@okrapdf/cli 0.3.7 → 0.4.0

package/dist/commands/elements.js

@@ -0,0 +1,857 @@
+/**
+ * Consolidated element management commands
+ *
+ * Combines entities, tables, toc, and review into a single top-level command.
+ * All extracted content from PDFs (tables, figures, footnotes, etc.) lives here.
+ */
+import { Command } from 'commander';
+import { writeFileSync, mkdirSync, existsSync, readFileSync } from 'fs';
+import { join, resolve } from 'path';
+import chalk from 'chalk';
+import { get, post, patch, OkraApiError, EXIT_CODES } from '../lib/client.js';
+import { formatOutput, formatDate, formatStatus as formatOutputStatus, success, error, info, warn, } from '../lib/output.js';
+import { withSpinner } from '../lib/progress.js';
+import { getDefaultFormat, shouldUseJsonOutput } from '../lib/config.js';
+import { openInBrowser, getJobWebUrl } from '../lib/browser.js';
+import { CacheManager } from '../lib/cache.js';
+import { validateJobReady, throwValidationError } from '../lib/validator.js';
+// ─── Table columns ───────────────────────────────────────────────────────────
+const ENTITY_COLUMNS = [
+    { key: 'id', header: 'Entity ID', width: 25 },
+    { key: 'type', header: 'Type', width: 12 },
+    { key: 'page', header: 'Page', width: 6 },
+    { key: 'title_fmt', header: 'Title', width: 40 },
+    { key: 'has_bbox', header: 'BBox', width: 6 },
+];
+const REVIEW_PAGE_COLUMNS = [
+    { key: 'page', header: 'Page', width: 6 },
+    { key: 'status', header: 'Status', width: 12 },
+    { key: 'resolution', header: 'Resolution', width: 12 },
+    { key: 'ocrLineCount', header: 'OCR Lines', width: 10 },
+    { key: 'hasOcr', header: 'Has OCR', width: 8 },
+];
+// ─── Helpers ─────────────────────────────────────────────────────────────────
+function truncate(str, maxLength) {
+    if (str.length <= maxLength)
+        return str;
+    return str.slice(0, maxLength - 3) + '...';
+}
+function formatReviewStatus(status) {
+    switch (status) {
+        case 'complete': return chalk.green(status);
+        case 'partial': return chalk.yellow(status);
+        case 'flagged': return chalk.red(status);
+        case 'pending': return chalk.yellow(status);
+        case 'gap': return chalk.magenta(status);
+        case 'empty': return chalk.dim(status);
+        default: return status;
+    }
+}
+function markdownTableToCsv(markdown) {
+    const lines = markdown.trim().split('\n');
+    const csvLines = [];
+    for (const line of lines) {
+        if (line.match(/^\|[\s-:|]+\|$/))
+            continue;
+        if (line.startsWith('|') && line.endsWith('|')) {
+            const cells = line
+                .slice(1, -1)
+                .split('|')
+                .map(cell => cell.trim());
+            const escapedCells = cells.map(cell => {
+                if (cell.includes(',') || cell.includes('"') || cell.includes('\n')) {
+                    return `"${cell.replace(/"/g, '""')}"`;
+                }
+                return cell;
+            });
+            csvLines.push(escapedCells.join(','));
+        }
+    }
+    return csvLines.join('\n');
+}
+function formatTocAsMarkdown(toc) {
+    const lines = [];
+    lines.push(chalk.bold('Table of Contents'));
+    lines.push('');
+    for (const entry of toc.toc) {
+        const indent = '  '.repeat(entry.level - 1);
+        const pageNum = chalk.dim(`(p. ${entry.page})`);
+        lines.push(`${indent}- ${entry.title} ${pageNum}`);
+    }
+    lines.push('');
+    lines.push(chalk.dim(`Total entries: ${toc.total_entries}`));
+    return lines.join('\n');
+}
+// ─── Main command ────────────────────────────────────────────────────────────
+export function createElementsCommand() {
+    const elements = new Command('elements')
+        .description('Manage extracted elements (tables, figures, footnotes, TOC, review)');
+    // ── elements list ──────────────────────────────────────────────────────────
+    elements
+        .command('list <jobId>')
+        .alias('ls')
+        .description('List elements from a job')
+        .option('-o, --output <format>', 'Output format (table, json, csv)', getDefaultFormat())
+        .option('-t, --type <type>', 'Filter by type (tables, figures, footnotes, summaries, signatures, all)', 'all')
+        .option('-p, --page <n>', 'Filter by page number')
+        .option('--with-bbox', 'Only show elements with bounding boxes')
+        .action(async (jobId, options) => {
+        const params = {
+            type: options.type,
+        };
+        const response = await withSpinner('Fetching elements', () => get(`api/ocr/jobs/${jobId}/entities`, params));
