAY,QAAQ;AACzC,iBAAa,KAAK,GAAG,MAAM;AAE3B,WAAO;AAAA,EACT;AAEA,QAAM,IAAI,MAAM,iBAAiB,UAAU,EAAE;AAC/C;","names":["readdir","readFile","join","basename","join","readFile","basename","readdir","readFile","readdir","stat","join","basename","stat","readFile","basename","readdir","join","stat","stat"]}

package/dist/index.d.ts

@@ -0,0 +1,78 @@
+import { R as RAGService, a as RAGOptions, b as RetrievalOptions, c as RAGResponse } from './types-CjnplPJD.js';
+export { C as ChunkMetadata, d as ChunkResult, e as ChunkingOptions, f as Citation, D as Document, g as DocumentChunk, E as EmbeddingBackend, h as EmbeddingOptions, i as EmbeddingResult, j as ExtractedDocument, k as RetrievalResult, S as SearchResult, V as VectorStore, l as VectorStoreOptions } from './types-CjnplPJD.js';
+export { BaseExtractor, CHMExtractor, HTMLExtractor, MarkdownExtractor, extractDocuments } from './extractors/index.js';
+export { HeaderAwareChunker } from './chunking/index.js';
+export { LOCAL_MODELS, LocalEmbeddingBackend, OPENAI_MODELS, OpenAIEmbeddingBackend, createEmbeddingBackend } from './embeddings/index.js';
+export { PostgresVectorStore, PostgresVectorStoreOptions, SQLiteVectorStore, SQLiteVectorStoreOptions, createVectorStore } from './vectorstore/index.js';
+export { DenseRetriever, HybridRetriever, HybridRetrieverOptions } from './retrieval/index.js';
+import 'pg';
+
+/**
+ * RAG Service
+ *
+ * Main service that orchestrates document indexing and question answering.
+ */
+
+declare class RAG implements RAGService {
+    private options;
+    private vectorStore;
+    private embeddings;
+    private retriever;
+    private llm;
+    private chunker;
+    private systemPrompt;
+    private initialized;
+    constructor(options: RAGOptions);
+    private createLLM;
+    /**
+     * Initialize the RAG service
+     */
+    initialize(): Promise<void>;
+    /**
+     * Index documents from a path
+     */
+    index(path: string, options?: {
+        sourceId?: string;
+    }): Promise<{
+        documentsIndexed: number;
+        chunksCreated: number;
+    }>;
+    /**
+     * Ask a question and get an answer with citations
+     */
+    ask(question: string, options?: RetrievalOptions): Promise<RAGResponse>;
+    /**
+     * Build context text from chunks
+     */
+    private buildContextText;
+    /**
+     * Generate answer using LLM
+     */
+    private generateAnswer;
+    /**
+     * Extract citations from answer text
+     */
+    private extractCitations;
+    /**
+     * Get document count
+     */
+    getDocumentCount(): Promise<number>;
+    /**
+     * Get chunk count
+     */
+    getChunkCount(): Promise<number>;
+    /**
+     * Clear all indexed data
+     */
+    clear(): Promise<void>;
+    /**
+     * Close connections
+     */
+    close(): Promise<void>;
+}
+/**
+ * Create a RAG service instance
+ */
+declare function createRAG(options: RAGOptions): Promise<RAG>;
+
+export { RAG, RAGOptions, RAGResponse, RAGService, RetrievalOptions, createRAG };