@ssmeo/mcp-local-rag 0.18.6

package/dist/parser/index.d.ts

@@ -0,0 +1,125 @@
+import type { Document as MupdfDocument } from 'mupdf';
+import { type AtomicTextRange } from '../chunker/index.js';
+import { AppError } from '../utils/errors.js';
+import type { EmbedderInterface, FilteredTextFragment } from './pdf-filter.js';
+/**
+ * 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 declare const SUPPORTED_EXTENSIONS: Set<string>;
+/**
+ * Result from parsing a document, containing both content and extracted title.
+ * Title is display-only metadata (NOT used for search scoring).
+ */
+export interface ParseResult {
+    content: string;
+    title: string;
+    atomicRanges?: readonly AtomicTextRange[];
+    imageAnchors?: readonly ParsedImageAnchor[];
+}
+export interface ParsedImageAnchor {
+    offset: number;
+    imageIndex: number;
+    mimeType: 'image/png' | 'image/jpeg';
+    bytes: Uint8Array;
+}
+export interface ParseFileOptions {
+    images?: boolean;
+}
+/**
+ * DocumentParser configuration. Exactly one of `baseDir` (legacy single root)
+ * or `baseDirs` must be supplied; the constructor rejects both, so a
+ * misconfiguration cannot silently pick one.
+ */
+export type ParserConfig = {
+    /** Security: single allowed base directory (legacy shape). */
+    baseDir: string;
+    baseDirs?: undefined;
+    /** Maximum file size (100MB). */
+    maxFileSize: number;
+} | {
+    /** Security: one or more allowed base directories (multi-root shape). */
+    baseDirs: readonly string[];
+    baseDir?: undefined;
+    /** Maximum file size (100MB). */
+    maxFileSize: number;
+};
+/**
+ * Validation error (equivalent to 400)
+ */
+export declare class ValidationError extends AppError {
+    constructor(message: string, options?: {
+        cause?: Error;
+    });
+}
+/**
+ * File operation error (equivalent to 500)
+ */
+export declare class FileOperationError extends AppError {
+    constructor(message: string, options?: {
+        cause?: Error;
+    });
+}
+/** Path and size validation plus PDF/DOCX/TXT/MD parsing. */
+export declare class DocumentParser {
+    private readonly config;
+    /** Raw allowed roots in input order (pre-realpath). Always non-empty. */
+    private readonly rawBaseDirs;
+    /**
+     * Realpath-normalized allowed roots, each with a trailing separator so the
+     * `startsWith` check cannot match a sibling (`/foo/bar/` vs `/foo/barista/`).
+     * Cached for the process lifetime.
+     */
+    private resolvedBaseDirs;
+    constructor(config: ParserConfig);
+    /**
+     * THE place realpath is used: following symlinks here is what makes prefix
+     * containment unforgeable. Everything else stores and looks up resolve()
+     * paths — see {@link BaseDirsConfig}.
+     *
+     * A file is accepted iff its realpath — or, for a not-yet-existing
+     * non-symlink, its resolve()d path — sits under any allowed root. A broken
+     * symlink is rejected outright.
+     */
+    validateFilePath(filePath: string): Promise<void>;
+    /** @throws ValidationError when the file exceeds the configured size limit. */
+    validateFileSize(filePath: string): void;
+    /** Parse a file, detecting the format from its extension. */
+    parseFile(filePath: string, options?: ParseFileOptions): Promise<ParseResult>;
+    /**
+     * PDF parsing. Headers and footers are detected semantically by embedding
+     * similarity across pages, not by position alone, and the title comes from
+     * PDF metadata with a first-page largest-font fallback.
+     */
+    parsePdf(filePath: string, embedder: EmbedderInterface): Promise<ParseResult>;
+    /**
+     * Per-page PDF parsing for the visual path. Adds `preserve-images` so mupdf
+     * emits the image blocks the visual detector needs, and returns the open
+     * `Document` so the renderer can work on the same handle.
+     *
+     * Disposal is asymmetric: on success the CALLER owns `doc` and must
+     * `finally { doc.destroy() }`; on throw `doc` is already destroyed here, so
+     * the caller must NOT destroy it.
