abb-opencode-local-rag 0.1.0

package/dist/parser/index.js

@@ -0,0 +1,454 @@
+// DocumentParser implementation with PDF/DOCX/TXT/MD support
+import { statSync } from 'node:fs';
+import { lstat, readFile, realpath } from 'node:fs/promises';
+import { basename, extname, isAbsolute, resolve } from 'node:path';
+import { JSDOM } from 'jsdom';
+import mammoth from 'mammoth';
+import { SemanticChunker } from '../chunker/index.js';
+import { withTrailingSeparator } from '../utils/base-dirs.js';
+import { AppError, isAppError } from '../utils/errors.js';
+import { convertDocxDocumentToText, extractDocxCoreTitle } from './docx-parser.js';
+import { extractPdfPages } from './pdf-extract.js';
+import { extractDocxTitle, extractMarkdownTitle, extractPdfTitle, extractTxtTitle, } from './title-extractor.js';
+// ============================================
+// Supported Extensions
+// ============================================
+/**
+ * File extensions supported by the parser module (parseFile + parsePdf).
+ * Exported so other modules (e.g. list_files) stay in sync automatically
+ * when new formats are added here.
+ */
+export const SUPPORTED_EXTENSIONS = new Set(['.pdf', '.docx', '.txt', '.md']);
+async function resolvePdfTitle(filePath, pages, metadataTitle, page1FontHint, embedder) {
+    const fileName = basename(filePath);
+    let firstPageChunkText;
+    try {
+        const filteredPage1 = pages[0]?.text;
+        if (filteredPage1 && filteredPage1.trim().length > 0) {
+            const page1Chunks = await new SemanticChunker().chunkText(filteredPage1, embedder);
+            firstPageChunkText = page1Chunks[0]?.text;
+        }
+    }
+    catch (titleError) {
+        if (isAppError(titleError))
+            throw titleError;
+        console.error(`Title extraction failed, falling back to filename: ${titleError}`);
+    }
+    return extractPdfTitle(metadataTitle, firstPageChunkText, fileName, page1FontHint).title;
+}
+/**
+ * Validation error (equivalent to 400)
+ */
+export class ValidationError extends AppError {
+    constructor(message, cause) {
+        super(message, 'parser', 'validation', cause);
+        this.name = 'ValidationError';
+    }
+}
+/**
+ * File operation error (equivalent to 500)
+ */
+export class FileOperationError extends AppError {
+    constructor(message, cause) {
+        super(message, 'parser', 'io', cause);
+        this.name = 'FileOperationError';
+    }
+}
+// ============================================
+// DocumentParser Class
+// ============================================
+/**
+ * Document parser class (PDF/DOCX/TXT/MD support)
+ *
+ * Responsibilities:
+ * - File path validation (path traversal prevention)
+ * - File size validation (100MB limit)
+ * - Parse 4 formats (PDF/DOCX/TXT/MD)
+ */
+export class DocumentParser {
+    config;
+    /** Raw allowed roots in input order (pre-realpath). Always non-empty. */
+    rawBaseDirs;
+    /**
+     * Lazily cached realpath-normalized allowed roots, each with a trailing
+     * path separator so the `startsWith` check is sibling-prefix safe (e.g.
+     * `/foo/bar/` must not match `/foo/barista/x.txt`). Order is preserved
+     * from `rawBaseDirs` so the legacy single-root rejection message keeps
+     * referencing the user-configured first root. Assumes the allowed roots
+     * are stable for the process lifetime.
+     */
+    resolvedBaseDirs = null;
+    constructor(config) {
+        this.config = config;
+        // Normalize the two accepted shapes into one internal raw-root list.
+        // The type system already rejects supplying both fields simultaneously,
+        // but defensively pick `baseDirs` first so a future relaxation does not
+        // accidentally fall back to the legacy single-root field.
+        //
+        // Empty `baseDirs` is accepted here so the parser can be constructed in
+        // the MCP server's degraded mode (configError present); `validateFilePath`
+        // fails closed in that case so no file is accepted while the empty root
+        // list stands. This is the only legitimate way to reach empty
+        // `rawBaseDirs`; production wiring always supplies a non-empty list when
+        // `configError` is absent.
+        if (config.baseDirs !== undefined) {
+            this.rawBaseDirs = config.baseDirs;
+        }
+        else {
+            this.rawBaseDirs = [config.baseDir];
+        }
+    }
+    /**
+     * File path validation (Absolute path requirement + Path traversal prevention).
+     *
+     * This is THE place realpath is used (with the base-dir resolver): the
+     * security/containment boundary. Following symlinks here makes prefix
+     * containment unforgeable. Stored/scanned/looked-up paths elsewhere use
+     * resolve() — see {@link BaseDirsConfig} for the path policy.
+     *
+     * Multi-root semantics: a file is accepted iff its realpath (or, for a
+     * non-symlink path that does not yet exist, its `resolve()`-normalized
+     * absolute path) is under ANY realpath-normalized allowed root using a
+     * trailing-separator prefix check. Broken symlinks are still rejected
+     * outright — the lstat-based detection mirrors the previous single-root
+     * behavior.
+     *
+     * Under a single allowed root the behavior is identical to the previous
+     * single-root implementation.
+     *
+     * @param filePath - File path to validate (must be absolute)
+     * @throws ValidationError - When path is not absolute or outside all allowed roots
+     */
+    async validateFilePath(filePath) {
+        // Fail-closed in degraded mode: when the parser was constructed with an
+        // empty allow-list (only legitimate when the MCP server is in degraded
+        // mode with a configError set), reject every path with a structured
+        // error rather than performing the realpath check against an empty
+        // surviving-roots set. Server-level `assertConfigOk` should have fired
+        // first; this is a defense-in-depth fallback for code paths that
+        // bypass that gate.
+        if (this.rawBaseDirs.length === 0) {
+            throw new ValidationError('No configured base directory: file access is disabled. Resolve the BASE_DIR / BASE_DIRS configuration error reported by the `status` tool before retrying.');
+        }
+        // Check if path is absolute (fast-fail without syscall)
+        if (!isAbsolute(filePath)) {
+            throw new ValidationError(`File path must be absolute path (received: ${filePath}). Please provide an absolute path within a configured base directory (BASE_DIR/BASE_DIRS/--base-dir).`);
+        }
+        // Lazily resolve and cache the real path of each allowed root (follows
+        // symlinks). Each entry gets a trailing separator so subsequent
+        // `startsWith` checks are sibling-prefix safe.