+        let entities = response.entities;
+        if (options.page) {
+            const pageNum = parseInt(options.page, 10);
+            entities = entities.filter(e => e.page === pageNum);
+        }
+        if (options.withBbox) {
+            entities = entities.filter(e => e.bbox != null);
+        }
+        if (entities.length === 0) {
+            console.log(chalk.dim('No elements found'));
+            return;
+        }
+        const formatted = entities.map(entity => ({
+            ...entity,
+            title_fmt: truncate(entity.title || '-', 40),
+            has_bbox: entity.bbox ? chalk.green('Yes') : chalk.dim('No'),
+        }));
+        console.log(formatOutput(formatted, options.output, ENTITY_COLUMNS));
+        if (!shouldUseJsonOutput(options.output)) {
+            console.log(chalk.dim(`\nTotal: ${entities.length} elements`));
+            console.log(chalk.dim(`Status: ${formatOutputStatus(response.extractionStatus)}`));
+        }
+    });
+    // ── elements get ───────────────────────────────────────────────────────────
+    elements
+        .command('get <elementId>')
+        .description('Get a single table/element by ID')
+        .option('-o, --output <format>', 'Output format (markdown, json)', 'markdown')
+        .action(async (elementId, options) => {
+        try {
+            const table = await withSpinner('Fetching element', () => get(`api/extractions/tables/${elementId}`));
+            if (options.output === 'json') {
+                console.log(formatOutput(table, 'json'));
+            }
+            else {
+                console.log(chalk.bold(`Table (Page ${table.page_number})`));
+                console.log(chalk.dim('-'.repeat(50)));
+                console.log(table.content_markdown);
+                console.log();
+                console.log(chalk.dim(`ID: ${table.id}`));
+                console.log(chalk.dim(`Source: ${table.processor_type}`));
+                if (table.confidence !== null) {
+                    console.log(chalk.dim(`Confidence: ${(table.confidence * 100).toFixed(1)}%`));
+                }
+            }
+        }
+        catch (err) {
+            if (err instanceof OkraApiError && err.statusCode === 404) {
+                error(`Element not found: ${elementId}`);
+                process.exit(EXIT_CODES.NOT_FOUND);
+            }
+            throw err;
+        }
+    });
+    // ── elements export ────────────────────────────────────────────────────────
+    elements
+        .command('export <jobId>')
+        .description('Export elements to disk (tables as CSV/JSON/MD, images as PNG/JPG)')
+        .option('-t, --type <type>', 'Element type to export (tables, figures, all)', 'all')
+        .option('-f, --format <fmt>', 'Export format (csv, json, md, png, jpg)', 'csv')
+        .option('-d, --output-dir <dir>', 'Output directory', './element-export')
+        .option('-q, --quality <n>', 'Image quality for JPEG (1-100)', '90')
+        .option('-s, --scale <n>', 'Image scale factor (1-4)', '2')
+        .option('--padding <n>', 'Padding around crop region in pixels', '10')
+        .option('-o, --output <format>', 'Result output format (table, json)', 'table')
+        .action(async (jobId, options) => {
+        const useJson = shouldUseJsonOutput(options.output);
+        const format = options.format.toLowerCase();
+        const isImageFormat = format === 'png' || format === 'jpg';
+        if (isImageFormat) {
+            // Image export path (from entities images)
+            await exportImages(jobId, { ...options, format }, useJson);
+        }
+        else {
+            // Table data export path (from tables export)
+            await exportTableData(jobId, { ...options, format }, useJson);
+        }
+    });
+    // ── elements toc ───────────────────────────────────────────────────────────
+    elements
+        .command('toc <jobId>')
+        .description('Extract table of contents from PDF')
+        .option('-f, --format <format>', 'Output format (markdown, json)', 'markdown')
+        .option('--max-depth <n>', 'Maximum TOC depth level', parseInt)
+        .option('--refresh', 'Force refresh from API (ignore cache)')
+        .action(async (jobId, options) => {
+        try {
+            const cache = new CacheManager();
+            const validation = await validateJobReady(jobId, cache);
+            if (!validation.valid) {
+                throwValidationError(validation.error);
+            }
+            if (!options.refresh && cache.getToc(jobId)) {
+                const cachedToc = cache.getToc(jobId);
+                if (cachedToc) {
+                    info('Using cached TOC data');