+     */
+    parsePdfPages(filePath: string, embedder: EmbedderInterface): Promise<{
+        doc: MupdfDocument;
+        title: string;
+        pages: Array<{
+            pageNum: number;
+            text: string;
+            textFragments: FilteredTextFragment[];
+            stextJson: unknown;
+        }>;
+    }>;
+    /**
+     * DOCX parsing. One Mammoth HTML conversion serves both title and body, so
+     * the document is not converted twice.
+     */
+    private parseDocx;
+    /** TXT parsing. */
+    private parseTxt;
+    /** MD parsing. */
+    private parseMd;
+}
+//# sourceMappingURL=index.d.ts.map

package/dist/parser/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/parser/index.ts"],"names":[],"mappings":"AAOA,OAAO,KAAK,EAAE,QAAQ,IAAI,aAAa,EAAE,MAAM,OAAO,CAAA;AACtD,OAAO,EAAE,KAAK,eAAe,EAAmB,MAAM,qBAAqB,CAAA;AAE3E,OAAO,EAAE,QAAQ,EAAuB,MAAM,oBAAoB,CAAA;AAIlE,OAAO,KAAK,EAAE,iBAAiB,EAAE,oBAAoB,EAAE,MAAM,iBAAiB,CAAA;AAY9E;;;;GAIG;AACH,eAAO,MAAM,oBAAoB,aAA4C,CAAA;AAM7E;;;GAGG;AACH,MAAM,WAAW,WAAW;IAC1B,OAAO,EAAE,MAAM,CAAA;IACf,KAAK,EAAE,MAAM,CAAA;IACb,YAAY,CAAC,EAAE,SAAS,eAAe,EAAE,CAAA;IACzC,YAAY,CAAC,EAAE,SAAS,iBAAiB,EAAE,CAAA;CAC5C;AAED,MAAM,WAAW,iBAAiB;IAChC,MAAM,EAAE,MAAM,CAAA;IACd,UAAU,EAAE,MAAM,CAAA;IAClB,QAAQ,EAAE,WAAW,GAAG,YAAY,CAAA;IACpC,KAAK,EAAE,UAAU,CAAA;CAClB;AAED,MAAM,WAAW,gBAAgB;IAC/B,MAAM,CAAC,EAAE,OAAO,CAAA;CACjB;AAgCD;;;;GAIG;AACH,MAAM,MAAM,YAAY,GACpB;IACE,8DAA8D;IAC9D,OAAO,EAAE,MAAM,CAAA;IACf,QAAQ,CAAC,EAAE,SAAS,CAAA;IACpB,iCAAiC;IACjC,WAAW,EAAE,MAAM,CAAA;CACpB,GACD;IACE,yEAAyE;IACzE,QAAQ,EAAE,SAAS,MAAM,EAAE,CAAA;IAC3B,OAAO,CAAC,EAAE,SAAS,CAAA;IACnB,iCAAiC;IACjC,WAAW,EAAE,MAAM,CAAA;CACpB,CAAA;AAEL;;GAEG;AACH,qBAAa,eAAgB,SAAQ,QAAQ;IAC3C,YAAY,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE;QAAE,KAAK,CAAC,EAAE,KAAK,CAAA;KAAE,EAGvD;CACF;AAED;;GAEG;AACH,qBAAa,kBAAmB,SAAQ,QAAQ;IAC9C,YAAY,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE;QAAE,KAAK,CAAC,EAAE,KAAK,CAAA;KAAE,EAGvD;CACF;AAMD,6DAA6D;AAC7D,qBAAa,cAAc;IACzB,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAc;IACrC,yEAAyE;IACzE,OAAO,CAAC,QAAQ,CAAC,WAAW,CAAmB;IAC/C;;;;OAIG;IACH,OAAO,CAAC,gBAAgB,CAAwB;IAEhD,YAAY,MAAM,EAAE,YAAY,EAW/B;IAED;;;;;;;;OAQG;IACG,gBAAgB,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC,CAkEtD;IAED,+EAA+E;IAC/E,gBAAgB,CAAC,QAAQ,EAAE,MAAM,GAAG,IAAI,CAoBvC;IAED,6DAA6D;IACvD,SAAS,CAAC,QAAQ,EAAE,MAAM,EAAE,OAAO,GAAE,gBAAqB,GAAG,OAAO,CAAC,WAAW,CAAC,CAiBtF;IAED;;;;OAIG;IACG,QAAQ,CAAC,QAAQ,EAAE,MAAM,EAAE,QAAQ,EAAE,iBAAiB,GAAG,OAAO,CAAC,WAAW,CAAC,CAgDlF;IAED;;;;;;;;OAQG;IACG,aAAa,CACjB,QAAQ,EAAE,MAAM,EAChB,QAAQ,EAAE,iBAAiB,GAC1B,OAAO,CAAC;QACT,GAAG,EAAE,aAAa,CAAA;QAClB,KAAK,EAAE,MAAM,CAAA;QACb,KAAK,EAAE,KAAK,CAAC;YACX,OAAO,EAAE,MAAM,CAAA;YACf,IAAI,EAAE,MAAM,CAAA;YACZ,aAAa,EAAE,oBAAoB,EAAE,CAAA;YACrC,SAAS,EAAE,OAAO,CAAA;SACnB,CAAC,CAAA;KACH,CAAC,CAqDD;IAED;;;OAGG;YACW,SAAS;IA4DvB,mBAAmB;YACL,QAAQ;IAYtB,kBAAkB;YACJ,OAAO;CAWtB"}

package/dist/parser/index.js

@@ -0,0 +1,364 @@
+// 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, toError } from '../utils/errors.js';
+import { errorCode } from '../utils/type-guards.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, hints, embedder) {
