@okrapdf/cli 0.1.3 → 0.1.5

package/dist/lib/scorer.test.js

@@ -0,0 +1,974 @@
+/**
+ * Tests for OCR Review Scorer
+ *
+ * These tests prove the deterministic scoring workflows work correctly
+ * for prioritizing pages in large document review.
+ */
+import { describe, it, expect, beforeEach } from 'vitest';
+import { OcrReviewScorer, createScorer, compareOcrToMarkdown, parsePagesJson, parseTablesJson, formatScoredPages, formatStats, DEFAULT_CONFIG, 
+// Financial verification
+getConfidenceTier, categorizeByConfidence, runCheckSums, detectAnomalies, validateCrossQuery, runFinancialVerification, formatVerificationReport, } from './scorer.js';
+// ============================================================================
+// Test Fixtures
+// ============================================================================
+function createMockPage(overrides = {}) {
+    return {
+        page: 1,
+        status: 'pending',
+        total: 0,
+        verified: 0,
+        pending: 0,
+        flagged: 0,
+        rejected: 0,
+        avgConfidence: 0.9,
+        hasOcr: true,
+        ocrLineCount: 50,
+        hasCoverageGaps: false,
+        uncoveredCount: 0,
+        resolution: null,
+        classification: null,
+        isStale: false,
+        ...overrides,
+    };
+}
+function createMockTable(overrides = {}) {
+    return {
+        id: 'table-1',
+        page_number: 1,
+        markdown: '| Col1 | Col2 |\n|------|------|\n| A | B |',
+        verification_status: 'pending',
+        confidence: 0.85,
+        ...overrides,
+    };
+}
+function createMockOcrBlocks() {
+    return [
+        { text: 'Header Text', bbox: { x: 100, y: 50, width: 200, height: 20 }, confidence: 0.95 },
+        { text: 'First paragraph content here.', bbox: { x: 50, y: 100, width: 400, height: 40 }, confidence: 0.88 },
+        { text: 'Table cell 1', bbox: { x: 50, y: 200, width: 100, height: 20 }, confidence: 0.82 },
+        { text: 'Table cell 2', bbox: { x: 150, y: 200, width: 100, height: 20 }, confidence: 0.80 },
+    ];
+}
+// ============================================================================
+// Core Scorer Tests
+// ============================================================================
+describe('OcrReviewScorer', () => {
+    let scorer;
+    beforeEach(() => {
+        scorer = new OcrReviewScorer();
+    });
+    describe('scorePage', () => {
+        it('should return zero score for high confidence page with no tables', () => {
+            const page = createMockPage({ avgConfidence: 0.95 });
+            const result = scorer.scorePage(page);
+            expect(result.score).toBeLessThan(10);
+            expect(result.tableCount).toBe(0);
+            expect(result.flags).not.toContain('has_tables');
+        });
+        it('should increase score for pages with tables', () => {
+            const page = createMockPage({ page: 1 });
+            const tables = [
+                createMockTable({ page_number: 1 }),
+                createMockTable({ page_number: 1, id: 'table-2' }),
+            ];
+            scorer.loadTables(tables);
+            const result = scorer.scorePage(page);
+            expect(result.score).toBeGreaterThan(0);
+            expect(result.tableCount).toBe(2);
+            expect(result.breakdown.tableScore).toBe(20); // 2 * 10 default weight
+            expect(result.flags).toContain('has_tables');
+        });
+        it('should increase score for low confidence pages', () => {
+            const highConfPage = createMockPage({ avgConfidence: 0.95, page: 1 });
+            const lowConfPage = createMockPage({ avgConfidence: 0.5, page: 2 });
+            const highResult = scorer.scorePage(highConfPage);
+            const lowResult = scorer.scorePage(lowConfPage);
+            expect(lowResult.score).toBeGreaterThan(highResult.score);
+            expect(lowResult.breakdown.confidenceScore).toBeGreaterThan(highResult.breakdown.confidenceScore);
+            expect(lowResult.flags).toContain('low_confidence');
+        });
+        it('should add penalty for coverage gaps', () => {
+            const noGaps = createMockPage({ hasCoverageGaps: false });
+            const hasGaps = createMockPage({ hasCoverageGaps: true });
+            const noGapsResult = scorer.scorePage(noGaps);
+            const hasGapsResult = scorer.scorePage(hasGaps);
+            expect(hasGapsResult.score).toBeGreaterThan(noGapsResult.score);
+            expect(hasGapsResult.breakdown.coverageScore).toBe(DEFAULT_CONFIG.weights.coverageGap);
+            expect(hasGapsResult.flags).toContain('coverage_gaps');
+        });
+        it('should add penalty for flagged entities', () => {
+            const noFlagged = createMockPage({ flagged: 0 });
+            const hasFlagged = createMockPage({ flagged: 3 });
+            const noFlaggedResult = scorer.scorePage(noFlagged);
+            const hasFlaggedResult = scorer.scorePage(hasFlagged);
+            expect(hasFlaggedResult.score).toBeGreaterThan(noFlaggedResult.score);
+            expect(hasFlaggedResult.breakdown.flaggedScore).toBe(3 * DEFAULT_CONFIG.weights.flaggedEntity);
+            expect(hasFlaggedResult.flags).toContain('has_flagged');
+        });