+                    if (options.format === 'json') {
+                        console.log(JSON.stringify(cachedToc, null, 2));
+                    }
+                    else {
+                        console.log(formatTocAsMarkdown(cachedToc));
+                    }
+                    return;
+                }
+            }
+            const params = {};
+            if (options.maxDepth) {
+                params.max_depth = options.maxDepth.toString();
+            }
+            info('Extracting TOC from PDF...');
+            const apiResponse = await withSpinner('Running TOC extraction (this may take ~30s)', () => get(`api/steps/table-of-content/${jobId}`, params));
+            if (!apiResponse.success) {
+                error('TOC extraction failed');
+                process.exit(EXIT_CODES.GENERAL_ERROR);
+            }
+            const tocResult = {
+                toc: apiResponse.toc.map(entry => ({
+                    level: entry.level,
+                    title: entry.title,
+                    page: entry.page,
+                })),
+                total_entries: apiResponse.total_entries,
+            };
+            cache.setToc(jobId, tocResult);
+            success('TOC cached locally');
+            cache.logCommand('toc', jobId, {
+                format: options.format,
+                maxDepth: options.maxDepth,
+                refresh: options.refresh,
+            });
+            cache.logResult(true, {
+                total_entries: tocResult.total_entries,
+                strategy: apiResponse.strategy,
+            });
+            if (options.format === 'json') {
+                console.log(JSON.stringify(tocResult, null, 2));
+            }
+            else {
+                console.log(formatTocAsMarkdown(tocResult));
+            }
+            console.log(chalk.dim(`\nExtraction: ${apiResponse.strategy} strategy, ${apiResponse.total_elapsed_ms}ms`));
+        }
+        catch (err) {
+            const cache = new CacheManager();
+            const errorMessage = err instanceof Error ? err.message : String(err);
+            cache.logResult(false, {}, errorMessage);
+            if (err instanceof OkraApiError) {
+                error(err.message);
+                process.exit(err.exitCode);
+            }
+            error(`TOC extraction failed: ${errorMessage}`);
+            process.exit(EXIT_CODES.GENERAL_ERROR);
+        }
+    });
+    // ── elements count ─────────────────────────────────────────────────────────
+    elements
+        .command('count <jobId>')
+        .description('Get element counts by type')
+        .option('-o, --output <format>', 'Output format (table, json)', getDefaultFormat())
+        .action(async (jobId, options) => {
+        const response = await withSpinner('Fetching element counts', () => get(`api/ocr/jobs/${jobId}/entities`));
+        if (options.output === 'json') {
+            console.log(formatOutput({
+                job_id: jobId,
+                counts: response.counts,
+                extraction_status: response.extractionStatus,
+                total_pages: response.totalPages,
+            }, 'json'));
+        }
+        else {
+            console.log(chalk.bold('\nElement Counts'));
+            console.log(chalk.dim('-'.repeat(30)));
+            console.log(chalk.bold('Tables:'), response.counts.tables);
+            console.log(chalk.bold('Figures:'), response.counts.figures);
+            console.log(chalk.bold('Footnotes:'), response.counts.footnotes);
+            console.log(chalk.bold('Summaries:'), response.counts.summaries);
+            console.log(chalk.bold('Signatures:'), response.counts.signatures);
+            console.log();
+            console.log(chalk.dim(`Status: ${response.extractionStatus}`));
+            if (response.totalPages) {
+                console.log(chalk.dim(`Total pages: ${response.totalPages}`));
+            }
+        }
+    });
+    // ── elements review ────────────────────────────────────────────────────────
+    const review = elements
+        .command('review')
+        .description('Review job verification status and page content');
+    // review status
+    review
+        .command('status <jobId>')
+        .description('Get verification status summary for a job')
+        .option('-o, --output <format>', 'Output format (table, json)', getDefaultFormat())
+        .option('-w, --web', 'Open job review page in browser')
+        .action(async (jobId, options) => {
+        if (options.web) {
+            const url = `${getJobWebUrl(jobId)}/review`;