+        if (!this.resolvedBaseDirs) {
+            const resolvedList = [];
+            for (const raw of this.rawBaseDirs) {
+                const resolved = await realpath(resolve(raw));
+                resolvedList.push(withTrailingSeparator(resolved));
+            }
+            this.resolvedBaseDirs = resolvedList;
+        }
+        // Resolve the real path of the file (follows symlinks)
+        let resolvedPath;
+        try {
+            resolvedPath = await realpath(filePath);
+        }
+        catch (error) {
+            // realpath fails if path doesn't exist on filesystem.
+            // Distinguish broken symlinks from genuinely non-existent paths:
+            // - Broken symlink: lstat succeeds (symlink entry exists) -> reject
+            // - Non-existent path: lstat fails -> fall back to resolve() for validation
+            const isSymlink = await lstat(filePath)
+                .then((stats) => stats.isSymbolicLink())
+                .catch(() => false);
+            if (isSymlink) {
+                throw new ValidationError(`Cannot resolve file path: ${filePath}. The file may not exist or is a broken symlink.`, error);
+            }
+            // File doesn't exist at all - fall back to resolve() for path validation.
+            // Note: resolve() is string-based and cannot detect symlinked parent directories.
+            // This is acceptable because non-existent files will fail at subsequent readFile/statSync.
+            resolvedPath = resolve(filePath);
+        }
+        // Check if resolved path is within any allowed root.
+        const allowed = this.resolvedBaseDirs.some((root) => resolvedPath.startsWith(root));
+        if (!allowed) {
+            const rootsDisplay = this.resolvedBaseDirs.length === 1
+                ? this.resolvedBaseDirs[0]
+                : this.resolvedBaseDirs.join(', ');
+            throw new ValidationError(`File path must be within a configured base directory (BASE_DIR/BASE_DIRS/--base-dir). Allowed roots: ${rootsDisplay}. Received path outside all configured roots: ${filePath}`);
+        }
+    }
+    /**
+     * File size validation (100MB limit)
+     *
+     * @param filePath - File path to validate
+     * @throws ValidationError - When file size exceeds limit
+     * @throws FileOperationError - When file read fails
+     */
+    validateFileSize(filePath) {
+        try {
+            const stats = statSync(filePath);
+            if (stats.size > this.config.maxFileSize) {
+                throw new ValidationError(`File size exceeds limit: ${stats.size} > ${this.config.maxFileSize}`);
+            }
+        }
+        catch (error) {
+            if (error instanceof ValidationError) {
+                throw error;
+            }
+            // Missing file is an input error, not an I/O fault.
+            if (error.code === 'ENOENT') {
+                throw new ValidationError(`File not found: ${filePath}`);
+            }
+            throw new FileOperationError(`Failed to check file size: ${filePath}`, error);
+        }
+    }
+    /**
+     * File parsing (auto format detection)
+     *
+     * @param filePath - File path to parse
+     * @returns ParseResult with content and extracted title
+     * @throws ValidationError - Path traversal, size exceeded, unsupported format
+     * @throws FileOperationError - File read failed, parse failed
+     */
+    async parseFile(filePath, options = {}) {
+        // Validation
+        await this.validateFilePath(filePath);
+        this.validateFileSize(filePath);
+        // Format detection (PDF uses parsePdf directly)
+        const ext = extname(filePath).toLowerCase();
+        switch (ext) {
+            case '.docx':
+                return await this.parseDocx(filePath, options.images === true);
+            case '.txt':
+                return await this.parseTxt(filePath);
+            case '.md':
+                return await this.parseMd(filePath);
+            default:
+                throw new ValidationError(`Unsupported file format: ${ext}`);
+        }
+    }
+    /**
+     * PDF parsing with header/footer filtering
+     *
+     * Features:
+     * - Extracts text with position information (x, y, fontSize)
+     * - Semantic header/footer detection using embedding similarity
+     * - Uses hasEOL for proper line break handling
+     * - Extracts document title from PDF metadata and first page font heuristic
+     *
+     * @param filePath - PDF file path
+     * @param embedder - Embedder for semantic header/footer detection
+     * @returns ParseResult with content and extracted title
+     * @throws FileOperationError - File read failed, parse failed
+     */
+    async parsePdf(filePath, embedder) {
+        // Validation
+        await this.validateFilePath(filePath);
+        this.validateFileSize(filePath);
+        // Hold `doc` outside the try so the `finally` block can dispose it after
+        // either a successful return or an error from `extractPdfPages` / the
+        // post-processing steps. `doc` stays `undefined` if `openDocument` itself
+        // throws — in that case there is no handle to destroy.
+        let doc;
+        try {
+            const buffer = await readFile(filePath);
+            const mupdf = await import('mupdf');
+            doc = mupdf.Document.openDocument(buffer, 'application/pdf');
+            const { pages, metadataTitle, page1FontHint } = await extractPdfPages(doc, embedder, 'preserve-whitespace');
+            const text = pages
+                .map((p) => p.text)
+                .filter((t) => t.length > 0)
+                .join('\n\n');
+            const title = await resolvePdfTitle(filePath, pages, metadataTitle, page1FontHint, embedder);
+            console.error(`Parsed PDF: ${filePath} (${text.length} characters, ${pages.length} pages)`);
+            return { content: text, title };
+        }
+        catch (error) {
+            // A foreign domain error (e.g. `EmbeddingError` raised while the parser
+            // uses the embedder) keeps its identity — rethrow it unchanged instead
+            // of relabeling it as a PDF parse failure. The parser's own
+            // `ValidationError`/`FileOperationError` are also `AppError` and so
+            // rethrow as-is, preserving their identity. Only a genuine non-`AppError`
+            // IO/mupdf failure wraps as `FileOperationError` with its `.cause` set.
+            if (isAppError(error)) {
+                throw error;
+            }
+            throw new FileOperationError(`Failed to parse PDF: ${filePath}`, error);
+        }
+        finally {
+            // Release the native WASM handle exactly once per invocation, on both
+            // success and error paths.