+    });
+    describe('scoreAll', () => {
+        it('should sort pages by score descending (highest priority first)', () => {
+            const pages = [
+                createMockPage({ page: 1, avgConfidence: 0.95 }), // Low priority
+                createMockPage({ page: 2, avgConfidence: 0.5 }), // High priority (low confidence)
+                createMockPage({ page: 3, avgConfidence: 0.8 }), // Medium priority
+            ];
+            const results = scorer.scoreAll(pages);
+            expect(results[0].page).toBe(2); // Lowest confidence = highest priority
+            expect(results[1].page).toBe(3);
+            expect(results[2].page).toBe(1); // Highest confidence = lowest priority
+        });
+        it('should correctly combine multiple scoring factors', () => {
+            const tables = [
+                createMockTable({ page_number: 2 }),
+                createMockTable({ page_number: 2, id: 'table-2' }),
+                createMockTable({ page_number: 2, id: 'table-3' }),
+            ];
+            scorer.loadTables(tables);
+            const pages = [
+                createMockPage({ page: 1, avgConfidence: 0.6, hasCoverageGaps: true }), // Low conf + gaps
+                createMockPage({ page: 2, avgConfidence: 0.9 }), // 3 tables but high conf
+                createMockPage({ page: 3, avgConfidence: 0.7, flagged: 2 }), // Low conf + flagged
+            ];
+            const results = scorer.scoreAll(pages);
+            // Page 3 should be highest: low conf + 2 flagged entities
+            // Page 1: low conf + gaps
+            // Page 2: 3 tables but high confidence
+            expect(results[0].page).toBe(3);
+            expect(results.map(r => r.page)).toEqual(expect.arrayContaining([1, 2, 3]));
+        });
+    });
+    describe('filter', () => {
+        it('should filter by status', () => {
+            const pages = [
+                createMockPage({ page: 1, status: 'pending' }),
+                createMockPage({ page: 2, status: 'complete' }),
+                createMockPage({ page: 3, status: 'flagged' }),
+            ];
+            const pending = scorer.filter(pages, { status: 'pending' });
+            expect(pending.length).toBe(1);
+            expect(pending[0].page).toBe(1);
+            const multiple = scorer.filter(pages, { status: ['pending', 'flagged'] });
+            expect(multiple.length).toBe(2);
+        });
+        it('should filter by confidence range', () => {
+            const pages = [
+                createMockPage({ page: 1, avgConfidence: 0.95 }),
+                createMockPage({ page: 2, avgConfidence: 0.75 }),
+                createMockPage({ page: 3, avgConfidence: 0.5 }),
+            ];
+            const lowConf = scorer.filter(pages, { maxConfidence: 0.8 });
+            expect(lowConf.length).toBe(2);
+            expect(lowConf.map(p => p.page)).toEqual([2, 3]);
+            const highConf = scorer.filter(pages, { minConfidence: 0.9 });
+            expect(highConf.length).toBe(1);
+            expect(highConf[0].page).toBe(1);
+        });
+        it('should filter by table count', () => {
+            const tables = [
+                createMockTable({ page_number: 1 }),
+                createMockTable({ page_number: 2, id: 't2' }),
+                createMockTable({ page_number: 2, id: 't3' }),
+            ];
+            scorer.loadTables(tables);
+            const pages = [
+                createMockPage({ page: 1 }),
+                createMockPage({ page: 2 }),
+                createMockPage({ page: 3 }),
+            ];
+            const withTables = scorer.filter(pages, { minTables: 1 });
+            expect(withTables.length).toBe(2);
+            const manyTables = scorer.filter(pages, { minTables: 2 });
+            expect(manyTables.length).toBe(1);
+            expect(manyTables[0].page).toBe(2);
+            const noTables = scorer.filter(pages, { maxTables: 0 });
+            expect(noTables.length).toBe(1);
+            expect(noTables[0].page).toBe(3);
+        });
+        it('should filter by coverage gaps', () => {
+            const pages = [
+                createMockPage({ page: 1, hasCoverageGaps: false }),
+                createMockPage({ page: 2, hasCoverageGaps: true }),
+            ];
+            const withGaps = scorer.filter(pages, { hasGaps: true });
+            expect(withGaps.length).toBe(1);
+            expect(withGaps[0].page).toBe(2);
+            const noGaps = scorer.filter(pages, { hasGaps: false });
+            expect(noGaps.length).toBe(1);
+            expect(noGaps[0].page).toBe(1);
+        });
+    });
+    describe('getAutoApprovable', () => {
+        it('should identify pages that can be auto-approved', () => {
+            const pages = [
+                createMockPage({ page: 1, avgConfidence: 0.98, status: 'pending', total: 0 }),
+                createMockPage({ page: 2, avgConfidence: 0.85, status: 'pending', total: 0 }), // Low conf
+                createMockPage({ page: 3, avgConfidence: 0.98, status: 'complete', total: 0 }), // Already done
+                createMockPage({ page: 4, avgConfidence: 0.98, status: 'pending', hasCoverageGaps: true }), // Has gaps