+            console.error(`Opening ${url} in your browser.`);
+            await openInBrowser(url);
+            return;
+        }
+        const tree = await withSpinner('Fetching verification status', () => get(`api/ocr/jobs/${jobId}/verification-tree`));
+        if (options.output === 'json') {
+            console.log(formatOutput(tree, 'json'));
+            return;
+        }
+        console.log(chalk.bold('Verification Status'));
+        console.log(chalk.dim('-'.repeat(50)));
+        console.log(chalk.bold('Job:'), jobId);
+        console.log(chalk.bold('Total Pages:'), tree.totalPages);
+        console.log();
+        console.log(chalk.bold('Summary:'));
+        console.log(`  ${chalk.green('Complete:')} ${tree.summary.complete}`);
+        console.log(`  ${chalk.yellow('Pending:')} ${tree.summary.pending}`);
+        console.log(`  ${chalk.red('Flagged:')} ${tree.summary.flagged}`);
+        console.log(`  ${chalk.magenta('Gap:')} ${tree.summary.gap}`);
+        console.log(`  ${chalk.blue('Resolved:')} ${tree.summary.resolved}`);
+        if (tree.summary.stale > 0) {
+            console.log(`  ${chalk.dim('Stale:')} ${tree.summary.stale}`);
+        }
+    });
+    // review pages
+    review
+        .command('pages <jobId>')
+        .description('List pages with verification status')
+        .option('-o, --output <format>', 'Output format (table, json)', getDefaultFormat())
+        .option('-s, --status <status>', 'Filter by status (complete, pending, flagged, gap)')
+        .action(async (jobId, options) => {
+        const tree = await withSpinner('Fetching pages', () => get(`api/ocr/jobs/${jobId}/verification-tree`));
+        let pages = tree.pages;
+        if (options.status) {
+            pages = pages.filter(p => p.status === options.status);
+        }
+        if (pages.length === 0) {
+            console.log(chalk.dim('No pages found'));
+            return;
+        }
+        if (options.output === 'json') {
+            console.log(formatOutput(pages, 'json'));
+            return;
+        }
+        const formatted = pages.map(p => ({
+            ...p,
+            status: formatReviewStatus(p.status),
+            resolution: p.resolution || chalk.dim('-'),
+            hasOcr: p.hasOcr ? chalk.green('\u2713') : chalk.dim('\u2717'),
+        }));
+        console.log(formatOutput(formatted, 'table', REVIEW_PAGE_COLUMNS));
+        console.log(chalk.dim(`\n${pages.length} pages`));
+    });
+    // review page
+    review
+        .command('page <jobId> <pageNum>')
+        .description('Get page content (markdown and OCR blocks)')
+        .option('-o, --output <format>', 'Output format (markdown, json)', 'markdown')
+        .option('--ocr', 'Show OCR blocks instead of markdown')
+        .option('--raw', 'Output raw content without formatting')
+        .action(async (jobId, pageNum, options) => {
+        const page = await withSpinner('Fetching page content', () => get(`api/ocr/jobs/${jobId}/pages/${pageNum}`));
+        if (options.output === 'json') {
+            console.log(formatOutput(page, 'json'));
+            return;
+        }
+        if (options.ocr) {
+            if (!page.blocks || page.blocks.length === 0) {
+                console.log(chalk.dim('No OCR blocks available'));
+                return;
+            }
+            if (options.raw) {
+                for (const block of page.blocks) {
+                    console.log(block.text);
+                }
+                return;
+            }
+            console.log(chalk.bold(`OCR Blocks - Page ${pageNum}`));
+            console.log(chalk.dim('-'.repeat(50)));
+            for (let i = 0; i < page.blocks.length; i++) {
+                const block = page.blocks[i];
+                const conf = block.confidence !== undefined ? ` (${(block.confidence * 100).toFixed(0)}%)` : '';
+                console.log(chalk.cyan(`[${i + 1}]${conf}`), block.text);
+            }
+            console.log(chalk.dim(`\n${page.blocks.length} blocks`));
+            return;
+        }
+        if (options.raw) {
+            console.log(page.content);
+            return;
+        }
+        console.log(chalk.bold(`Page ${pageNum} Content`));
+        console.log(chalk.dim('-'.repeat(50)));
+        console.log(page.content);
+        console.log();
+        console.log(chalk.dim(`Version: ${page.version}`));
+        if (page.dimension) {
+            console.log(chalk.dim(`Dimension: ${page.dimension.width}x${page.dimension.height}`));
+        }
+        if (page.blocks) {
+            console.log(chalk.dim(`OCR Blocks: ${page.blocks.length}`));
+        }
+    });
+    // review resolve
