llm-testrunner-components 1.0.6 → 1.0.9

package/dist/collection/lib/evaluation/evaluators/bleu/tests/bleu.test.js

@@ -0,0 +1,352 @@
+import { jest, describe, it, expect } from "@jest/globals";
+import { performBleuEvaluation } from "../bleu-evaluator";
+import { DEFAULT_BLEU_PASS_SCORE, EvaluationApproach, } from "../../../constants";
+describe('performBleuEvaluation', () => {
+    // Helper function to create a base request with optional overrides
+    const createRequest = (overrides = {}) => {
+        const defaults = {
+            testCaseId: 'test-001',
+            question: 'Test question',
+            expectedOutcome: 'keyword',
+            actualResponse: 'response with keyword',
+            evaluationParameters: {
+                approach: EvaluationApproach.BLEU,
+                threshold: DEFAULT_BLEU_PASS_SCORE,
+            },
+        };
+        return {
+            ...defaults,
+            ...overrides,
+            evaluationParameters: {
+                ...defaults.evaluationParameters,
+                ...overrides.evaluationParameters,
+            },
+        };
+    };
+    describe('basic functionality', () => {
+        it('should return a valid EvaluationResult structure', async () => {
+            const request = createRequest({
+                actualResponse: 'AI stands for artificial intelligence',
+                expectedOutcome: 'artificial intelligence',
+            });
+            const result = performBleuEvaluation(request);
+            expect(result).toMatchObject({
+                testCaseId: 'test-001',
+                passed: expect.any(Boolean),
+                keywordMatches: expect.any(Array),
+                timestamp: expect.any(String),
+                evaluationParameters: expect.any(Object),
+                evaluationApproachResult: expect.any(Object),
+            });
+        });
+        it('should use default threshold when not provided', async () => {
+            const request = createRequest({
+                expectedOutcome: 'test evaluation system works',
+                actualResponse: 'test evaluation system works well',
+                evaluationParameters: { approach: EvaluationApproach.BLEU },
+            });
+            const result = performBleuEvaluation(request);
+            expect(result.evaluationParameters.threshold).toBe(DEFAULT_BLEU_PASS_SCORE);
+        });
+        it('should use provided threshold when specified', async () => {
+            const customThreshold = 0.85;
+            const request = createRequest({
+                actualResponse: 'response text with multiple words',
+                expectedOutcome: 'response text with multiple',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: customThreshold,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            expect(result.evaluationParameters.threshold).toBe(customThreshold);
+        });
+    });
+    describe('single keyword evaluation', () => {
+        it('should pass when keyword with 4+ words matches exactly', async () => {
+            const request = createRequest({
+                expectedOutcome: 'the machine learning algorithm works',
+                actualResponse: 'the machine learning algorithm works',
+            });
+            const result = performBleuEvaluation(request);
+            expect(result).toMatchObject({
+                passed: true,
+                keywordMatches: [
+                    {
+                        keyword: 'the machine learning algorithm works',
+                        found: true,
+                        evaluationApproachResult: {
+                            score: expect.any(Number),
+                            approachUsed: EvaluationApproach.BLEU,
+                        },
+                    },
+                ],
+            });
+            expect(result.keywordMatches[0].evaluationApproachResult.score).toBeGreaterThanOrEqual(0.7);
+        });
+        it('should pass when keyword with 4+ words is found with high n-gram overlap', async () => {
+            const request = createRequest({
+                expectedOutcome: 'the cat sat on',
+                actualResponse: 'the cat sat on the mat',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.5,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            expect(result.keywordMatches[0]).toMatchObject({
+                found: true,
+                evaluationApproachResult: {
+                    score: expect.any(Number),
+                    approachUsed: EvaluationApproach.BLEU,
+                },
+            });
+            expect(result.keywordMatches[0].evaluationApproachResult.score).toBeGreaterThan(0);
+        });
+        it('should fail when keyword has low n-gram overlap', async () => {
+            const request = createRequest({
+                expectedOutcome: 'quantum physics research continues',
+                actualResponse: 'This is about machine learning algorithms',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.7,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            expect(result).toMatchObject({