+                createMockPage({ page: 5, avgConfidence: 0.98, status: 'pending', flagged: 1 }), // Has flagged
+            ];
+            // Add tables to page 1 to test table threshold
+            scorer.loadTables([createMockTable({ page_number: 6 })]);
+            pages.push(createMockPage({ page: 6, avgConfidence: 0.98, status: 'pending', total: 0 }));
+            const autoApprovable = scorer.getAutoApprovable(pages);
+            // Only page 1 should be auto-approvable (page 6 has tables)
+            expect(autoApprovable.length).toBe(1);
+            expect(autoApprovable[0].page).toBe(1);
+        });
+    });
+    describe('getRequireReview', () => {
+        it('should identify pages that need human review', () => {
+            const tables = [createMockTable({ page_number: 3 })];
+            scorer.loadTables(tables);
+            const pages = [
+                createMockPage({ page: 1, avgConfidence: 0.98 }), // High conf, no issues
+                createMockPage({ page: 2, avgConfidence: 0.5 }), // Low confidence
+                createMockPage({ page: 3, avgConfidence: 0.95 }), // Has table
+                createMockPage({ page: 4, hasCoverageGaps: true }), // Has gaps
+                createMockPage({ page: 5, flagged: 2 }), // Has flagged
+            ];
+            const needsReview = scorer.getRequireReview(pages);
+            // Pages 2, 3, 4, 5 should need review
+            expect(needsReview.length).toBe(4);
+            expect(needsReview.map(p => p.page)).not.toContain(1);
+        });
+    });
+    describe('computeStats', () => {
+        it('should compute accurate statistics', () => {
+            const tables = [
+                createMockTable({ page_number: 1 }),
+                createMockTable({ page_number: 2, id: 't2' }),
+                createMockTable({ page_number: 2, id: 't3' }),
+            ];
+            scorer.loadTables(tables);
+            const pages = [
+                createMockPage({ page: 1, status: 'pending', avgConfidence: 0.95 }),
+                createMockPage({ page: 2, status: 'pending', avgConfidence: 0.6 }),
+                createMockPage({ page: 3, status: 'complete', avgConfidence: 0.9 }),
+                createMockPage({ page: 4, status: 'flagged', avgConfidence: 0.4, hasCoverageGaps: true }),
+            ];
+            const stats = scorer.computeStats(pages);
+            expect(stats.totalPages).toBe(4);
+            expect(stats.pagesWithTables).toBe(2);
+            expect(stats.totalTables).toBe(3);
+            expect(stats.avgConfidence).toBeCloseTo(0.7125, 2);
+            expect(stats.lowConfidencePages).toBe(2); // Pages with conf < 0.7
+            expect(stats.pagesWithGaps).toBe(1);
+            expect(stats.byStatus).toEqual({
+                pending: 2,
+                complete: 1,
+                flagged: 1,
+            });
+        });
+    });
+});
+// ============================================================================
+// Scoring Strategy Tests
+// ============================================================================
+describe('Scoring Strategies', () => {
+    it('should create scorer with direct strategy (ignores tables)', () => {
+        const scorer = createScorer('direct');
+        const tables = [createMockTable({ page_number: 1 })];
+        scorer.loadTables(tables);
+        const page = createMockPage({ page: 1, avgConfidence: 0.8 });
+        const result = scorer.scorePage(page);
+        expect(result.breakdown.tableScore).toBe(0);
+        expect(result.breakdown.structureScore).toBe(0);
+    });
+    it('should create scorer with structure strategy (emphasizes tables)', () => {
+        const scorer = createScorer('structure');
+        const tables = [createMockTable({ page_number: 1 })];
+        scorer.loadTables(tables);
+        const page = createMockPage({ page: 1 });
+        const result = scorer.scorePage(page);
+        // Structure strategy has tableCount weight of 20 vs default 10
+        expect(result.breakdown.tableScore).toBe(20);
+    });
+    it('should create scorer with comparison strategy', () => {
+        const scorer = createScorer('comparison');
+        const result = scorer.scorePage(createMockPage());
+        // Comparison strategy has lower table weight
+        expect(result.score).toBeDefined();
+    });
+});
+// ============================================================================
+// OCR vs Markdown Comparison Tests
+// ============================================================================
+describe('compareOcrToMarkdown', () => {
+    it('should detect minimal difference for matching content', () => {
+        const ocrBlocks = [
+            { text: 'Hello World', confidence: 0.95 },
+            { text: 'This is a test.', confidence: 0.90 },
+        ];
+        const markdown = 'Hello World\n\nThis is a test.';
+        const result = compareOcrToMarkdown(ocrBlocks, markdown);
+        expect(result.deltaPct).toBeLessThan(0.1);
+        expect(result.flags).not.toContain('high_char_delta');
+    });
+    it('should detect significant difference when content differs', () => {
+        const ocrBlocks = [