+    review
+        .command('resolve <jobId> <pageNum>')
+        .description('Mark a page as reviewed')
+        .option('-r, --resolution <type>', 'Resolution type (reviewed, skipped, flagged)', 'reviewed')
+        .option('-n, --note <text>', 'Add a note to the resolution')
+        .option('-o, --output <format>', 'Output format (table, json)', 'table')
+        .action(async (jobId, pageNum, options) => {
+        const useJson = shouldUseJsonOutput(options.output);
+        try {
+            await withSpinner(`Resolving page ${pageNum}`, () => post(`api/ocr/jobs/${jobId}/pages/${pageNum}/resolve`, {
+                resolution: options.resolution,
+                note: options.note,
+            }));
+            if (useJson) {
+                console.log(formatOutput({
+                    success: true,
+                    job_id: jobId,
+                    page: parseInt(pageNum),
+                    resolution: options.resolution,
+                }, 'json'));
+            }
+            else {
+                success(`Page ${pageNum} marked as ${options.resolution}`);
+            }
+        }
+        catch (err) {
+            if (err instanceof OkraApiError) {
+                if (useJson) {
+                    console.log(formatOutput({ success: false, error: err.message }, 'json'));
+                }
+                else {
+                    error(err.message);
+                }
+                process.exit(err.exitCode);
+            }
+            throw err;
+        }
+    });
+    // review history
+    review
+        .command('history <jobId>')
+        .description('Get verification audit trail')
+        .option('-o, --output <format>', 'Output format (table, json)', getDefaultFormat())
+        .option('-l, --limit <n>', 'Limit results', '20')
+        .option('-p, --page <n>', 'Filter by page number')
+        .action(async (jobId, options) => {
+        const params = { limit: options.limit };
+        if (options.page)
+            params.page = options.page;
+        const response = await withSpinner('Fetching history', () => get(`api/ocr/jobs/${jobId}/history`, params));
+        if (response.history.length === 0) {
+            console.log(chalk.dim('No history found'));
+            return;
+        }
+        if (options.output === 'json') {
+            console.log(formatOutput(response.history, 'json'));
+            return;
+        }
+        console.log(chalk.bold('Verification History'));
+        console.log(chalk.dim('-'.repeat(60)));
+        for (const entry of response.history) {
+            const page = entry.pageNumber ? `Page ${entry.pageNumber}` : '';
+            const entity = entry.entityType ? `${entry.entityType}` : '';
+            const target = [page, entity].filter(Boolean).join(' - ') || 'Job';
+            console.log(chalk.dim(formatDate(entry.createdAt)), chalk.cyan(entry.action), chalk.white(target), chalk.dim(`by ${entry.triggeredBy}`));
+        }
+    });
+    // review save
+    review
+        .command('save <jobId> <pageNum>')
+        .description('Save/update page markdown content')
+        .option('-f, --file <path>', 'Read content from file')
+        .option('-c, --content <text>', 'Content to save (use - for stdin)')
+        .option('-o, --output <format>', 'Output format (table, json)', 'table')
+        .action(async (jobId, pageNum, options) => {
+        const useJson = shouldUseJsonOutput(options.output);
+        let content;
+        if (options.file) {
+            if (!existsSync(options.file)) {
+                if (useJson) {
+                    console.log(formatOutput({ success: false, error: `File not found: ${options.file}` }, 'json'));
+                }
+                else {
+                    error(`File not found: ${options.file}`);
+                }
+                process.exit(EXIT_CODES.INVALID_ARGS);
+            }
+            content = readFileSync(options.file, 'utf-8');
+        }
+        else if (options.content) {
+            if (options.content === '-') {
+                const chunks = [];
+                for await (const chunk of process.stdin) {
+                    chunks.push(chunk);
+                }
+                content = Buffer.concat(chunks).toString('utf-8');
+            }
+            else {
+                content = options.content;
+            }
+        }
+        else {
+            if (useJson) {
+                console.log(formatOutput({ success: false, error: 'Either --file or --content is required' }, 'json'));
+            }
+            else {
+                error('Either --file or --content is required');