+                passed: false,
+                keywordMatches: [
+                    {
+                        found: false,
+                        evaluationApproachResult: {
+                            score: expect.any(Number),
+                        },
+                    },
+                ],
+            });
+            expect(result.keywordMatches[0].evaluationApproachResult.score).toBeLessThan(0.7);
+        });
+    });
+    describe('4-gram limitation', () => {
+        it('should fail when keyword has fewer than 4 words', async () => {
+            const request = createRequest({
+                expectedOutcome: 'machine learning',
+                actualResponse: 'machine learning is important',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.7,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            // BLEU uses 4-gram matching, so keywords with fewer than 4 words will have very low scores
+            expect(result.passed).toBe(false);
+            expect(result.keywordMatches[0].evaluationApproachResult.score).toBeLessThan(0.7);
+        });
+        it('should work correctly with 4-word keywords', async () => {
+            const request = createRequest({
+                expectedOutcome: 'the cat sat on',
+                actualResponse: 'the cat sat on the mat',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.5,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            // 4-word keywords can produce proper 4-gram BLEU scores
+            expect(result.keywordMatches[0].found).toBe(true);
+            expect(result.keywordMatches[0].evaluationApproachResult.score).toBeGreaterThan(0);
+        });
+    });
+    describe('multiple keywords evaluation', () => {
+        it('should pass when all keywords meet threshold', async () => {
+            const request = createRequest({
+                expectedOutcome: 'Machine learning algorithms work. Artificial intelligence systems are. Neural network models perform.',
+                actualResponse: 'Machine learning algorithms work. Artificial intelligence systems are. Neural network models perform.',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.15,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            expect(result).toMatchObject({
+                passed: true,
+                evaluationApproachResult: {
+                    score: 1,
+                    approachUsed: EvaluationApproach.BLEU,
+                },
+            });
+            expect(result.keywordMatches).toHaveLength(3);
+            expect(result.keywordMatches.every(match => match.found)).toBe(true);
+        });
+        it('should fail when not all keywords meet threshold', async () => {
+            const request = createRequest({
+                expectedOutcome: 'machine learning algorithms work\nquantum computing research continues\nartificial intelligence systems',
+                actualResponse: 'Machine learning algorithms work. Artificial intelligence systems are growing.',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.7,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            expect(result.passed).toBe(false);
+            expect(result.keywordMatches).toHaveLength(3);
+            const foundCount = result.keywordMatches.filter(match => match.found).length;
+            expect(foundCount).toBeLessThan(3);
+            expect(result.evaluationApproachResult.score).toBe(foundCount / 3);
+        });
+        it('should calculate overall score as ratio of passed keywords', async () => {
+            const request = createRequest({
+                expectedOutcome: 'alpha beta gamma delta\nepsilon zeta eta theta\niota kappa lambda mu\nnu xi omicron pi',
+                actualResponse: 'alpha beta gamma delta and epsilon zeta eta theta are here',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.3,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            const foundCount = result.keywordMatches.filter(match => match.found).length;
+            expect(result).toMatchObject({
+                passed: false,
+                evaluationApproachResult: {
+                    score: foundCount / 4,
+                },
+            });
+        });
+    });
+    describe('threshold handling', () => {
+        it('should pass all keywords with threshold 0.0', async () => {
+            const request = createRequest({
+                actualResponse: 'completely unrelated text about cooking',
+                expectedOutcome: 'quantum physics research continues\nmathematics and statistics are',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.0,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            expect(result.passed).toBe(true);
+            expect(result.keywordMatches.every(m => m.found)).toBe(true);
+            expect(result.evaluationParameters.threshold).toBe(0.0);
+        });
+        it('should fail when threshold is 1.0 and match is not perfect', async () => {