+            { text: 'This is a very long paragraph with lots of content that was captured by OCR.', confidence: 0.85 },
+        ];
+        const markdown = 'Short.';
+        const result = compareOcrToMarkdown(ocrBlocks, markdown);
+        expect(result.deltaPct).toBeGreaterThan(0.5);
+        expect(result.flags).toContain('high_char_delta');
+    });
+    it('should evaluate spatial integrity', () => {
+        const ocrBlocks = [
+            { text: 'Line 1', bbox: { x: 0, y: 0, width: 100, height: 20 }, confidence: 0.9 },
+            { text: 'Line 2', bbox: { x: 0, y: 30, width: 100, height: 20 }, confidence: 0.9 },
+            { text: 'Line 3', bbox: { x: 0, y: 60, width: 100, height: 20 }, confidence: 0.9 },
+        ];
+        const markdown = 'Line 1\n\nLine 2\n\nLine 3';
+        const result = compareOcrToMarkdown(ocrBlocks, markdown);
+        expect(result.spatialIntegrity).toBeGreaterThan(0.5);
+    });
+    it('should evaluate table structure preservation', () => {
+        const ocrBlocks = [
+            { text: 'Col1', bbox: { x: 50, y: 0, width: 50, height: 20 } },
+            { text: 'Col2', bbox: { x: 150, y: 0, width: 50, height: 20 } },
+            { text: 'A', bbox: { x: 50, y: 30, width: 50, height: 20 } },
+            { text: 'B', bbox: { x: 150, y: 30, width: 50, height: 20 } },
+        ];
+        const markdown = '| Col1 | Col2 |\n|------|------|\n| A | B |';
+        const result = compareOcrToMarkdown(ocrBlocks, markdown);
+        expect(result.tableStructureScore).toBeGreaterThan(0.5);
+    });
+    it('should flag missing OCR', () => {
+        const ocrBlocks = [];
+        const markdown = 'Some content that exists in markdown but not OCR';
+        const result = compareOcrToMarkdown(ocrBlocks, markdown);
+        expect(result.flags).toContain('missing_ocr');
+    });
+    it('should flag missing markdown', () => {
+        const ocrBlocks = [
+            { text: 'Content in OCR but not in markdown', confidence: 0.9 },
+        ];
+        const markdown = '';
+        const result = compareOcrToMarkdown(ocrBlocks, markdown);
+        expect(result.flags).toContain('missing_markdown');
+    });
+});
+// ============================================================================
+// JSON Parsing Tests
+// ============================================================================
+describe('JSON Parsing', () => {
+    it('should parse pages JSON array', () => {
+        const json = JSON.stringify([
+            { page: 1, status: 'pending', avgConfidence: 0.9, total: 0, verified: 0, pending: 0, flagged: 0, rejected: 0, hasOcr: true, ocrLineCount: 50, hasCoverageGaps: false, uncoveredCount: 0, resolution: null, classification: null, isStale: false },
+        ]);
+        const pages = parsePagesJson(json);
+        expect(pages.length).toBe(1);
+        expect(pages[0].page).toBe(1);
+    });
+    it('should parse pages JSON object with pages key', () => {
+        const json = JSON.stringify({
+            pages: [{ page: 1, status: 'pending', avgConfidence: 0.9, total: 0, verified: 0, pending: 0, flagged: 0, rejected: 0, hasOcr: true, ocrLineCount: 50, hasCoverageGaps: false, uncoveredCount: 0, resolution: null, classification: null, isStale: false }],
+        });
+        const pages = parsePagesJson(json);
+        expect(pages.length).toBe(1);
+    });
+    it('should parse tables JSON array', () => {
+        const json = JSON.stringify([
+            { id: 't1', page_number: 1, markdown: '| A |', verification_status: 'pending' },
+        ]);
+        const tables = parseTablesJson(json);
+        expect(tables.length).toBe(1);
+        expect(tables[0].id).toBe('t1');
+    });
+    it('should parse tables JSON object with tables key', () => {
+        const json = JSON.stringify({
+            tables: [{ id: 't1', page_number: 1, markdown: '| A |', verification_status: 'pending' }],
+        });
+        const tables = parseTablesJson(json);
+        expect(tables.length).toBe(1);
+    });
+});
+// ============================================================================
+// Output Formatting Tests
+// ============================================================================
+describe('Output Formatting', () => {
+    const sampleScored = [
+        {
+            page: 1,
+            score: 45.5,
+            breakdown: { tableScore: 20, confidenceScore: 15.5, coverageScore: 0, flaggedScore: 0, structureScore: 10, comparisonScore: 0 },
+            tableCount: 2,
+            avgConfidence: 0.69,
+            status: 'pending',
+            flags: ['has_tables', 'low_confidence'],
+        },
+        {
+            page: 2,
+            score: 10,
+            breakdown: { tableScore: 0, confidenceScore: 5, coverageScore: 0, flaggedScore: 0, structureScore: 5, comparisonScore: 0 },
+            tableCount: 0,
+            avgConfidence: 0.9,
+            status: 'complete',
+            flags: [],
+        },
+    ];
+    it('should format as JSON', () => {
+        const output = formatScoredPages(sampleScored, 'json');
+        const parsed = JSON.parse(output);
+        expect(parsed).toBeInstanceOf(Array);
+        expect(parsed.length).toBe(2);
+        expect(parsed[0].page).toBe(1);
+    });
+    it('should format as JSONL', () => {