+            }
+            process.exit(EXIT_CODES.INVALID_ARGS);
+        }
+        try {
+            const result = await withSpinner(`Saving page ${pageNum}`, () => patch(`api/ocr/jobs/${jobId}/pages/${pageNum}`, { content }));
+            if (useJson) {
+                console.log(formatOutput({
+                    success: true,
+                    job_id: jobId,
+                    page: parseInt(pageNum),
+                    version: result.version,
+                }, 'json'));
+            }
+            else {
+                success(`Page ${pageNum} saved (version ${result.version})`);
+            }
+        }
+        catch (err) {
+            if (err instanceof OkraApiError) {
+                if (useJson) {
+                    console.log(formatOutput({ success: false, error: err.message }, 'json'));
+                }
+                else {
+                    error(err.message);
+                }
+                process.exit(err.exitCode);
+            }
+            throw err;
+        }
+    });
+    // review versions
+    review
+        .command('versions <jobId> <pageNum>')
+        .description('List page content versions')
+        .option('-o, --output <format>', 'Output format (table, json)', getDefaultFormat())
+        .action(async (jobId, pageNum, options) => {
+        const response = await withSpinner('Fetching versions', () => get(`api/ocr/jobs/${jobId}/pages/${pageNum}/versions`));
+        if (response.versions.length === 0) {
+            console.log(chalk.dim('No versions found'));
+            return;
+        }
+        if (options.output === 'json') {
+            console.log(formatOutput(response.versions, 'json'));
+            return;
+        }
+        console.log(chalk.bold(`Page ${pageNum} Versions`));
+        console.log(chalk.dim('-'.repeat(50)));
+        for (const v of response.versions) {
+            console.log(`  v${v.version} - ${formatDate(v.createdAt)} by ${v.createdBy}`);
+        }
+    });
+    // review version
+    review
+        .command('version <jobId> <pageNum> <version>')
+        .description('Get specific version of page content')
+        .option('-o, --output <format>', 'Output format (markdown, json)', 'markdown')
+        .option('--raw', 'Output raw content without formatting')
+        .action(async (jobId, pageNum, version, options) => {
+        const page = await withSpinner(`Fetching version ${version}`, () => get(`api/ocr/jobs/${jobId}/pages/${pageNum}/versions/${version}`));
+        if (options.output === 'json') {
+            console.log(formatOutput(page, 'json'));
+            return;
+        }
+        if (options.raw) {
+            console.log(page.content);
+            return;
+        }
+        console.log(chalk.bold(`Page ${pageNum} - Version ${version}`));
+        console.log(chalk.dim('-'.repeat(50)));
+        console.log(page.content);
+    });
+    // review diff
+    review
+        .command('diff <jobId> <pageNum>')
+        .description('Show diff between current and previous version')
+        .option('--from <v>', 'Compare from version')
+        .option('--to <v>', 'Compare to version')
+        .option('-o, --output <format>', 'Output format (table, json)', 'table')
+        .action(async (jobId, pageNum, options) => {
+        const useJson = shouldUseJsonOutput(options.output);
+        const current = await get(`api/ocr/jobs/${jobId}/pages/${pageNum}`);
+        let previousVersion = (current.version || 1) - 1;
+        if (options.from)
+            previousVersion = parseInt(options.from);
+        if (previousVersion < 1) {
+            if (useJson) {
+                console.log(formatOutput({
+                    job_id: jobId,
+                    page: parseInt(pageNum),
+                    current_version: current.version,
+                    previous_version: null,
+                    message: 'No previous version to compare',
+                    changes: [],
+                }, 'json'));
+            }
+            else {
+                console.log(chalk.dim('No previous version to compare'));
+            }
+            return;
+        }
+        const previous = await get(`api/ocr/jobs/${jobId}/pages/${pageNum}/versions/${previousVersion}`);
+        const currentLines = current.content.split('\n');
+        const previousLines = previous.content.split('\n');
+        const changes = [];
+        for (let i = 0; i < Math.max(currentLines.length, previousLines.length); i++) {
+            const curr = currentLines[i] ?? '';
+            const prev = previousLines[i] ?? '';
+            if (curr !== prev) {
+                if (prev)