+            const request = createRequest({
+                actualResponse: 'This is about machine learning concepts',
+                expectedOutcome: 'machine learning algorithms work well',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 1.0,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            expect(result.evaluationParameters.threshold).toBe(1.0);
+            expect(result.keywordMatches[0].evaluationApproachResult.score).toBeLessThan(1.0);
+            expect(result.keywordMatches[0].found).toBe(false);
+        });
+    });
+    describe('edge cases', () => {
+        it('should handle empty keywords array', async () => {
+            const request = createRequest({
+                expectedOutcome: '',
+                actualResponse: 'Some response',
+            });
+            const result = performBleuEvaluation(request);
+            expect(result).toMatchObject({
+                passed: true,
+                keywordMatches: [],
+                evaluationApproachResult: {
+                    score: 1,
+                },
+            });
+        });
+        it('should handle empty actual response', async () => {
+            const request = createRequest({
+                expectedOutcome: 'machine learning algorithms work',
+                actualResponse: '',
+            });
+            // Suppress expected warning
+            const consoleWarnSpy = jest
+                .spyOn(console, 'warn')
+                .mockImplementation(() => { });
+            const result = performBleuEvaluation(request);
+            expect(result).toMatchObject({
+                passed: false,
+                keywordMatches: [
+                    {
+                        found: false,
+                        evaluationApproachResult: {
+                            score: 0,
+                        },
+                    },
+                ],
+            });
+            consoleWarnSpy.mockRestore();
+        });
+        it('should handle whitespace-only keyword', async () => {
+            const request = createRequest({
+                expectedOutcome: '   ',
+                actualResponse: 'Some response',
+            });
+            // Suppress expected warning
+            const consoleWarnSpy = jest
+                .spyOn(console, 'warn')
+                .mockImplementation(() => { });
+            const result = performBleuEvaluation(request);
+            expect(result.keywordMatches[0]).toMatchObject({
+                found: false,
+                evaluationApproachResult: {
+                    score: 0,
+                },
+            });
+            consoleWarnSpy.mockRestore();
+        });
+        it('should handle null/undefined actualResponse gracefully', async () => {
+            const request = createRequest({
+                expectedOutcome: 'machine learning algorithms work',
+                actualResponse: null,
+            });
+            // Suppress expected warning
+            const consoleWarnSpy = jest
+                .spyOn(console, 'warn')
+                .mockImplementation(() => { });
+            const result = performBleuEvaluation(request);
+            expect(result).toMatchObject({
+                passed: false,
+                keywordMatches: [
+                    {
+                        found: false,
+                    },
+                ],
+            });
+            consoleWarnSpy.mockRestore();
+        });
+    });
+    describe('BLEU score calculation', () => {
+        it('should calculate BLEU score for partial match', async () => {
+            const request = createRequest({
+                expectedOutcome: 'the cat sat on the mat',
+                actualResponse: 'the cat sat on',
+                evaluationParameters: {
+                    approach: EvaluationApproach.BLEU,
+                    threshold: 0.3,
+                },
+            });
+            const result = performBleuEvaluation(request);
+            // Partial match should have lower BLEU score than perfect match
+            const score = result.keywordMatches[0].evaluationApproachResult.score;
+            expect(score).toBeGreaterThan(0);
+            expect(score).toBeLessThan(1.0);
+        });
+    });
+    describe('timestamp', () => {
+        it('should include a valid ISO timestamp', async () => {
+            const request = createRequest({
+                expectedOutcome: 'test evaluation system works',
+                actualResponse: 'test evaluation system works well',
+            });
+            const result = performBleuEvaluation(request);
+            expect(result.timestamp).toBeDefined();
+            expect(new Date(result.timestamp).toISOString()).toBe(result.timestamp);
+        });
+    });
+});
+//# sourceMappingURL=bleu.test.js.map