+            doc?.destroy();
+        }
+    }
+    /**
+     * Per-page PDF parsing for the visual-enrichment path.
+     *
+     * Opens a mupdf `Document`, delegates per-page extraction to the shared
+     * `extractPdfPages` helper with the `'preserve-whitespace,preserve-images'`
+     * stext option string so mupdf emits `block.type === 'image'` blocks for
+     * the downstream visual-candidate detector.
+     *
+     * Returns the open `Document` handle alongside the per-page records and
+     * title-resolution materials so the caller can:
+     *   - run the renderer (`page.toPixmap()`) on the same handle,
+     *   - feed `metadataTitle` + `pages[0].page1FontHint` into `extractPdfTitle`
+     *     after `buildChunksAndEmbeddings` returns.
+     *
+     * Disposal contract (asymmetric — read carefully):
+     *   - SUCCESS path: this method returns the open `doc` handle. The caller
+     *     owns disposal and MUST wrap the call site in
+     *     `try { ... } finally { doc.destroy() }`.
+     *   - ERROR path: when this method throws, `doc` has already been destroyed
+     *     internally before the exception propagates (so the caller never
+     *     receives a handle it would not know to clean up). Callers MUST NOT
+     *     call `doc.destroy()` on an error from this method.
+     * Title resolution stays here so visual captions cannot change PDF metadata.
+     * Visual candidate selection remains `pdf-visual/detector`'s responsibility.
+     *
+     * @param filePath - PDF file path (validated against BASE_DIR and size limit)
+     * @param embedder - Embedder for semantic header/footer detection
+     * @returns Open mupdf `Document`, resolved title, and per-page text/layout records.
+     * @throws ValidationError - Path traversal, size exceeded
+     * @throws FileOperationError - File read or parse failed (after destroying `doc` internally)
+     */
+    async parsePdfPages(filePath, embedder) {
+        // Validation (mirrors parsePdf's entry-point contract so the visual path
+        // does not bypass BASE_DIR / size checks).
+        await this.validateFilePath(filePath);
+        this.validateFileSize(filePath);
+        // Open the doc and run per-page extraction. Success-path disposal of
+        // `doc` stays with the caller.
+        // For the error-path window between `openDocument` and the return below,
+        // destroy `doc` here before re-throwing so a failure in `extractPdfPages`
+        // (or any future pre-return step) does not leak the mupdf WASM handle.
+        let doc;
+        try {
+            const buffer = await readFile(filePath);
+            const mupdf = await import('mupdf');
+            doc = mupdf.Document.openDocument(buffer, 'application/pdf');
+            const extracted = await extractPdfPages(doc, embedder, 'preserve-whitespace,preserve-images');
+            const { pages: helperPages, metadataTitle, page1FontHint } = extracted;
+            const pages = helperPages.map((page) => ({
+                pageNum: page.pageNum,
+                text: page.text,
+                textFragments: page.textFragments,
+                stextJson: page.stextJson,
+            }));
+            const title = await resolvePdfTitle(filePath, helperPages, metadataTitle, page1FontHint, embedder);
+            console.error(`Parsed PDF pages: ${filePath} (${pages.length} pages; caller owns doc disposal)`);
+            return { doc, title, pages };
+        }
+        catch (error) {
+            // `doc` is undefined when `openDocument` itself threw — nothing to free.
+            // When it is defined, dispose before re-throwing (on BOTH the foreign and
+            // the genuine error paths) so the caller never receives the handle and
+            // cannot be expected to clean it up.
+            doc?.destroy();
+            // A foreign domain error (e.g. `EmbeddingError`) keeps its identity —
+            // rethrow it unchanged. Only a genuine non-`AppError` IO/mupdf failure
+            // wraps as `FileOperationError` with its `.cause` set.
+            if (isAppError(error)) {
+                throw error;
+            }
+            throw new FileOperationError(`Failed to parse PDF pages: ${filePath}`, error);
+        }