+    const { metadataTitle, page1FontHint } = hints;
+    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, options) {
+        super(message, 'parser', 'validation', options);
+        this.name = 'ValidationError';
+    }
+}
+/**
+ * File operation error (equivalent to 500)
+ */
+export class FileOperationError extends AppError {
+    constructor(message, options) {
+        super(message, 'parser', 'io', options);
+        this.name = 'FileOperationError';
+    }
+}
+// ============================================
+// DocumentParser Class
+// ============================================
+/** Path and size validation plus PDF/DOCX/TXT/MD parsing. */
+export class DocumentParser {
+    config;
+    /** Raw allowed roots in input order (pre-realpath). Always non-empty. */
+    rawBaseDirs;
+    /**
+     * Realpath-normalized allowed roots, each with a trailing separator so the
+     * `startsWith` check cannot match a sibling (`/foo/bar/` vs `/foo/barista/`).
+     * Cached for the process lifetime.
+     */
+    resolvedBaseDirs = null;
+    constructor(config) {
+        this.config = config;
+        // `baseDirs` wins so a future relaxation of the type cannot fall back to
+        // the legacy field. An empty list is accepted only so the parser is
+        // constructible in the server's degraded mode; `validateFilePath` then
+        // fails closed, accepting nothing.
+        if (config.baseDirs !== undefined) {
+            this.rawBaseDirs = config.baseDirs;
+        }
+        else {
+            this.rawBaseDirs = [config.baseDir];
+        }
+    }
+    /**
+     * THE place realpath is used: following symlinks here is what makes prefix
+     * containment unforgeable. Everything else stores and looks up resolve()
+     * paths — see {@link BaseDirsConfig}.
+     *
+     * A file is accepted iff its realpath — or, for a not-yet-existing
+     * non-symlink, its resolve()d path — sits under any allowed root. A broken
+     * symlink is rejected outright.
+     */
+    async validateFilePath(filePath) {
+        // Fail closed in degraded mode: an empty allow-list must reject every
+        // path, not run a prefix check against no roots. `assertConfigOk` should
+        // fire first; this covers paths that bypass it.
+        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.`, { cause: toError(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}`);
+        }
+    }
+    /** @throws ValidationError when the file exceeds the configured size limit. */
+    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 (errorCode(error) === 'ENOENT') {
+                throw new ValidationError(`File not found: ${filePath}`);
+            }
+            throw new FileOperationError(`Failed to check file size: ${filePath}`, {
+                cause: toError(error),
+            });
+        }
+    }
+    /** Parse a file, detecting the format from its extension. */
+    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. Headers and footers are detected semantically by embedding
+     * similarity across pages, not by position alone, and the title comes from
+     * PDF metadata with a first-page largest-font fallback.