+        const output = formatScoredPages(sampleScored, 'jsonl');
+        const lines = output.split('\n');
+        expect(lines.length).toBe(2);
+        expect(JSON.parse(lines[0]).page).toBe(1);
+        expect(JSON.parse(lines[1]).page).toBe(2);
+    });
+    it('should format as CSV', () => {
+        const output = formatScoredPages(sampleScored, 'csv');
+        const lines = output.split('\n');
+        expect(lines[0]).toBe('page,score,tableCount,avgConfidence,status,flags');
+        expect(lines[1]).toContain('1');
+        expect(lines[1]).toContain('45.50');
+    });
+    it('should format as table', () => {
+        const output = formatScoredPages(sampleScored, 'table');
+        expect(output).toContain('Page');
+        expect(output).toContain('Score');
+        expect(output).toContain('Tables');
+        expect(output).toContain('has_tables');
+    });
+});
+describe('Stats Formatting', () => {
+    it('should format stats as JSON', () => {
+        const stats = {
+            totalPages: 100,
+            pagesWithTables: 20,
+            totalTables: 45,
+            avgConfidence: 0.85,
+            lowConfidencePages: 10,
+            pagesWithGaps: 5,
+            byStatus: { pending: 50, complete: 40, flagged: 10 },
+            byPriority: { high: 5, medium: 15, low: 80 },
+            estimatedReviewTime: { pages: 20, minutes: 25 },
+        };
+        const output = formatStats(stats, 'json');
+        const parsed = JSON.parse(output);
+        expect(parsed.totalPages).toBe(100);
+        expect(parsed.byPriority.high).toBe(5);
+    });
+    it('should format stats as table', () => {
+        const stats = {
+            totalPages: 100,
+            pagesWithTables: 20,
+            totalTables: 45,
+            avgConfidence: 0.85,
+            lowConfidencePages: 10,
+            pagesWithGaps: 5,
+            byStatus: { pending: 50, complete: 40, flagged: 10 },
+            byPriority: { high: 5, medium: 15, low: 80 },
+            estimatedReviewTime: { pages: 20, minutes: 25 },
+        };
+        const output = formatStats(stats, 'table');
+        expect(output).toContain('Total pages: 100');
+        expect(output).toContain('Pages with tables: 20');
+        expect(output).toContain('High (score >= 50): 5 pages');
+        expect(output).toContain('Estimated review: 20 pages');
+    });
+});
+// ============================================================================
+// Workflow Integration Tests
+// ============================================================================
+describe('Workflow Integration', () => {
+    it('should complete full table-priority workflow', () => {
+        // Simulate a 50-page document with varying characteristics
+        const pages = [];
+        const tables = [];
+        for (let i = 1; i <= 50; i++) {
+            const confidence = 0.5 + Math.random() * 0.5; // 0.5 - 1.0
+            const hasTables = i % 5 === 0; // Every 5th page has tables
+            const hasGaps = i % 10 === 0; // Every 10th page has gaps
+            pages.push(createMockPage({
+                page: i,
+                avgConfidence: confidence,
+                hasCoverageGaps: hasGaps,
+                status: i <= 40 ? 'pending' : 'complete',
+            }));
+            if (hasTables) {
+                tables.push(createMockTable({ page_number: i, id: `table-${i}` }));
+                if (i % 15 === 0) {
+                    // Some pages have multiple tables
+                    tables.push(createMockTable({ page_number: i, id: `table-${i}-2` }));
+                }
+            }
+        }
+        const scorer = new OcrReviewScorer();
+        scorer.loadTables(tables);
+        // Step 1: Get statistics
+        const stats = scorer.computeStats(pages);
+        expect(stats.totalPages).toBe(50);
+        expect(stats.pagesWithTables).toBe(10); // 50/5 = 10 pages with tables
+        // Step 2: Get auto-approvable pages (high confidence, no tables, no issues)
+        const autoApprovable = scorer.getAutoApprovable(pages);
+        expect(autoApprovable.length).toBeGreaterThan(0);
+        // Step 3: Get pages requiring review
+        const needsReview = scorer.getRequireReview(pages);
+        expect(needsReview.length).toBeGreaterThan(0);
+        // Step 4: Score all and verify priority ordering
+        const scored = scorer.scoreAll(pages);
+        expect(scored[0].score).toBeGreaterThanOrEqual(scored[scored.length - 1].score);
+        // Step 5: Filter for table-heavy pages
+        const tablePages = scorer.filter(pages, { minTables: 1 });
+        expect(tablePages.length).toBe(10);
+        // Verify the highest priority pages have tables or low confidence
+        const topPriority = scored.slice(0, 5);
+        for (const page of topPriority) {
+            const hasIssue = page.tableCount > 0 ||
+                page.avgConfidence < 0.7 ||
+                page.flags.includes('coverage_gaps');
+            expect(hasIssue).toBe(true);
+        }
+    });
+    it('should complete OCR comparison workflow', () => {
+        const ocrBlocks = createMockOcrBlocks();
+        const markdown = `# Header Text
+
+First paragraph content here.