+    }
+    /**
+     * DOCX parsing (using mammoth)
+     *
+     * Uses DOCX package metadata plus one Mammoth HTML conversion for title and body extraction.
+     *
+     * @param filePath - DOCX file path
+     * @returns ParseResult with content and extracted title
+     * @throws FileOperationError - File read failed, parse failed
+     */
+    async parseDocx(filePath, includeImages) {
+        try {
+            const buffer = await readFile(filePath);
+            const capturedImages = new Map();
+            let nextImageIndex = 0;
+            let skippedImageCount = 0;
+            const htmlResult = includeImages
+                ? await mammoth.convertToHtml({ buffer }, {
+                    convertImage: mammoth.images.imgElement(async (image) => {
+                        const imageIndex = nextImageIndex++;
+                        if (image.contentType !== 'image/png' && image.contentType !== 'image/jpeg') {
+                            skippedImageCount += 1;
+                            return { src: `data:${image.contentType};base64,` };
+                        }
+                        const bytes = await image.readAsBuffer().catch(() => null);
+                        if (bytes === null) {
+                            skippedImageCount += 1;
+                            return { src: `data:${image.contentType};base64,` };
+                        }
+                        capturedImages.set(imageIndex, { mimeType: image.contentType, bytes });
+                        return {
+                            src: `data:${image.contentType};base64,`,
+                            'data-rag-image-index': String(imageIndex),
+                        };
+                    }),
+                })
+                : await mammoth.convertToHtml({ buffer });
+            const htmlDocument = new JSDOM(htmlResult.value).window.document;
+            const coreTitle = await extractDocxCoreTitle(buffer);
+            const body = convertDocxDocumentToText(htmlDocument);
+            const fileName = basename(filePath);
+            const titleResult = extractDocxTitle(htmlDocument, fileName, coreTitle);
+            console.error(`Parsed DOCX: ${filePath} (${body.content.length} characters)`);
+            if (skippedImageCount > 0) {
+                console.warn(`Skipped ${skippedImageCount} unsupported or unreadable DOCX image(s) in ${filePath}; only embedded PNG and JPEG are supported`);
+            }
+            const imageAnchors = (body.imageAnchors ?? []).flatMap((anchor) => {
+                const captured = capturedImages.get(anchor.imageIndex);
+                return captured ? [{ ...anchor, ...captured }] : [];
+            });
+            return {
+                content: body.content,
+                title: titleResult.title,
+                ...(body.atomicRanges.length === 0 ? {} : { atomicRanges: body.atomicRanges }),
+                ...(imageAnchors.length === 0 ? {} : { imageAnchors }),
+            };
+        }
+        catch (error) {
+            throw new FileOperationError(`Failed to parse DOCX: ${filePath}`, error);
+        }
+    }
+    /**
+     * TXT parsing (using fs.readFile)
+     *
+     * @param filePath - TXT file path
+     * @returns ParseResult with content and extracted title
+     * @throws FileOperationError - File read failed
+     */
+    async parseTxt(filePath) {
+        try {
+            const text = await readFile(filePath, 'utf-8');
+            const fileName = basename(filePath);
+            const titleResult = extractTxtTitle(text, fileName);
+            console.error(`Parsed TXT: ${filePath} (${text.length} characters)`);
+            return { content: text, title: titleResult.title };
+        }
+        catch (error) {
+            throw new FileOperationError(`Failed to parse TXT: ${filePath}`, error);
+        }
+    }
+    /**
+     * MD parsing (using fs.readFile)
+     *
+     * @param filePath - MD file path
+     * @returns ParseResult with content and extracted title
+     * @throws FileOperationError - File read failed
+     */
+    async parseMd(filePath) {
+        try {
+            const text = await readFile(filePath, 'utf-8');
+            const fileName = basename(filePath);