+     */
+    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 (an `EmbeddingError` raised while the parser
+            // used the embedder) keeps its identity rather than being relabelled a
+            // PDF failure. Only a genuine non-`AppError` wraps as `FileOperationError`.
+            if (isAppError(error)) {
+                throw error;
+            }
+            throw new FileOperationError(`Failed to parse PDF: ${filePath}`, { cause: toError(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 path. Adds `preserve-images` so mupdf
+     * emits the image blocks the visual detector needs, and returns the open
+     * `Document` so the renderer can work on the same handle.
+     *
+     * Disposal is asymmetric: on success the CALLER owns `doc` and must
+     * `finally { doc.destroy() }`; on throw `doc` is already destroyed here, so
+     * the caller must NOT destroy it.
+     */
+    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: ${filePath}`, {
+                cause: toError(error),
+            });
+        }
+    }
+    /**
+     * DOCX parsing. One Mammoth HTML conversion serves both title and body, so
+     * the document is not converted twice.
+     */
+    async parseDocx(filePath, includeImages) {
+        try {
+            const buffer = await readFile(filePath);
+            const capturedImages = new Map();
+            let nextImageIndex = 0;
+            let skippedImageCount = 0;
+            const htmlResult = includeImages
+                ? await mammoth.convertToHtml({ buffer }, {
+                    convertImage: mammoth.images.imgElement(async (image) => {
+                        const imageIndex = nextImageIndex++;
+                        if (image.contentType !== 'image/png' && image.contentType !== 'image/jpeg') {
+                            skippedImageCount += 1;
+                            return { src: `data:${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}`, { cause: toError(error) });
+        }
+    }
+    /** TXT parsing. */
+    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}`, { cause: toError(error) });
+        }
+    }
+    /** MD parsing. */
+    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}`, { cause: toError(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,OAAO,EAAE,MAAM,oBAAoB,CAAA;AAClE,OAAO,EAAE,SAAS,EAAE,MAAM,yBAAyB,CAAA;AACnD,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;AAkC7E,KAAK,UAAU,eAAe,CAC5B,QAAgB,EAChB,KAAkC,EAClC,KAAoB,EACpB,QAA2B;IAE3B,MAAM,EAAE,aAAa,EAAE,aAAa,EAAE,GAAG,KAAK,CAAA;IAC9C,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,EAAE,CAAC;YAC3B,MAAM,UAAU,CAAA;QAClB,CAAC;QACD,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;AAuBD;;GAEG;AACH,MAAM,OAAO,eAAgB,SAAQ,QAAQ;IAC3C,YAAY,OAAe,EAAE,OAA2B;QACtD,KAAK,CAAC,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,OAAO,CAAC,CAAA;QAC/C,IAAI,CAAC,IAAI,GAAG,iBAAiB,CAAA;IAC/B,CAAC;CACF;AAED;;GAEG;AACH,MAAM,OAAO,kBAAmB,SAAQ,QAAQ;IAC9C,YAAY,OAAe,EAAE,OAA2B;QACtD,KAAK,CAAC,OAAO,EAAE,QAAQ,EAAE,IAAI,EAAE,OAAO,CAAC,CAAA;QACvC,IAAI,CAAC,IAAI,GAAG,oBAAoB,CAAA;IAClC,CAAC;CACF;AAED,+CAA+C;AAC/C,uBAAuB;AACvB,+CAA+C;AAE/C,6DAA6D;AAC7D,MAAM,OAAO,cAAc;IACR,MAAM,CAAc;IACrC,yEAAyE;IACxD,WAAW,CAAmB;IAC/C;;;;OAIG;IACK,gBAAgB,GAAoB,IAAI,CAAA;IAEhD,YAAY,MAAoB;QAC9B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAA;QACpB,yEAAyE;QACzE,oEAAoE;QACpE,uEAAuE;QACvE,mCAAmC;QACnC,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;;;;;;;;OAQG;IACH,KAAK,CAAC,gBAAgB,CAAC,QAAgB;QACrC,sEAAsE;QACtE,yEAAyE;QACzE,gDAAgD;QAChD,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,