+                    changes.push({ line: i + 1, type: 'removed', content: prev });
+                if (curr)
+                    changes.push({ line: i + 1, type: 'added', content: curr });
+            }
+        }
+        if (useJson) {
+            console.log(formatOutput({
+                job_id: jobId,
+                page: parseInt(pageNum),
+                from_version: previousVersion,
+                to_version: current.version,
+                changes,
+            }, 'json'));
+            return;
+        }
+        console.log(chalk.bold(`Diff: v${previousVersion} \u2192 v${current.version}`));
+        console.log(chalk.dim('-'.repeat(50)));
+        for (const change of changes) {
+            if (change.type === 'removed') {
+                console.log(chalk.red(`- ${change.content}`));
+            }
+            else {
+                console.log(chalk.green(`+ ${change.content}`));
+            }
+        }
+    });
+    return elements;
+}
+// ─── Export helpers ──────────────────────────────────────────────────────────
+async function exportImages(jobId, options, useJson) {
+    const format = options.format;
+    if (format !== 'png' && format !== 'jpg') {
+        if (useJson) {
+            console.log(formatOutput({ success: false, error: 'Invalid image format. Use: png or jpg' }, 'json'));
+        }
+        else {
+            error('Invalid image format. Use: png or jpg');
+        }
+        process.exit(EXIT_CODES.INVALID_ARGS);
+    }
+    const entitiesResponse = await withSpinner('Fetching elements', () => get(`api/ocr/jobs/${jobId}/entities`, { type: options.type }));
+    let entities = entitiesResponse.entities.filter(e => e.bbox != null);
+    if (options.page) {
+        const pageNum = parseInt(options.page, 10);
+        entities = entities.filter(e => e.page === pageNum);
+    }
+    if (entities.length === 0) {
+        if (useJson) {
+            console.log(formatOutput({
+                success: true,
+                job_id: jobId,
+                message: 'No elements with bounding boxes found',
+                exported: 0,
+            }, 'json'));
+        }
+        else {
+            warn('No elements with bounding boxes found');
+        }
+        return;
+    }
+    info(`Found ${entities.length} elements with bounding boxes`);
+    const job = await withSpinner('Fetching job info', () => get(`api/v1/jobs/${jobId}`));
+    const jobData = job;
+    if (!jobData.document_uuid) {
+        if (useJson) {
+            console.log(formatOutput({ success: false, error: 'Job has no associated document' }, 'json'));
+        }
+        else {
+            error('Job has no associated document');
+        }
+        process.exit(EXIT_CODES.GENERAL_ERROR);
+    }
+    const docInfo = await withSpinner('Getting document URL', () => get(`api/documents/${jobData.document_uuid}/download`));
+    const got = (await import('got')).default;
+    info('Downloading PDF...');
+    const pdfResponse = await got(docInfo.signed_url, { responseType: 'buffer' });
+    const pdfBuffer = pdfResponse.body;
+    const { PdfImageRenderer } = await import('../lib/pdf-image.js');
+    const outDir = resolve(options.outputDir);
+    if (!existsSync(outDir)) {
+        mkdirSync(outDir, { recursive: true });
+    }
+    const renderer = await PdfImageRenderer.fromBuffer(pdfBuffer);
+    const exported = [];
+    const errors = [];
+    const renderOptions = {
+        format,
+        quality: parseInt(options.quality, 10),
+        scale: parseFloat(options.scale),
+        padding: parseInt(options.padding, 10),
+    };
+    console.log(chalk.dim(`Exporting ${entities.length} elements to ${outDir}...`));
+    for (const entity of entities) {
+        try {
+            const result = await renderer.renderRegion(entity.page, entity.bbox, renderOptions);
+            const safeTitle = (entity.title || 'untitled')
+                .toLowerCase()
+                .replace(/[^a-z0-9]+/g, '-')
+                .slice(0, 30);
+            const filename = `${entity.type}-p${entity.page}-${safeTitle}.${format}`;
+            const filepath = join(outDir, filename);
+            writeFileSync(filepath, result.buffer);
+            exported.push({
+                entity_id: entity.id,
+                file: filepath,
+                page: entity.page,
+                type: entity.type,
+            });
+            if (!useJson) {
+                console.log(chalk.green('  \u2713'), chalk.dim(`${entity.type} p${entity.page}:`), filename);
+            }
+        }
+        catch (err) {
+            const errorMsg = err instanceof Error ? err.message : 'Unknown error';