+            const titleResult = extractMarkdownTitle(text, fileName);
+            console.error(`Parsed MD: ${filePath} (${text.length} characters)`);
+            return { content: text, title: titleResult.title };
+        }
+        catch (error) {
+            throw new FileOperationError(`Failed to parse MD: ${filePath}`, error);
+        }
+    }
+}
+//# sourceMappingURL=index.js.map

package/dist/parser/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/parser/index.ts"],"names":[],"mappings":"AAAA,6DAA6D;AAE7D,OAAO,EAAE,QAAQ,EAAE,MAAM,SAAS,CAAA;AAClC,OAAO,EAAE,KAAK,EAAE,QAAQ,EAAE,QAAQ,EAAE,MAAM,kBAAkB,CAAA;AAC5D,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,UAAU,EAAE,OAAO,EAAE,MAAM,WAAW,CAAA;AAClE,OAAO,EAAE,KAAK,EAAE,MAAM,OAAO,CAAA;AAC7B,OAAO,OAAO,MAAM,SAAS,CAAA;AAE7B,OAAO,EAAwB,eAAe,EAAE,MAAM,qBAAqB,CAAA;AAC3E,OAAO,EAAE,qBAAqB,EAAE,MAAM,uBAAuB,CAAA;AAC7D,OAAO,EAAE,QAAQ,EAAE,UAAU,EAAE,MAAM,oBAAoB,CAAA;AACzD,OAAO,EAAE,yBAAyB,EAAE,oBAAoB,EAAE,MAAM,kBAAkB,CAAA;AAClF,OAAO,EAAE,eAAe,EAAE,MAAM,kBAAkB,CAAA;AAElD,OAAO,EACL,gBAAgB,EAChB,oBAAoB,EACpB,eAAe,EACf,eAAe,GAChB,MAAM,sBAAsB,CAAA;AAE7B,+CAA+C;AAC/C,uBAAuB;AACvB,+CAA+C;AAE/C;;;;GAIG;AACH,MAAM,CAAC,MAAM,oBAAoB,GAAG,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,CAAA;AA4B7E,KAAK,UAAU,eAAe,CAC5B,QAAgB,EAChB,KAAkC,EAClC,aAAiC,EACjC,aAA6D,EAC7D,QAA2B;IAE3B,MAAM,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC,CAAA;IACnC,IAAI,kBAAsC,CAAA;IAC1C,IAAI,CAAC;QACH,MAAM,aAAa,GAAG,KAAK,CAAC,CAAC,CAAC,EAAE,IAAI,CAAA;QACpC,IAAI,aAAa,IAAI,aAAa,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YACrD,MAAM,WAAW,GAAG,MAAM,IAAI,eAAe,EAAE,CAAC,SAAS,CAAC,aAAa,EAAE,QAAQ,CAAC,CAAA;YAClF,kBAAkB,GAAG,WAAW,CAAC,CAAC,CAAC,EAAE,IAAI,CAAA;QAC3C,CAAC;IACH,CAAC;IAAC,OAAO,UAAU,EAAE,CAAC;QACpB,IAAI,UAAU,CAAC,UAAU,CAAC;YAAE,MAAM,UAAU,CAAA;QAC5C,OAAO,CAAC,KAAK,CAAC,sDAAsD,UAAU,EAAE,CAAC,CAAA;IACnF,CAAC;IACD,OAAO,eAAe,CAAC,aAAa,EAAE,kBAAkB,EAAE,QAAQ,EAAE,aAAa,CAAC,CAAC,KAAK,CAAA;AAC1F,CAAC;AAiCD;;GAEG;AACH,MAAM,OAAO,eAAgB,SAAQ,QAAQ;IAC3C,YAAY,OAAe,EAAE,KAAa;QACxC,KAAK,CAAC,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,KAAK,CAAC,CAAA;QAC7C,IAAI,CAAC,IAAI,GAAG,iBAAiB,CAAA;IAC/B,CAAC;CACF;AAED;;GAEG;AACH,MAAM,OAAO,kBAAmB,SAAQ,QAAQ;IAC9C,YAAY,OAAe,EAAE,KAAa;QACxC,KAAK,CAAC,OAAO,EAAE,QAAQ,EAAE,IAAI,EAAE,KAAK,CAAC,CAAA;QACrC,IAAI,CAAC,IAAI,GAAG,oBAAoB,CAAA;IAClC,CAAC;CACF;AAED,+CAA+C;AAC/C,uBAAuB;AACvB,+CAA+C;AAE/C;;;;;;;GAOG;AACH,MAAM,OAAO,cAAc;IACR,MAAM,CAAc;IACrC,yEAAyE;IACxD,WAAW,CAAmB;IAC/C;;;;;;;OAOG;IACK,gBAAgB,GAAoB,IAAI,CAAA;IAEhD,YAAY,MAAoB;QAC9B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAA;QACpB,qEAAqE;QACrE,wEAAwE;QACxE,wEAAwE;QACxE,0DAA0D;QAC1D,EAAE;QACF,wEAAwE;QACxE,2EAA2E;QAC3E,wEAAwE;QACxE,8DAA8D;QAC9D,yEAAyE;QACzE,2BAA2B;QAC3B,IAAI,MAAM,CAAC,QAAQ,KAAK,SAAS,EAAE,CAAC;YAClC,IAAI,CAAC,WAAW,GAAG,MAAM,CAAC,QAAQ,CAAA;QACpC,CAAC;aAAM,CAAC;YACN,IAAI,CAAC,WAAW,GAAG,CAAC,MAAM,CAAC,OAAO,CAAC,CAAA;QACrC,CAAC;IACH,CAAC;IAED;;;;;;;;;;;;;;;;;;;;OAoBG;IACH,KAAK,CAAC,gBAAgB,CAAC,QAAgB;QACrC,wEAAwE;QACxE,uEAAuE;QACvE,oEAAoE;QACpE,mEAAmE;QACnE,uEAAuE;QACvE,iEAAiE;QACjE,oBAAoB;QACpB,IAAI,IAAI,CAAC,WAAW,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAClC,MAAM,IAAI,eAAe,CACvB,4JAA4J,CAC7J,CAAA;QACH,CAAC;QAED,wDAAwD;QACxD,IAAI,CAAC,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC;YAC1B,MAAM,IAAI,eAAe,CACvB,8CAA8C,QAAQ,wGAAwG,CAC/J,CAAA;QACH,CAAC;QAED,uEAAuE;QACvE,gEAAgE;QAChE,+CAA+C;QAC/C,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC3B,MAAM,YAAY,GAAa,EAAE,CAAA;YACjC,KAAK,MAAM,GAAG,IAAI,IAAI,CAAC,WAAW,EAAE,CAAC;gBACnC,MAAM,QAAQ,GAAG,MAAM,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAA;gBAC7C,YAAY,CAAC,IAAI,CAAC,qBAAqB,CAAC,QAAQ,CAAC,CAAC,CAAA;YACpD,CAAC;YACD,IAAI,CAAC,gBAAgB,GAAG,YAAY,CAAA;QACtC,CAAC;QAED,uDAAuD;QACvD,IAAI,YAAoB,CAAA;QACxB,IAAI,CAAC;YACH,YAAY,GAAG,MAAM,QAAQ,CAAC,QAAQ,CAAC,CAAA;QACzC,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,sDAAsD;YACtD,iEAAiE;YACjE,oEAAoE;YACpE,4EAA4E;YAC5E,MAAM,SAAS,GAAG,MAAM,KAAK,CAAC,QAAQ,CAAC;iBACpC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,cAAc,EAAE,CAAC;iBACvC,KAAK,CAAC,GAAG,EAAE,CAAC,KAAK,CAAC,CAAA;YAErB,IAAI,SAAS,EAAE,CAAC;gBACd,MAAM,IAAI,eAAe,CACvB,6BAA6B,QAAQ,kDAAkD,EACvF,KAAc,CACf,CAAA;YACH,CAAC;YAED,0EAA0E;YAC1E,kFAAkF;YAClF,2FAA2F;YAC3F,YAAY,GAAG,OAAO,CAAC,QAAQ,CAAC,CAAA;QAClC,CAAC;QAED,qDAAqD;QACrD,MAAM,OAAO,GAAG,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,YAAY,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,CAAA;QACnF,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,MAAM,YAAY,GAChB,IAAI,CAAC,gBAAgB,CAAC,MAAM,KAAK,CAAC;gBAChC,CAAC,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC,CAAC;gBAC1B,CAAC,CAAC,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;YACtC,MAAM,IAAI,eAAe,CACvB,wGAAwG,YAAY,iDAAiD,QAAQ,EAAE,CAChL,CAAA;QACH,CAAC;IACH,CAAC;IAED;;;;;;OAMG;IACH,gBAAgB,CAAC,QAAgB;QAC/B,IAAI,CAAC;YACH,MAAM,KAAK,GAAG,QAAQ,CAAC,QAAQ,CAAC,CAAA;YAChC,IAAI,KAAK,CAAC,IAAI,GAAG,IAAI,CAAC,MAAM,CAAC,WAAW,EAAE,CAAC;gBACzC,MAAM,IAAI,eAAe,CACvB,4BAA4B,KAAK,CAAC,IAAI,MAAM,IAAI,CAAC,MAAM,CAAC,WAAW,EAAE,CACtE,CAAA;YACH,CAAC;QACH,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,IAAI,KAAK,YAAY,eAAe,EAAE,CAAC;gBACrC,MAAM,KAAK,CAAA;YACb,CAAC;YACD,oDAAoD;YACpD,IAAK,KAA+B,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;gBACvD,MAAM,IAAI,eAAe,CAAC,mBAAmB,QAAQ,EAAE,CAAC,CAAA;YAC1D,CAAC;YACD,MAAM,IAAI,kBAAkB,CAAC,8BAA8B,QAAQ,EAAE,EAAE,KAAc,CAAC,CAAA;QACxF,CAAC;IACH,CAAC;IAED;;;;;;;OAOG;IACH,KAAK,CAAC,SAAS,CAAC,QAAgB,EAAE,OAAO,GAAqB,EAAE;QAC9D,aAAa;QACb,MAAM,IAAI,CAAC,gBAAgB,CAAC,QAAQ,CAAC,CAAA;QACrC,IAAI,CAAC,gBAAgB,CAAC,QAAQ,CAAC,CAAA;QAE/B,gDAAgD;QAChD,MAAM,GAAG,GAAG,OAAO,CAAC,QAAQ,CAAC,CAAC,WAAW,EAAE,CAAA;QAC3C,QAAQ,GAAG,EAAE,CAAC;YACZ,KAAK,OAAO;gBACV,OAAO,MAAM,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,CAAA;YAChE,KAAK,MAAM;gBACT,OAAO,MAAM,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAA;YACtC,KAAK,KAAK;gBACR,OAAO,MAAM,IAAI,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAA;YACrC;gBACE,MAAM,IAAI,eAAe,CAAC,4BAA4B,GAAG,EAAE,CAAC,CAAA;QAChE,CAAC;IACH,CAAC;IAED;;;;;;;;;;;;;OAaG;IACH,KAAK,CAAC,QAAQ,CAAC,QAAgB,EAAE,QAA2B;QAC1D,aAAa;QACb,MAAM,IAAI,CAAC,gBAAgB,CAAC,QAAQ,CAAC,CAAA;QACrC,IAAI,CAAC,gBAAgB,CAAC,QAAQ,CAAC,CAAA;QAE/B,yEAAyE;QACzE,sEAAsE;QACtE,0EAA0E;QAC1E,uDAAuD;QACvD,IAAI,GAA8B,CAAA;QAClC,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,QAAQ,CAAC,QAAQ,CAAC,CAAA;YACvC,MAAM,KAAK,GAAG,MAAM,MAAM,CAAC,OAAO,CAAC,CAAA;YACnC,GAAG,GAAG,KAAK,CAAC,QAAQ,CAAC,YAAY,CAAC,MAAM,EAAE,iBAAiB,CAAkB,CAAA;YAE7E,MAAM,EAAE,KAAK,EAAE,aAAa,EAAE,aAAa,EAAE,GAAG,MAAM,eAAe,CACnE,GAAG,EACH,QAAQ,EACR,qBAAqB,CACtB,CAAA;YACD,MAAM,IAAI,GAAG,KAAK;iBACf,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC;iBAClB,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC;iBAC3B,IAAI,CAAC,MAAM,CAAC,CAAA;YAEf,MAAM,KAAK,GAAG,MAAM,eAAe,CAAC,QAAQ,EAAE,KAAK,EAAE,aAAa,EAAE,aAAa,EAAE,QAAQ,CAAC,CAAA;YAE5F,OAAO,CAAC,KAAK,CAAC,eAAe,QAAQ,KAAK,IAAI,CAAC,MAAM,gBAAgB,KAAK,CAAC,MAAM,SAAS,CAAC,CAAA;YAE3F,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,KAAK,EAAE,CAAA;QACjC,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,wEAAwE;YACxE,uEAAuE;YACvE,4DAA4D;YAC5D,oEAAoE;YACpE,0EAA0E;YAC1E,wEAAwE;YACxE,IAAI,UAAU,CAAC,KAAK,CAAC,EAAE,CAAC;gBACtB,MAAM,KAAK,CAAA;YACb,CAAC;YACD,MAAM,IAAI,kBAAkB,CAAC,wBAAwB,QAAQ,EAAE,EAAE,KAAc,CAAC,CAAA;QAClF,CAAC;gBAAS,CAAC;YACT,sEAAsE;YACtE,2BAA2B;YAC3B,GAAG,EAAE,OAAO,EAAE,CAAA;QAChB,CAAC;IACH,CAAC;IAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;OA8BG;IACH,KAAK,CAAC,aAAa,CACjB,QAAgB,EAChB,QAA2B;QAW3B,yEAAyE;QACzE,2CAA2C;QAC3C,MAAM,IAAI,CAAC,gBAAgB,CAAC,QAAQ,CAAC,CAAA;QACrC,IAAI,CAAC,gBAAgB,CAAC,QAAQ,CAAC,CAAA;QAE/B,qEAAqE;QACrE,+BAA+B;QAC/B,yEAAyE;QACzE,0EAA0E;QAC1E,uEAAuE;QACvE,IAAI,GAA8B,CAAA;QAClC,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,QAAQ,CAAC,QAAQ,CAAC,CAAA;YACvC,MAAM,KAAK,GAAG,MAAM,MAAM,CAAC,OAAO,CAAC,CAAA;YACnC,GAAG,GAAG,KAAK,CAAC,QAAQ,CAAC,YAAY,CAAC,MAAM,EAAE,iBAAiB,CAAkB,CAAA;YAC7E,MAAM,SAAS,GAAG,MAAM,eAAe,CAAC,GAAG,EAAE,QAAQ,EAAE,qCAAqC,CAAC,CAAA;YAE7F,MAAM,EAAE,KAAK,EAAE,WAAW,EAAE,aAAa,EAAE,aAAa,EAAE,GAAG,SAAS,CAAA;YACtE,MAAM,KAAK,GAAG,WAAW,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;gBACvC,OAAO,EAAE,IAAI,CAAC,OAAO;gBACrB,IAAI,EAAE,IAAI,CAAC,IAAI;gBACf,aAAa,EAAE,IAAI,CAAC,aAAa;gBACjC,SAAS,EAAE,IAAI,CAAC,SAAS;aAC1B,CAAC,CAAC,CAAA;YACH,MAAM,KAAK,GAAG,MAAM,eAAe,CACjC,QAAQ,EACR,WAAW,EACX,aAAa,EACb,aAAa,EACb,QAAQ,CACT,CAAA;YAED,OAAO,CAAC,KAAK,CACX,qBAAqB,QAAQ,KAAK,KAAK,CAAC,MAAM,mCAAmC,CAClF,CAAA;YAED,OAAO,EAAE,GAAG,EAAE,KAAK,EAAE,KAAK,EAAE,CAAA;QAC9B,