+
+| Table cell 1 | Table cell 2 |
+|--------------|--------------|`;
+        const comparison = compareOcrToMarkdown(ocrBlocks, markdown);
+        // Should detect reasonable similarity
+        expect(comparison.deltaPct).toBeLessThan(0.5);
+        expect(comparison.spatialIntegrity).toBeGreaterThan(0);
+        expect(comparison.tableStructureScore).toBeGreaterThan(0.5);
+    });
+    it('should handle empty document gracefully', () => {
+        const scorer = new OcrReviewScorer();
+        scorer.loadTables([]);
+        const pages = [];
+        const stats = scorer.computeStats(pages);
+        expect(stats.totalPages).toBe(0);
+        expect(stats.avgConfidence).toBe(0);
+        expect(stats.byPriority).toEqual({ high: 0, medium: 0, low: 0 });
+    });
+    it('should handle all pages with tables', () => {
+        const pages = [
+            createMockPage({ page: 1 }),
+            createMockPage({ page: 2 }),
+            createMockPage({ page: 3 }),
+        ];
+        const tables = [
+            createMockTable({ page_number: 1 }),
+            createMockTable({ page_number: 2 }),
+            createMockTable({ page_number: 3 }),
+        ];
+        const scorer = new OcrReviewScorer();
+        scorer.loadTables(tables);
+        const autoApprovable = scorer.getAutoApprovable(pages);
+        expect(autoApprovable.length).toBe(0); // All have tables
+        const needsReview = scorer.getRequireReview(pages);
+        expect(needsReview.length).toBe(3);
+    });
+});
+// ============================================================================
+// Custom Configuration Tests
+// ============================================================================
+describe('Custom Configuration', () => {
+    it('should apply custom weights', () => {
+        const scorer = new OcrReviewScorer({
+            weights: {
+                tableCount: 100, // Much higher table weight
+                inverseConfidence: 10,
+                coverageGap: 5,
+                flaggedEntity: 5,
+                structurePenalty: 5,
+                comparisonDelta: 5,
+            },
+        });
+        const tables = [createMockTable({ page_number: 1 })];
+        scorer.loadTables(tables);
+        const page = createMockPage({ page: 1 });
+        const result = scorer.scorePage(page);
+        expect(result.breakdown.tableScore).toBe(100);
+    });
+    it('should apply custom thresholds', () => {
+        const scorer = new OcrReviewScorer({
+            thresholds: {
+                autoApproveConfidence: 0.99, // Very strict
+                requireReviewConfidence: 0.9,
+                maxTablesForAutoApprove: 0,
+                minOcrLinesForContent: 5,
+            },
+        });
+        const pages = [
+            createMockPage({ page: 1, avgConfidence: 0.98, status: 'pending' }),
+            createMockPage({ page: 2, avgConfidence: 0.995, status: 'pending' }),
+        ];
+        const autoApprovable = scorer.getAutoApprovable(pages);
+        expect(autoApprovable.length).toBe(1);
+        expect(autoApprovable[0].page).toBe(2); // Only 0.995 passes 0.99 threshold
+    });
+});
+// ============================================================================
+// Financial Verification Tests
+// ============================================================================
+describe('Financial Verification', () => {
+    describe('getConfidenceTier', () => {
+        it('should categorize high confidence as auto_approve', () => {
+            expect(getConfidenceTier(0.96)).toBe('auto_approve');
+            expect(getConfidenceTier(0.99)).toBe('auto_approve');
+            expect(getConfidenceTier(1.0)).toBe('auto_approve');
+        });
+        it('should categorize medium confidence as spot_check', () => {
+            expect(getConfidenceTier(0.70)).toBe('spot_check');
+            expect(getConfidenceTier(0.85)).toBe('spot_check');
+            expect(getConfidenceTier(0.94)).toBe('spot_check');
+        });
+        it('should categorize low confidence as manual_review', () => {
+            expect(getConfidenceTier(0.69)).toBe('manual_review');
+            expect(getConfidenceTier(0.50)).toBe('manual_review');
+            expect(getConfidenceTier(0.0)).toBe('manual_review');
+        });
+    });
+    describe('categorizeByConfidence', () => {
+        it('should categorize pages into tiers', () => {
+            const pages = [
+                createMockPage({ page: 1, avgConfidence: 0.98 }),
+                createMockPage({ page: 2, avgConfidence: 0.85 }),
+                createMockPage({ page: 3, avgConfidence: 0.60 }),
+                createMockPage({ page: 4, avgConfidence: 0.75 }),
+                createMockPage({ page: 5, avgConfidence: 0.96 }),
+            ];
+            const categorized = categorizeByConfidence(pages);
+            expect(categorized.auto_approve.length).toBe(2); // pages 1, 5
+            expect(categorized.spot_check.length).toBe(2); // pages 2, 4
+            expect(categorized.manual_review.length).toBe(1); // page 3
+        });
+    });
+    describe('runCheckSums', () => {
+        it('should validate balance sheet identity', () => {
+            // Valid balance sheet: Assets = Liabilities + Equity
+            const validValues = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 400000,
+            };
+            const results = runCheckSums(validValues);
+            const balanceSheet = results.find(r => r.rule.name === 'Balance Sheet Identity');
+            expect(balanceSheet).toBeDefined();
+            expect(balanceSheet.result.passed).toBe(true);
+        });
+        it('should detect balance sheet imbalance', () => {
+            // Invalid: Assets don't equal Liabilities + Equity
+            const invalidValues = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 300000, // Should be 400000
+            };
+            const results = runCheckSums(invalidValues);
+            const balanceSheet = results.find(r => r.rule.name === 'Balance Sheet Identity');
+            expect(balanceSheet).toBeDefined();
+            expect(balanceSheet.result.passed).toBe(false);
+            expect(balanceSheet.result.tolerancePct).toBeGreaterThan(1);