+            errors.push({ entity_id: entity.id, error: errorMsg });
+            if (!useJson) {
+                console.log(chalk.red('  \u2717'), chalk.dim(`${entity.type} p${entity.page}:`), errorMsg);
+            }
+        }
+    }
+    renderer.close();
+    if (useJson) {
+        console.log(formatOutput({
+            success: errors.length === 0,
+            job_id: jobId,
+            output_dir: outDir,
+            format,
+            exported: exported.length,
+            errors: errors.length,
+            files: exported,
+            error_details: errors.length > 0 ? errors : undefined,
+        }, 'json'));
+    }
+    else {
+        console.log();
+        if (exported.length > 0) {
+            success(`Exported ${exported.length} images to: ${outDir}`);
+        }
+        if (errors.length > 0) {
+            warn(`${errors.length} elements failed to export`);
+        }
+    }
+}
+async function exportTableData(jobId, options, useJson) {
+    const format = options.format;
+    if (!['csv', 'json', 'md', 'markdown'].includes(format)) {
+        if (useJson) {
+            console.log(formatOutput({ success: false, error: `Invalid format: ${format}. Use: csv, json, md, png, jpg` }, 'json'));
+        }
+        else {
+            error(`Invalid format: ${format}. Use: csv, json, md, png, jpg`);
+        }
+        process.exit(EXIT_CODES.INVALID_ARGS);
+    }
+    // Fetch entities to get table IDs
+    const entitiesResponse = await withSpinner('Fetching elements', () => get(`api/ocr/jobs/${jobId}/entities`, { type: 'tables' }));
+    const tableEntities = entitiesResponse.entities.filter(e => e.type === 'table');
+    if (tableEntities.length === 0) {
+        if (useJson) {
+            console.log(formatOutput({ success: true, job_id: jobId, message: 'No tables found', exported: 0 }, 'json'));
+        }
+        else {
+            warn('No tables found to export');
+        }
+        return;
+    }
+    const outDir = resolve(options.outputDir);
+    if (!existsSync(outDir)) {
+        mkdirSync(outDir, { recursive: true });
+    }
+    info(`Exporting ${tableEntities.length} tables to ${outDir}...`);
+    const exported = [];
+    const errors = [];
+    for (const entity of tableEntities) {
+        try {
+            const table = await get(`api/extractions/tables/${entity.id}`);
+            let content;
+            let ext;
+            switch (format) {
+                case 'json':
+                    content = JSON.stringify(table, null, 2);
+                    ext = 'json';
+                    break;
+                case 'csv':
+                    content = markdownTableToCsv(table.content_markdown);
+                    ext = 'csv';
+                    break;
+                case 'markdown':
+                case 'md':
+                    content = table.content_markdown;
+                    ext = 'md';
+                    break;
+                default:
+                    continue;
+            }
+            const safeTitle = (entity.title || 'untitled')
+                .toLowerCase()
+                .replace(/[^a-z0-9]+/g, '-')
+                .slice(0, 30);
+            const filename = `table-p${entity.page}-${safeTitle}.${ext}`;
+            const filepath = join(outDir, filename);
+            writeFileSync(filepath, content, 'utf-8');
+            exported.push({ table_id: entity.id, file: filepath, page: entity.page });
+            if (!useJson) {
+                console.log(chalk.green('  \u2713'), chalk.dim(`p${entity.page}:`), filename);
+            }
+        }
+        catch (err) {
+            const errorMsg = err instanceof Error ? err.message : 'Unknown error';
+            errors.push({ table_id: entity.id, error: errorMsg });
+            if (!useJson) {
+                console.log(chalk.red('  \u2717'), chalk.dim(`p${entity.page}:`), errorMsg);
+            }
+        }
+    }
+    if (useJson) {
+        console.log(formatOutput({
+            success: errors.length === 0,
+            job_id: jobId,
+            output_dir: outDir,
+            format,
+            exported: exported.length,
+            errors: errors.length,
+            files: exported,
+            error_details: errors.length > 0 ? errors : undefined,
+        }, 'json'));
+    }
+    else {
+        console.log();
+        if (exported.length > 0) {
+            success(`Exported ${exported.length} tables to: ${outDir}`);
+        }
+        if (errors.length > 0) {
+            warn(`${errors.length} tables failed to export`);
+        }
+    }
+}
+//# sourceMappingURL=elements.js.map