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,yEAAyE;YACzE,0EAA0E;YAC1E,uEAAuE;YACvE,qCAAqC;YACrC,GAAG,EAAE,OAAO,EAAE,CAAA;YACd,sEAAsE;YACtE,uEAAuE;YACvE,uDAAuD;YACvD,IAAI,UAAU,CAAC,KAAK,CAAC,EAAE,CAAC;gBACtB,MAAM,KAAK,CAAA;YACb,CAAC;YACD,MAAM,IAAI,kBAAkB,CAAC,8BAA8B,QAAQ,EAAE,EAAE,KAAc,CAAC,CAAA;QACxF,CAAC;IACH,CAAC;IAED;;;;;;;;OAQG;IACK,KAAK,CAAC,SAAS,CAAC,QAAgB,EAAE,aAAsB;QAC9D,IAAI,CAAC;YACH,MAAM,MAAM,GAAG,MAAM,QAAQ,CAAC,QAAQ,CAAC,CAAA;YACvC,MAAM,cAAc,GAAG,IAAI,GAAG,EAG3B,CAAA;YACH,IAAI,cAAc,GAAG,CAAC,CAAA;YACtB,IAAI,iBAAiB,GAAG,CAAC,CAAA;YACzB,MAAM,UAAU,GAAG,aAAa;gBAC9B,CAAC,CAAC,MAAM,OAAO,CAAC,aAAa,CACzB,EAAE,MAAM,EAAE,EACV;oBACE,YAAY,EAAE,OAAO,CAAC,MAAM,CAAC,UAAU,CAAC,KAAK,EAAE,KAAK,EAAE,EAAE;wBACtD,MAAM,UAAU,GAAG,cAAc,EAAE,CAAA;wBACnC,IAAI,KAAK,CAAC,WAAW,KAAK,WAAW,IAAI,KAAK,CAAC,WAAW,KAAK,YAAY,EAAE,CAAC;4BAC5E,iBAAiB,IAAI,CAAC,CAAA;4BACtB,OAAO,EAAE,GAAG,EAAE,QAAQ,KAAK,CAAC,WAAW,UAAU,EAAE,CAAA;wBACrD,CAAC;wBACD,MAAM,KAAK,GAAG,MAAM,KAAK,CAAC,YAAY,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,CAAA;wBAC1D,IAAI,KAAK,KAAK,IAAI,EAAE,CAAC;4BACnB,iBAAiB,IAAI,CAAC,CAAA;4BACtB,OAAO,EAAE,GAAG,EAAE,QAAQ,KAAK,CAAC,WAAW,UAAU,EAAE,CAAA;wBACrD,CAAC;wBACD,cAAc,CAAC,GAAG,CAAC,UAAU,EAAE,EAAE,QAAQ,EAAE,KAAK,CAAC,WAAW,EAAE,KAAK,EAAE,CAAC,CAAA;wBACtE,OAAO;4BACL,GAAG,EAAE,QAAQ,KAAK,CAAC,WAAW,UAAU;4BACxC,sBAAsB,EAAE,MAAM,CAAC,UAAU,CAAC;yBAC3C,CAAA;oBACH,CAAC,CAAC;iBACH,CACF;gBACH,CAAC,CAAC,MAAM,OAAO,CAAC,aAAa,CAAC,EAAE,MAAM,EAAE,CAAC,CAAA;YAC3C,MAAM,YAAY,GAAG,IAAI,KAAK,CAAC,UAAU,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAA;YAChE,MAAM,SAAS,GAAG,MAAM,oBAAoB,CAAC,MAAM,CAAC,CAAA;YACpD,MAAM,IAAI,GAAG,yBAAyB,CAAC,YAAY,CAAC,CAAA;YACpD,MAAM,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC,CAAA;YACnC,MAAM,WAAW,GAAG,gBAAgB,CAAC,YAAY,EAAE,QAAQ,EAAE,SAAS,CAAC,CAAA;YAEvE,OAAO,CAAC,KAAK,CAAC,gBAAgB,QAAQ,KAAK,IAAI,CAAC,OAAO,CAAC,MAAM,cAAc,CAAC,CAAA;YAC7E,IAAI,iBAAiB,GAAG,CAAC,EAAE,CAAC;gBAC1B,OAAO,CAAC,IAAI,CACV,WAAW,iBAAiB,+CAA+C,QAAQ,4CAA4C,CAChI,CAAA;YACH,CAAC;YACD,MAAM,YAAY,GAAG,CAAC,IAAI,CAAC,YAAY,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,MAAM,EAAE,EAAE;gBAChE,MAAM,QAAQ,GAAG,cAAc,CAAC,GAAG,CAAC,MAAM,CAAC,UAAU,CAAC,CAAA;gBACtD,OAAO,QAAQ,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG,MAAM,EAAE,GAAG,QAAQ,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,CAAA;YACrD,CAAC,CAAC,CAAA;YACF,OAAO;gBACL,OAAO,EAAE,IAAI,CAAC,OAAO;gBACrB,KAAK,EAAE,WAAW,CAAC,KAAK;gBACxB,GAAG,CAAC,IAAI,CAAC,YAAY,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,YAAY,EAAE,IAAI,CAAC,YAAY,EAAE,CAAC;gBAC9E,GAAG,CAAC,YAAY,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,YAAY,EAAE,CAAC;aACvD,CAAA;QACH,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,MAAM,IAAI,kBAAkB,CAAC,yBAAyB,QAAQ,EAAE,EAAE,KAAc,CAAC,CAAA;QACnF,CAAC;IACH,CAAC;IAED;;;;;;OAMG;IACK,KAAK,CAAC,QAAQ,CAAC,QAAgB;QACrC,IAAI,CAAC;YACH,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,QAAQ,EAAE,OAAO,CAAC,CAAA;YAC9C,MAAM,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC,CAAA;YACnC,MAAM,WAAW,GAAG,eAAe,CAAC,IAAI,EAAE,QAAQ,CAAC,CAAA;YACnD,OAAO,CAAC,KAAK,CAAC,eAAe,QAAQ,KAAK,IAAI,CAAC,MAAM,cAAc,CAAC,CAAA;YACpE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,KAAK,EAAE,WAAW,CAAC,KAAK,EAAE,CAAA;QACpD,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,MAAM,IAAI,kBAAkB,CAAC,wBAAwB,QAAQ,EAAE,EAAE,KAAc,CAAC,CAAA;QAClF,CAAC;IACH,CAAC;IAED;;;;;;OAMG;IACK,KAAK,CAAC,OAAO,CAAC,QAAgB;QACpC,IAAI,CAAC;YACH,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,QAAQ,EAAE,OAAO,CAAC,CAAA;YAC9C,MAAM,QAAQ,GAAG,QAAQ,CAAC,QAAQ,CAAC,CAAA;YACnC,MAAM,WAAW,GAAG,oBAAoB,CAAC,IAAI,EAAE,QAAQ,CAAC,CAAA;YACxD,OAAO,CAAC,KAAK,CAAC,cAAc,QAAQ,KAAK,IAAI,CAAC,MAAM,cAAc,CAAC,CAAA;YACnE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,KAAK,EAAE,WAAW,CAAC,KAAK,EAAE,CAAA;QACpD,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,MAAM,IAAI,kBAAkB,CAAC,uBAAuB,QAAQ,EAAE,EAAE,KAAc,CAAC,CAAA;QACjF,CAAC;IACH,CAAC;CACF"}

package/dist/parser/pdf-extract.d.ts

@@ -0,0 +1,81 @@
+import type { Document as MupdfDocument } from 'mupdf';
+import { type EmbedderInterface, type FilteredTextFragment } from './pdf-filter.js';
+interface StextBbox {
+    x: number;
+    y: number;
+    w: number;
+    h: number;
+}
+/**
+ * Shape of mupdf's structured-text JSON used by the per-page loop.