+        });
+        it('should validate net income calculation', () => {
+            const values = {
+                total_revenue: 5000000,
+                total_expenses: 4200000,
+                net_income: 800000,
+            };
+            const results = runCheckSums(values);
+            const netIncomeCheck = results.find(r => r.rule.name === 'Net Income Check');
+            expect(netIncomeCheck).toBeDefined();
+            expect(netIncomeCheck.result.passed).toBe(true);
+        });
+        it('should allow small tolerance for rounding', () => {
+            // 0.5% rounding difference should pass
+            const values = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 405000, // 0.5% over
+            };
+            const results = runCheckSums(values);
+            const balanceSheet = results.find(r => r.rule.name === 'Balance Sheet Identity');
+            expect(balanceSheet.result.passed).toBe(true);
+        });
+        it('should skip rules with insufficient data', () => {
+            const values = {
+                total_assets: 1000000, // Only one field
+            };
+            const results = runCheckSums(values);
+            // Should not validate balance sheet without liabilities/equity
+            expect(results.length).toBe(0);
+        });
+    });
+    describe('detectAnomalies', () => {
+        it('should detect decimal point error', () => {
+            const historical = [
+                { revenue: 5000000 },
+                { revenue: 5200000 },
+                { revenue: 4800000 },
+            ];
+            const current = {
+                revenue: 50000, // Missing two zeros - 100x error
+            };
+            const anomalies = detectAnomalies(current, historical);
+            expect(anomalies.length).toBe(1);
+            expect(anomalies[0].field).toBe('revenue');
+            expect(anomalies[0].possibleCause).toContain('decimal point error');
+        });
+        it('should not flag normal variance', () => {
+            const historical = [
+                { revenue: 5000000 },
+                { revenue: 5200000 },
+                { revenue: 4800000 },
+            ];
+            const current = {
+                revenue: 5100000, // Normal year-over-year change
+            };
+            const anomalies = detectAnomalies(current, historical);
+            expect(anomalies.length).toBe(0);
+        });
+        it('should handle insufficient historical data', () => {
+            const historical = [{ revenue: 5000000 }]; // Only one data point
+            const current = { revenue: 50000 };
+            const anomalies = detectAnomalies(current, historical);
+            expect(anomalies.length).toBe(0); // Can't detect anomalies without enough history
+        });
+        it('should detect significant outliers', () => {
+            const historical = [
+                { profit: 100000 },
+                { profit: 120000 },
+                { profit: 110000 },
+                { profit: 105000 },
+            ];
+            const current = {
+                profit: 1000000, // 10x higher than historical
+            };
+            const anomalies = detectAnomalies(current, historical);
+            expect(anomalies.length).toBe(1);
+            expect(anomalies[0].zScore).toBeGreaterThan(3);
+        });
+    });
+    describe('validateCrossQuery', () => {
+        it('should validate total assets breakdown', () => {
+            const values = {
+                total_assets: 1000000,
+                current_assets: 400000,
+                non_current_assets: 600000,
+            };
+            const results = validateCrossQuery(values);
+            const assetsCheck = results.find(r => r.rule.name === 'Total Assets Components');
+            expect(assetsCheck).toBeDefined();
+            expect(assetsCheck.passed).toBe(true);
+        });
+        it('should detect missing component in total', () => {
+            const values = {
+                total_assets: 1000000,
+                current_assets: 400000,
+                // non_current_assets missing but total doesn't match
+            };
+            const results = validateCrossQuery(values);
+            const assetsCheck = results.find(r => r.rule.name === 'Total Assets Components');
+            expect(assetsCheck).toBeDefined();
+            expect(assetsCheck.passed).toBe(false);
+            expect(assetsCheck.differencePercent).toBeGreaterThan(2);
+        });
+        it('should allow tolerance for incomplete components', () => {
+            const values = {
+                total_revenue: 1000000,
+                product_revenue: 800000,
+                service_revenue: 180000,
+                // other_revenue = 20000 is missing but within tolerance
+            };
+            const results = validateCrossQuery(values);
+            const revenueCheck = results.find(r => r.rule.name === 'Total Revenue Components');
+            expect(revenueCheck).toBeDefined();
+            expect(revenueCheck.passed).toBe(true); // Within 2% tolerance
+        });
+    });
+    describe('runFinancialVerification', () => {
+        it('should pass for valid financial data with high confidence', () => {
+            const values = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 400000,
+            };
+            const report = runFinancialVerification(values, 0.98);
+            expect(report.overallStatus).toBe('pass');
+            expect(report.confidenceTier).toBe('auto_approve');
+            expect(report.issues.length).toBe(0);
+        });
+        it('should fail for low confidence', () => {
+            const values = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 400000,
+            };
+            const report = runFinancialVerification(values, 0.50);
+            expect(report.overallStatus).toBe('fail');
+            expect(report.confidenceTier).toBe('manual_review');
+            expect(report.issues.some(i => i.includes('Low OCR confidence'))).toBe(true);
+        });
+        it('should fail for check-sum violations', () => {
+            const values = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 200000, // Wrong!