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,EAAE,CAC1B,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,+EAA+E;IAC/E,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,IAAI,SAAS,CAAC,KAAK,CAAC,KAAK,QAAQ,EAAE,CAAC;gBAClC,MAAM,IAAI,eAAe,CAAC,mBAAmB,QAAQ,EAAE,CAAC,CAAA;YAC1D,CAAC;YACD,MAAM,IAAI,kBAAkB,CAAC,8BAA8B,QAAQ,EAAE,EAAE;gBACrE,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC;aACtB,CAAC,CAAA;QACJ,CAAC;IACH,CAAC;IAED,6DAA6D;IAC7D,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;;;;OAIG;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,CAAC,CAAA;YAE5D,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,CACjC,QAAQ,EACR,KAAK,EACL,EAAE,aAAa,EAAE,aAAa,EAAE,EAChC,QAAQ,CACT,CAAA;YAED,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,sEAAsE;YACtE,uEAAuE;YACvE,4EAA4E;YAC5E,IAAI,UAAU,CAAC,KAAK,CAAC,EAAE,CAAC;gBACtB,MAAM,KAAK,CAAA;YACb,CAAC;YACD,MAAM,IAAI,kBAAkB,CAAC,wBAAwB,QAAQ,EAAE,EAAE,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC,CAAA;QAC7F,CAAC;gBAAS,CAAC;YACT,sEAAsE;YACtE,2BAA2B;YAC3B,GAAG,EAAE,OAAO,EAAE,CAAA;QAChB,CAAC;IACH,CAAC;IAED;;;;;;;;OAQG;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,CAAC,CAAA;YAC5D,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,EAAE,aAAa,EAAE,aAAa,EAAE,EAChC,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;gBACrE,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC;aACtB,CAAC,CAAA;QACJ,CAAC;IACH,CAAC;IAED;;;OAGG;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,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC,CAAA;QAC9F,CAAC;IACH,CAAC;IAED,mBAAmB;IACX,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,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC,CAAA;QAC7F,CAAC;IACH,CAAC;IAED,kBAAkB;IACV,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,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC,CAAA;QAC5F,CAAC;IACH,CAAC;CACF"}

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

@@ -0,0 +1,49 @@
+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;
+}
+/** The part of mupdf's structured-text JSON this module reads. */
+interface StextJson {
+    blocks: Array<{
+        type: string;
+        /** Present on image blocks; the visual detector reads it to locate rasters. */
+        bbox?: StextBbox;
+        lines?: Array<{
+            text: string;
+            x: number;
+            y: number;
+            bbox?: StextBbox;
+            font: {
+                size: number;
+                name?: string;
+                weight?: string;
+            };
+        }>;
+    }>;
+}
+/**
+ * `text` is post-header/footer-filtering; `stextJson` is the raw mupdf JSON,
+ * kept for the downstream visual detector.
+ */
+interface ExtractedPage {
+    pageNum: number;
+    text: string;
+    textFragments: FilteredTextFragment[];
+    stextJson: StextJson;
+}
+/** Result returned by `extractPdfPages`. */
+interface ExtractedPdf {
+    pages: ExtractedPage[];
+    metadataTitle: string | undefined;
+    page1FontHint: {
+        text: string;
+        fontSize: number;
+    } | undefined;
+}
+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,EAAqB,MAAM,OAAO,CAAA;AAEzE,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,kEAAkE;AAClE,UAAU,SAAS;IACjB,MAAM,EAAE,KAAK,CAAC;QACZ,IAAI,EAAE,MAAM,CAAA;QACZ,+EAA+E;QAC/E,IAAI,CAAC,EAAE,SAAS,CAAA;QAChB,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;;;GAGG;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,4CAA4C;AAC5C,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;AAwFD,wBAAsB,eAAe,CACnC,GAAG,EAAE,aAAa,EAClB,QAAQ,EAAE,iBAAiB,EAC3B,YAAY,EAAE,MAAM,GACnB,OAAO,CAAC,YAAY,CAAC,CAgCvB"}