+ * Captured here so both the items-extraction step and the raw `stextJson`
+ * we return remain typed.
+ */
+interface StextJson {
+    blocks: Array<{
+        type: string;
+        lines?: Array<{
+            text: string;
+            x: number;
+            y: number;
+            bbox?: StextBbox;
+            font: {
+                size: number;
+                name?: string;
+                weight?: string;
+            };
+        }>;
+    }>;
+}
+/**
+ * Per-page record produced by `extractPdfPages`. `text` is the page's text
+ * after semantic header/footer filtering. `textFragments` carries every
+ * survivor's native provenance and exact range in `text`; `stextJson` is the
+ * copied raw mupdf structured-text JSON for downstream visual detection.
+ */
+interface ExtractedPage {
+    pageNum: number;
+    text: string;
+    textFragments: FilteredTextFragment[];
+    stextJson: StextJson;
+}
+/**
+ * Result returned by `extractPdfPages`. The helper lifts three concerns
+ * out of the legacy `parsePdf` body:
+ *   1. the per-page `toStructuredText` + `block.type === 'text'` loop;
+ *   2. `filterPageBoundarySentences` for header/footer removal;
+ *   3. title-resolution materials (`metadataTitle` and `page1FontHint`).
+ *
+ * Both `parsePdf` and `parsePdfPages` consume this helper; they differ only
+ * in the `stextOptions` argument they pass to `page.toStructuredText(...)`.
+ */
+interface ExtractedPdf {
+    pages: ExtractedPage[];
+    metadataTitle: string | undefined;
+    page1FontHint: {
+        text: string;
+        fontSize: number;
+    } | undefined;
+}
+/**
+ * Per-page extraction shared by `parsePdf` and `parsePdfPages`.
+ *
+ * Takes an already-open mupdf `Document` and:
+ *   - reads `info:Title` once,
+ *   - iterates pages calling `toStructuredText(stextOptions)`,
+ *   - builds `PageData` items (only `block.type === 'text'` lines),
+ *   - runs `filterPageBoundarySentences` to drop semantic headers/footers,
+ *   - derives `page1FontHint` from page 1's largest-font lines.
+ *
+ * The two callers differ ONLY in `stextOptions`: `parsePdf` passes
+ * `'preserve-whitespace'`;
+ * `parsePdfPages` passes `'preserve-whitespace,preserve-images'` so mupdf
+ * emits `block.type === 'image'` entries for the downstream visual-candidate
+ * detector.
+ *
+ * Lifecycle: this helper does NOT call `doc.destroy()` — disposal stays
+ * with the caller.
+ */
+export declare function extractPdfPages(doc: MupdfDocument, embedder: EmbedderInterface, stextOptions: string): Promise<ExtractedPdf>;
+export {};
+//# sourceMappingURL=pdf-extract.d.ts.map

package/dist/parser/pdf-extract.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"pdf-extract.d.ts","sourceRoot":"","sources":["../../src/parser/pdf-extract.ts"],"names":[],"mappings":"AAUA,OAAO,KAAK,EAAE,QAAQ,IAAI,aAAa,EAAE,MAAM,OAAO,CAAA;AACtD,OAAO,EACL,KAAK,iBAAiB,EACtB,KAAK,oBAAoB,EAG1B,MAAM,iBAAiB,CAAA;AAExB,UAAU,SAAS;IACjB,CAAC,EAAE,MAAM,CAAA;IACT,CAAC,EAAE,MAAM,CAAA;IACT,CAAC,EAAE,MAAM,CAAA;IACT,CAAC,EAAE,MAAM,CAAA;CACV;AAED;;;;GAIG;AACH,UAAU,SAAS;IACjB,MAAM,EAAE,KAAK,CAAC;QACZ,IAAI,EAAE,MAAM,CAAA;QACZ,KAAK,CAAC,EAAE,KAAK,CAAC;YACZ,IAAI,EAAE,MAAM,CAAA;YACZ,CAAC,EAAE,MAAM,CAAA;YACT,CAAC,EAAE,MAAM,CAAA;YACT,IAAI,CAAC,EAAE,SAAS,CAAA;YAChB,IAAI,EAAE;gBAAE,IAAI,EAAE,MAAM,CAAC;gBAAC,IAAI,CAAC,EAAE,MAAM,CAAC;gBAAC,MAAM,CAAC,EAAE,MAAM,CAAA;aAAE,CAAA;SACvD,CAAC,CAAA;KACH,CAAC,CAAA;CACH;AAED;;;;;GAKG;AACH,UAAU,aAAa;IACrB,OAAO,EAAE,MAAM,CAAA;IACf,IAAI,EAAE,MAAM,CAAA;IACZ,aAAa,EAAE,oBAAoB,EAAE,CAAA;IACrC,SAAS,EAAE,SAAS,CAAA;CACrB;AAED;;;;;;;;;GASG;AACH,UAAU,YAAY;IACpB,KAAK,EAAE,aAAa,EAAE,CAAA;IACtB,aAAa,EAAE,MAAM,GAAG,SAAS,CAAA;IACjC,aAAa,EAAE;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,CAAA;KAAE,GAAG,SAAS,CAAA;CAC9D;AAED;;;;;;;;;;;;;;;;;;GAkBG;AACH,wBAAsB,eAAe,CACnC,GAAG,EAAE,aAAa,EAClB,QAAQ,EAAE,iBAAiB,EAC3B,YAAY,EAAE,MAAM,GACnB,OAAO,CAAC,YAAY,CAAC,CAoFvB"}