+            };
+            const report = runFinancialVerification(values, 0.98);
+            expect(report.overallStatus).toBe('fail');
+            expect(report.issues.some(i => i.includes('Balance Sheet Identity failed'))).toBe(true);
+        });
+        it('should fail for anomalies with historical data', () => {
+            const values = {
+                revenue: 50000, // Should be ~5 million
+            };
+            const historical = [
+                { revenue: 5000000 },
+                { revenue: 5200000 },
+                { revenue: 4800000 },
+            ];
+            const report = runFinancialVerification(values, 0.98, historical);
+            expect(report.overallStatus).toBe('fail');
+            expect(report.anomalies.length).toBe(1);
+        });
+        it('should provide recommendations for issues', () => {
+            const values = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 200000,
+            };
+            const report = runFinancialVerification(values, 0.75);
+            expect(report.recommendations.length).toBeGreaterThan(0);
+            expect(report.recommendations.some(r => r.includes('spot-check'))).toBe(true);
+        });
+    });
+    describe('formatVerificationReport', () => {
+        it('should format passing report', () => {
+            const values = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 400000,
+            };
+            const report = runFinancialVerification(values, 0.98);
+            const formatted = formatVerificationReport(report);
+            expect(formatted).toContain('Overall Status: PASS');
+            expect(formatted).toContain('AUTO APPROVE');
+        });
+        it('should format failing report with details', () => {
+            const values = {
+                total_assets: 1000000,
+                total_liabilities: 600000,
+                equity: 200000,
+            };
+            const report = runFinancialVerification(values, 0.50);
+            const formatted = formatVerificationReport(report);
+            expect(formatted).toContain('Overall Status: FAIL');
+            expect(formatted).toContain('Check-Sum Validations');
+            expect(formatted).toContain('Issues:');
+            expect(formatted).toContain('Recommendations:');
+        });
+    });
+});
+// ============================================================================
+// End-to-End Financial Workflow Test
+// ============================================================================
+describe('Financial Document Review Workflow', () => {
+    it('should complete full financial verification workflow', () => {
+        // Simulate extracting data from a financial report
+        // Step 1: OCR extracts pages with confidence scores
+        const pages = [
+            createMockPage({ page: 1, avgConfidence: 0.98, total: 2 }), // Cover page
+            createMockPage({ page: 2, avgConfidence: 0.92, total: 5 }), // Balance sheet
+            createMockPage({ page: 3, avgConfidence: 0.95, total: 4 }), // Income statement
+            createMockPage({ page: 4, avgConfidence: 0.65, total: 8 }), // Notes (low quality)
+            createMockPage({ page: 5, avgConfidence: 0.88, total: 3 }), // Cash flow
+        ];
+        const tables = [
+            createMockTable({ page_number: 2, id: 'balance-sheet-1' }),
+            createMockTable({ page_number: 2, id: 'balance-sheet-2' }),
+            createMockTable({ page_number: 3, id: 'income-statement' }),
+            createMockTable({ page_number: 5, id: 'cash-flow' }),
+        ];
+        // Step 2: Categorize by confidence tier
+        const categorized = categorizeByConfidence(pages);
+        expect(categorized.auto_approve.length).toBe(2); // Pages 1 (98%), 3 (95%)
+        expect(categorized.spot_check.length).toBe(2); // Pages 2, 5
+        expect(categorized.manual_review.length).toBe(1); // Page 4 (65%)
+        // Step 3: Score pages for review priority
+        const scorer = new OcrReviewScorer();
+        scorer.loadTables(tables);
+        const scored = scorer.scoreAll(pages);
+        // Highest priority pages should include those with low confidence or tables
+        // Page 4 has lowest confidence (0.65), pages 2,3,5 have tables
+        const topPriorityPages = scored.slice(0, 3).map(s => s.page);
+        expect(topPriorityPages).toContain(4); // Low confidence page
+        expect(topPriorityPages).toContain(2); // Has 2 tables
+        // Step 4: Simulate extracted financial values
+        const extractedValues = {
+            total_assets: 10500000,
+            current_assets: 3500000,
+            non_current_assets: 7000000,
+            total_liabilities: 6200000,
+            current_liabilities: 2100000,
+            non_current_liabilities: 4100000,
+            equity: 4300000,
+            working_capital: 1400000, // current_assets - current_liabilities
+            total_revenue: 25000000,
+            total_expenses: 22500000,
+            net_income: 2500000,
+        };
+        // Step 5: Run financial verification
+        const report = runFinancialVerification(extractedValues, 0.96);
+        // Should pass - data is consistent and confidence is high enough
+        expect(report.overallStatus).toBe('pass');
+        expect(report.checkSums.every(c => c.result.passed)).toBe(true);
+        // Step 6: Run cross-query validation
+        const crossResults = validateCrossQuery(extractedValues);
+        expect(crossResults.every(r => r.passed)).toBe(true);
+        // Step 7: Test anomaly detection with historical data
+        const historicalData = [
+            { total_revenue: 24000000, net_income: 2400000 },
+            { total_revenue: 23000000, net_income: 2300000 },
+            { total_revenue: 22000000, net_income: 2200000 },
+        ];
+        const anomalies = detectAnomalies(extractedValues, historicalData);
+        expect(anomalies.length).toBe(0); // No anomalies - growth is normal
+        // Step 8: Format report for analyst
+        const formatted = formatVerificationReport(report);
+        expect(formatted).toContain('PASS');
+    });
+    it('should catch common OCR errors in financial data', () => {
+        // Simulate OCR errors that commonly occur
+        // Error 1: Decimal point shift
+        const historicalRevenue = [
+            { revenue: 5000000 },
+            { revenue: 5200000 },
+            { revenue: 4800000 },
+        ];
+        const decimalError = { revenue: 50000 }; // Missing 2 zeros
+        const anomalies = detectAnomalies(decimalError, historicalRevenue);
+        expect(anomalies.length).toBe(1);
+        expect(anomalies[0].possibleCause).toContain('decimal point error');
+        // Error 2: Balance sheet doesn't balance
+        const unbalanced = {
+            total_assets: 1000000,
+            total_liabilities: 500000,
+            equity: 400000, // Should be 500000
+        };
+        const checksumResults = runCheckSums(unbalanced);
+        const balanceCheck = checksumResults.find(r => r.rule.name === 'Balance Sheet Identity');
+        expect(balanceCheck.result.passed).toBe(false);
+        // Error 3: Components don't sum to total
+        const missingComponent = {
+            total_assets: 1000000,
+            current_assets: 300000,
+            // Missing non_current_assets that should be 700000
+        };
+        const crossResults = validateCrossQuery(missingComponent);
+        const assetsCheck = crossResults.find(r => r.rule.name === 'Total Assets Components');
+        expect(assetsCheck.passed).toBe(false);
+    });
+});
+//# sourceMappingURL=scorer.test.js.map