package/dist/collection/lib/evaluation/evaluators/bleu/tests/bleu.test.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"bleu.test.js","sourceRoot":"","sources":["../../../../../../src/lib/evaluation/evaluators/bleu/tests/bleu.test.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,MAAM,EAAE,MAAM,eAAe,CAAC;AAC3D,OAAO,EAAE,qBAAqB,EAAE,MAAM,mBAAmB,CAAC;AAE1D,OAAO,EACL,uBAAuB,EACvB,kBAAkB,GACnB,MAAM,oBAAoB,CAAC;AAE5B,QAAQ,CAAC,uBAAuB,EAAE,GAAG,EAAE;IACrC,mEAAmE;IACnE,MAAM,aAAa,GAAG,CACpB,YAAwC,EAAE,EACvB,EAAE;QACrB,MAAM,QAAQ,GAAsB;YAClC,UAAU,EAAE,UAAU;YACtB,QAAQ,EAAE,eAAe;YACzB,eAAe,EAAE,SAAS;YAC1B,cAAc,EAAE,uBAAuB;YACvC,oBAAoB,EAAE;gBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;gBACjC,SAAS,EAAE,uBAAuB;aACnC;SACF,CAAC;QAEF,OAAO;YACL,GAAG,QAAQ;YACX,GAAG,SAAS;YACZ,oBAAoB,EAAE;gBACpB,GAAG,QAAQ,CAAC,oBAAoB;gBAChC,GAAG,SAAS,CAAC,oBAAoB;aAClC;SACF,CAAC;IACJ,CAAC,CAAC;IAEF,QAAQ,CAAC,qBAAqB,EAAE,GAAG,EAAE;QACnC,EAAE,CAAC,kDAAkD,EAAE,KAAK,IAAI,EAAE;YAChE,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,cAAc,EAAE,uCAAuC;gBACvD,eAAe,EAAE,yBAAyB;aAC3C,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,CAAC,aAAa,CAAC;gBAC3B,UAAU,EAAE,UAAU;gBACtB,MAAM,EAAE,MAAM,CAAC,GAAG,CAAC,OAAO,CAAC;gBAC3B,cAAc,EAAE,MAAM,CAAC,GAAG,CAAC,KAAK,CAAC;gBACjC,SAAS,EAAE,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC;gBAC7B,oBAAoB,EAAE,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC;gBACxC,wBAAwB,EAAE,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC;aAC7C,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,gDAAgD,EAAE,KAAK,IAAI,EAAE;YAC9D,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,8BAA8B;gBAC/C,cAAc,EAAE,mCAAmC;gBACnD,oBAAoB,EAAE,EAAE,QAAQ,EAAE,kBAAkB,CAAC,IAAI,EAAE;aAC5D,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,oBAAoB,CAAC,SAAS,CAAC,CAAC,IAAI,CAChD,uBAAuB,CACxB,CAAC;QACJ,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,8CAA8C,EAAE,KAAK,IAAI,EAAE;YAC5D,MAAM,eAAe,GAAG,IAAI,CAAC;YAC7B,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,cAAc,EAAE,mCAAmC;gBACnD,eAAe,EAAE,6BAA6B;gBAC9C,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,eAAe;iBAC3B;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,oBAAoB,CAAC,SAAS,CAAC,CAAC,IAAI,CAAC,eAAe,CAAC,CAAC;QACtE,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;IAEH,QAAQ,CAAC,2BAA2B,EAAE,GAAG,EAAE;QACzC,EAAE,CAAC,wDAAwD,EAAE,KAAK,IAAI,EAAE;YACtE,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,sCAAsC;gBACvD,cAAc,EAAE,sCAAsC;aACvD,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,CAAC,aAAa,CAAC;gBAC3B,MAAM,EAAE,IAAI;gBACZ,cAAc,EAAE;oBACd;wBACE,OAAO,EAAE,sCAAsC;wBAC/C,KAAK,EAAE,IAAI;wBACX,wBAAwB,EAAE;4BACxB,KAAK,EAAE,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC;4BACzB,YAAY,EAAE,kBAAkB,CAAC,IAAI;yBACtC;qBACF;iBACF;aACF,CAAC,CAAC;YACH,MAAM,CACJ,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,KAAK,CACxD,CAAC,sBAAsB,CAAC,GAAG,CAAC,CAAC;QAChC,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,0EAA0E,EAAE,KAAK,IAAI,EAAE;YACxF,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,gBAAgB;gBACjC,cAAc,EAAE,wBAAwB;gBACxC,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,CAAC,aAAa,CAAC;gBAC7C,KAAK,EAAE,IAAI;gBACX,wBAAwB,EAAE;oBACxB,KAAK,EAAE,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC;oBACzB,YAAY,EAAE,kBAAkB,CAAC,IAAI;iBACtC;aACF,CAAC,CAAC;YACH,MAAM,CACJ,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,KAAK,CACxD,CAAC,eAAe,CAAC,CAAC,CAAC,CAAC;QACvB,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,iDAAiD,EAAE,KAAK,IAAI,EAAE;YAC/D,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,oCAAoC;gBACrD,cAAc,EAAE,2CAA2C;gBAC3D,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,CAAC,aAAa,CAAC;gBAC3B,MAAM,EAAE,KAAK;gBACb,cAAc,EAAE;oBACd;wBACE,KAAK,EAAE,KAAK;wBACZ,wBAAwB,EAAE;4BACxB,KAAK,EAAE,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC;yBAC1B;qBACF;iBACF;aACF,CAAC,CAAC;YACH,MAAM,CACJ,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,KAAK,CACxD,CAAC,YAAY,CAAC,GAAG,CAAC,CAAC;QACtB,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;IAEH,QAAQ,CAAC,mBAAmB,EAAE,GAAG,EAAE;QACjC,EAAE,CAAC,iDAAiD,EAAE,KAAK,IAAI,EAAE;YAC/D,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,kBAAkB;gBACnC,cAAc,EAAE,+BAA+B;gBAC/C,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,2FAA2F;YAC3F,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YAClC,MAAM,CACJ,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,KAAK,CACxD,CAAC,YAAY,CAAC,GAAG,CAAC,CAAC;QACtB,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,4CAA4C,EAAE,KAAK,IAAI,EAAE;YAC1D,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,gBAAgB;gBACjC,cAAc,EAAE,wBAAwB;gBACxC,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,wDAAwD;YACxD,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAClD,MAAM,CACJ,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,KAAK,CACxD,CAAC,eAAe,CAAC,CAAC,CAAC,CAAC;QACvB,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;IAEH,QAAQ,CAAC,8BAA8B,EAAE,GAAG,EAAE;QAC5C,EAAE,CAAC,8CAA8C,EAAE,KAAK,IAAI,EAAE;YAC5D,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EACb,uGAAuG;gBACzG,cAAc,EACZ,uGAAuG;gBACzG,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,IAAI;iBAChB;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,CAAC,aAAa,CAAC;gBAC3B,MAAM,EAAE,IAAI;gBACZ,wBAAwB,EAAE;oBACxB,KAAK,EAAE,CAAC;oBACR,YAAY,EAAE,kBAAkB,CAAC,IAAI;iBACtC;aACF,CAAC,CAAC;YACH,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC;YAC9C,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,KAAK,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACvE,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,kDAAkD,EAAE,KAAK,IAAI,EAAE;YAChE,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EACb,yGAAyG;gBAC3G,cAAc,EACZ,gFAAgF;gBAClF,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YAClC,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC;YAC9C,MAAM,UAAU,GAAG,MAAM,CAAC,cAAc,CAAC,MAAM,CAC7C,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,KAAK,CACrB,CAAC,MAAM,CAAC;YACT,MAAM,CAAC,UAAU,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC;YACnC,MAAM,CAAC,MAAM,CAAC,wBAAwB,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,UAAU,GAAG,CAAC,CAAC,CAAC;QACrE,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,4DAA4D,EAAE,KAAK,IAAI,EAAE;YAC1E,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EACb,wFAAwF;gBAC1F,cAAc,EACZ,4DAA4D;gBAC9D,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,UAAU,GAAG,MAAM,CAAC,cAAc,CAAC,MAAM,CAC7C,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,KAAK,CACrB,CAAC,MAAM,CAAC;YACT,MAAM,CAAC,MAAM,CAAC,CAAC,aAAa,CAAC;gBAC3B,MAAM,EAAE,KAAK;gBACb,wBAAwB,EAAE;oBACxB,KAAK,EAAE,UAAU,GAAG,CAAC;iBACtB;aACF,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;IAEH,QAAQ,CAAC,oBAAoB,EAAE,GAAG,EAAE;QAClC,EAAE,CAAC,6CAA6C,EAAE,KAAK,IAAI,EAAE;YAC3D,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,cAAc,EAAE,yCAAyC;gBACzD,eAAe,EACb,oEAAoE;gBACtE,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YACjC,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAC7D,MAAM,CAAC,MAAM,CAAC,oBAAoB,CAAC,SAAS,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;QAC1D,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,4DAA4D,EAAE,KAAK,IAAI,EAAE;YAC1E,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,cAAc,EAAE,yCAAyC;gBACzD,eAAe,EAAE,uCAAuC;gBACxD,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,oBAAoB,CAAC,SAAS,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;YACxD,MAAM,CACJ,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,KAAK,CACxD,CAAC,YAAY,CAAC,GAAG,CAAC,CAAC;YACpB,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACrD,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;IAEH,QAAQ,CAAC,YAAY,EAAE,GAAG,EAAE;QAC1B,EAAE,CAAC,oCAAoC,EAAE,KAAK,IAAI,EAAE;YAClD,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,EAAE;gBACnB,cAAc,EAAE,eAAe;aAChC,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,CAAC,aAAa,CAAC;gBAC3B,MAAM,EAAE,IAAI;gBACZ,cAAc,EAAE,EAAE;gBAClB,wBAAwB,EAAE;oBACxB,KAAK,EAAE,CAAC;iBACT;aACF,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,qCAAqC,EAAE,KAAK,IAAI,EAAE;YACnD,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,kCAAkC;gBACnD,cAAc,EAAE,EAAE;aACnB,CAAC,CAAC;YAEH,4BAA4B;YAC5B,MAAM,cAAc,GAAG,IAAI;iBACxB,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC;iBACtB,kBAAkB,CAAC,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;YAEhC,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,CAAC,aAAa,CAAC;gBAC3B,MAAM,EAAE,KAAK;gBACb,cAAc,EAAE;oBACd;wBACE,KAAK,EAAE,KAAK;wBACZ,wBAAwB,EAAE;4BACxB,KAAK,EAAE,CAAC;yBACT;qBACF;iBACF;aACF,CAAC,CAAC;YAEH,cAAc,CAAC,WAAW,EAAE,CAAC;QAC/B,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,uCAAuC,EAAE,KAAK,IAAI,EAAE;YACrD,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,KAAK;gBACtB,cAAc,EAAE,eAAe;aAChC,CAAC,CAAC;YAEH,4BAA4B;YAC5B,MAAM,cAAc,GAAG,IAAI;iBACxB,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC;iBACtB,kBAAkB,CAAC,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;YAEhC,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,CAAC,aAAa,CAAC;gBAC7C,KAAK,EAAE,KAAK;gBACZ,wBAAwB,EAAE;oBACxB,KAAK,EAAE,CAAC;iBACT;aACF,CAAC,CAAC;YAEH,cAAc,CAAC,WAAW,EAAE,CAAC;QAC/B,CAAC,CAAC,CAAC;QAEH,EAAE,CAAC,wDAAwD,EAAE,KAAK,IAAI,EAAE;YACtE,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,kCAAkC;gBACnD,cAAc,EAAE,IAAyB;aAC1C,CAAC,CAAC;YAEH,4BAA4B;YAC5B,MAAM,cAAc,GAAG,IAAI;iBACxB,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC;iBACtB,kBAAkB,CAAC,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC;YAEhC,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,CAAC,aAAa,CAAC;gBAC3B,MAAM,EAAE,KAAK;gBACb,cAAc,EAAE;oBACd;wBACE,KAAK,EAAE,KAAK;qBACb;iBACF;aACF,CAAC,CAAC;YAEH,cAAc,CAAC,WAAW,EAAE,CAAC;QAC/B,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;IAEH,QAAQ,CAAC,wBAAwB,EAAE,GAAG,EAAE;QACtC,EAAE,CAAC,+CAA+C,EAAE,KAAK,IAAI,EAAE;YAC7D,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,wBAAwB;gBACzC,cAAc,EAAE,gBAAgB;gBAChC,oBAAoB,EAAE;oBACpB,QAAQ,EAAE,kBAAkB,CAAC,IAAI;oBACjC,SAAS,EAAE,GAAG;iBACf;aACF,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,gEAAgE;YAChE,MAAM,KAAK,GAAG,MAAM,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,KAAK,CAAC;YACtE,MAAM,CAAC,KAAK,CAAC,CAAC,eAAe,CAAC,CAAC,CAAC,CAAC;YACjC,MAAM,CAAC,KAAK,CAAC,CAAC,YAAY,CAAC,GAAG,CAAC,CAAC;QAClC,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;IAEH,QAAQ,CAAC,WAAW,EAAE,GAAG,EAAE;QACzB,EAAE,CAAC,sCAAsC,EAAE,KAAK,IAAI,EAAE;YACpD,MAAM,OAAO,GAAG,aAAa,CAAC;gBAC5B,eAAe,EAAE,8BAA8B;gBAC/C,cAAc,EAAE,mCAAmC;aACpD,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,qBAAqB,CAAC,OAAO,CAAC,CAAC;YAE9C,MAAM,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC,WAAW,EAAE,CAAC;YACvC,MAAM,CAAC,IAAI,IAAI,CAAC,MAAM,CAAC,SAAU,CAAC,CAAC,WAAW,EAAE,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QAC3E,CAAC,CAAC,CAAC;IACL,CAAC,CAAC,CAAC;AACL,CAAC,CAAC,CAAC","sourcesContent":["import { jest, describe, it, expect } from '@jest/globals';\nimport { performBleuEvaluation } from '../bleu-evaluator';\nimport { EvaluationRequest } from '../../../types';\nimport {\n  DEFAULT_BLEU_PASS_SCORE,\n  EvaluationApproach,\n} from '../../../constants';\n\ndescribe('performBleuEvaluation', () => {\n  // Helper function to create a base request with optional overrides\n  const createRequest = (\n    overrides: Partial<EvaluationRequest> = {},\n  ): EvaluationRequest => {\n    const defaults: EvaluationRequest = {\n      testCaseId: 'test-001',\n      question: 'Test question',\n      expectedOutcome: 'keyword',\n      actualResponse: 'response with keyword',\n      evaluationParameters: {\n        approach: EvaluationApproach.BLEU,\n        threshold: DEFAULT_BLEU_PASS_SCORE,\n      },\n    };\n\n    return {\n      ...defaults,\n      ...overrides,\n      evaluationParameters: {\n        ...defaults.evaluationParameters,\n        ...overrides.evaluationParameters,\n      },\n    };\n  };\n\n  describe('basic functionality', () => {\n    it('should return a valid EvaluationResult structure', async () => {\n      const request = createRequest({\n        actualResponse: 'AI stands for artificial intelligence',\n        expectedOutcome: 'artificial intelligence',\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result).toMatchObject({\n        testCaseId: 'test-001',\n        passed: expect.any(Boolean),\n        keywordMatches: expect.any(Array),\n        timestamp: expect.any(String),\n        evaluationParameters: expect.any(Object),\n        evaluationApproachResult: expect.any(Object),\n      });\n    });\n\n    it('should use default threshold when not provided', async () => {\n      const request = createRequest({\n        expectedOutcome: 'test evaluation system works',\n        actualResponse: 'test evaluation system works well',\n        evaluationParameters: { approach: EvaluationApproach.BLEU },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result.evaluationParameters.threshold).toBe(\n        DEFAULT_BLEU_PASS_SCORE,\n      );\n    });\n\n    it('should use provided threshold when specified', async () => {\n      const customThreshold = 0.85;\n      const request = createRequest({\n        actualResponse: 'response text with multiple words',\n        expectedOutcome: 'response text with multiple',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: customThreshold,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result.evaluationParameters.threshold).toBe(customThreshold);\n    });\n  });\n\n  describe('single keyword evaluation', () => {\n    it('should pass when keyword with 4+ words matches exactly', async () => {\n      const request = createRequest({\n        expectedOutcome: 'the machine learning algorithm works',\n        actualResponse: 'the machine learning algorithm works',\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result).toMatchObject({\n        passed: true,\n        keywordMatches: [\n          {\n            keyword: 'the machine learning algorithm works',\n            found: true,\n            evaluationApproachResult: {\n              score: expect.any(Number),\n              approachUsed: EvaluationApproach.BLEU,\n            },\n          },\n        ],\n      });\n      expect(\n        result.keywordMatches[0].evaluationApproachResult.score,\n      ).toBeGreaterThanOrEqual(0.7);\n    });\n\n    it('should pass when keyword with 4+ words is found with high n-gram overlap', async () => {\n      const request = createRequest({\n        expectedOutcome: 'the cat sat on',\n        actualResponse: 'the cat sat on the mat',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.5,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result.keywordMatches[0]).toMatchObject({\n        found: true,\n        evaluationApproachResult: {\n          score: expect.any(Number),\n          approachUsed: EvaluationApproach.BLEU,\n        },\n      });\n      expect(\n        result.keywordMatches[0].evaluationApproachResult.score,\n      ).toBeGreaterThan(0);\n    });\n\n    it('should fail when keyword has low n-gram overlap', async () => {\n      const request = createRequest({\n        expectedOutcome: 'quantum physics research continues',\n        actualResponse: 'This is about machine learning algorithms',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.7,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result).toMatchObject({\n        passed: false,\n        keywordMatches: [\n          {\n            found: false,\n            evaluationApproachResult: {\n              score: expect.any(Number),\n            },\n          },\n        ],\n      });\n      expect(\n        result.keywordMatches[0].evaluationApproachResult.score,\n      ).toBeLessThan(0.7);\n    });\n  });\n\n  describe('4-gram limitation', () => {\n    it('should fail when keyword has fewer than 4 words', async () => {\n      const request = createRequest({\n        expectedOutcome: 'machine learning',\n        actualResponse: 'machine learning is important',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.7,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      // BLEU uses 4-gram matching, so keywords with fewer than 4 words will have very low scores\n      expect(result.passed).toBe(false);\n      expect(\n        result.keywordMatches[0].evaluationApproachResult.score,\n      ).toBeLessThan(0.7);\n    });\n\n    it('should work correctly with 4-word keywords', async () => {\n      const request = createRequest({\n        expectedOutcome: 'the cat sat on',\n        actualResponse: 'the cat sat on the mat',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.5,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      // 4-word keywords can produce proper 4-gram BLEU scores\n      expect(result.keywordMatches[0].found).toBe(true);\n      expect(\n        result.keywordMatches[0].evaluationApproachResult.score,\n      ).toBeGreaterThan(0);\n    });\n  });\n\n  describe('multiple keywords evaluation', () => {\n    it('should pass when all keywords meet threshold', async () => {\n      const request = createRequest({\n        expectedOutcome:\n          'Machine learning algorithms work. Artificial intelligence systems are. Neural network models perform.',\n        actualResponse:\n          'Machine learning algorithms work. Artificial intelligence systems are. Neural network models perform.',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.15,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result).toMatchObject({\n        passed: true,\n        evaluationApproachResult: {\n          score: 1,\n          approachUsed: EvaluationApproach.BLEU,\n        },\n      });\n      expect(result.keywordMatches).toHaveLength(3);\n      expect(result.keywordMatches.every(match => match.found)).toBe(true);\n    });\n\n    it('should fail when not all keywords meet threshold', async () => {\n      const request = createRequest({\n        expectedOutcome:\n          'machine learning algorithms work\\nquantum computing research continues\\nartificial intelligence systems',\n        actualResponse:\n          'Machine learning algorithms work. Artificial intelligence systems are growing.',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.7,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result.passed).toBe(false);\n      expect(result.keywordMatches).toHaveLength(3);\n      const foundCount = result.keywordMatches.filter(\n        match => match.found,\n      ).length;\n      expect(foundCount).toBeLessThan(3);\n      expect(result.evaluationApproachResult.score).toBe(foundCount / 3);\n    });\n\n    it('should calculate overall score as ratio of passed keywords', async () => {\n      const request = createRequest({\n        expectedOutcome:\n          'alpha beta gamma delta\\nepsilon zeta eta theta\\niota kappa lambda mu\\nnu xi omicron pi',\n        actualResponse:\n          'alpha beta gamma delta and epsilon zeta eta theta are here',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.3,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      const foundCount = result.keywordMatches.filter(\n        match => match.found,\n      ).length;\n      expect(result).toMatchObject({\n        passed: false,\n        evaluationApproachResult: {\n          score: foundCount / 4,\n        },\n      });\n    });\n  });\n\n  describe('threshold handling', () => {\n    it('should pass all keywords with threshold 0.0', async () => {\n      const request = createRequest({\n        actualResponse: 'completely unrelated text about cooking',\n        expectedOutcome:\n          'quantum physics research continues\\nmathematics and statistics are',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.0,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result.passed).toBe(true);\n      expect(result.keywordMatches.every(m => m.found)).toBe(true);\n      expect(result.evaluationParameters.threshold).toBe(0.0);\n    });\n\n    it('should fail when threshold is 1.0 and match is not perfect', async () => {\n      const request = createRequest({\n        actualResponse: 'This is about machine learning concepts',\n        expectedOutcome: 'machine learning algorithms work well',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 1.0,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result.evaluationParameters.threshold).toBe(1.0);\n      expect(\n        result.keywordMatches[0].evaluationApproachResult.score,\n      ).toBeLessThan(1.0);\n      expect(result.keywordMatches[0].found).toBe(false);\n    });\n  });\n\n  describe('edge cases', () => {\n    it('should handle empty keywords array', async () => {\n      const request = createRequest({\n        expectedOutcome: '',\n        actualResponse: 'Some response',\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result).toMatchObject({\n        passed: true,\n        keywordMatches: [],\n        evaluationApproachResult: {\n          score: 1,\n        },\n      });\n    });\n\n    it('should handle empty actual response', async () => {\n      const request = createRequest({\n        expectedOutcome: 'machine learning algorithms work',\n        actualResponse: '',\n      });\n\n      // Suppress expected warning\n      const consoleWarnSpy = jest\n        .spyOn(console, 'warn')\n        .mockImplementation(() => {});\n\n      const result = performBleuEvaluation(request);\n\n      expect(result).toMatchObject({\n        passed: false,\n        keywordMatches: [\n          {\n            found: false,\n            evaluationApproachResult: {\n              score: 0,\n            },\n          },\n        ],\n      });\n\n      consoleWarnSpy.mockRestore();\n    });\n\n    it('should handle whitespace-only keyword', async () => {\n      const request = createRequest({\n        expectedOutcome: '   ',\n        actualResponse: 'Some response',\n      });\n\n      // Suppress expected warning\n      const consoleWarnSpy = jest\n        .spyOn(console, 'warn')\n        .mockImplementation(() => {});\n\n      const result = performBleuEvaluation(request);\n\n      expect(result.keywordMatches[0]).toMatchObject({\n        found: false,\n        evaluationApproachResult: {\n          score: 0,\n        },\n      });\n\n      consoleWarnSpy.mockRestore();\n    });\n\n    it('should handle null/undefined actualResponse gracefully', async () => {\n      const request = createRequest({\n        expectedOutcome: 'machine learning algorithms work',\n        actualResponse: null as unknown as string,\n      });\n\n      // Suppress expected warning\n      const consoleWarnSpy = jest\n        .spyOn(console, 'warn')\n        .mockImplementation(() => {});\n\n      const result = performBleuEvaluation(request);\n\n      expect(result).toMatchObject({\n        passed: false,\n        keywordMatches: [\n          {\n            found: false,\n          },\n        ],\n      });\n\n      consoleWarnSpy.mockRestore();\n    });\n  });\n\n  describe('BLEU score calculation', () => {\n    it('should calculate BLEU score for partial match', async () => {\n      const request = createRequest({\n        expectedOutcome: 'the cat sat on the mat',\n        actualResponse: 'the cat sat on',\n        evaluationParameters: {\n          approach: EvaluationApproach.BLEU,\n          threshold: 0.3,\n        },\n      });\n\n      const result = performBleuEvaluation(request);\n\n      // Partial match should have lower BLEU score than perfect match\n      const score = result.keywordMatches[0].evaluationApproachResult.score;\n      expect(score).toBeGreaterThan(0);\n      expect(score).toBeLessThan(1.0);\n    });\n  });\n\n  describe('timestamp', () => {\n    it('should include a valid ISO timestamp', async () => {\n      const request = createRequest({\n        expectedOutcome: 'test evaluation system works',\n        actualResponse: 'test evaluation system works well',\n      });\n\n      const result = performBleuEvaluation(request);\n\n      expect(result.timestamp).toBeDefined();\n      expect(new Date(result.timestamp!).toISOString()).toBe(result.timestamp);\n    });\n  });\n});\n"]}

package/dist/collection/lib/evaluation/evaluators/exact/exact.js

@@ -0,0 +1,44 @@
+import { EvaluationApproach } from "../../constants";
+export async function performEvaluation(request) {
+    const { testCaseId, expectedOutcome, actualResponse } = request;
+    // Split expectedOutcome by newlines to create keywords array
+    const expectedKeywords = expectedOutcome
+        ? expectedOutcome
+            .split(/[\n,]+/)
+            .map(k => k.trim())
+            .filter(k => k.length > 0)
+        : [];
+    const keywordMatches = evaluateKeywords(expectedKeywords, actualResponse);
+    // Test passes only if ALL expected keywords are found
+    const totalItems = keywordMatches.length;
+    const foundItems = keywordMatches.filter(m => m.found).length;
+    const passed = foundItems === totalItems;
+    return {
+        testCaseId,
+        passed,
+        keywordMatches,
+        timestamp: new Date().toISOString(),
+        evaluationParameters: request.evaluationParameters,
+        evaluationApproachResult: {
+            score: totalItems > 0 ? foundItems / totalItems : 1,
+            approachUsed: EvaluationApproach.EXACT,
+        },
+    };
+}
+function evaluateKeywords(expectedKeywords, actualResponse) {
+    // Case-insensitive keyword matching
+    const response = actualResponse.toLowerCase();
+    return expectedKeywords.map(keyword => {
+        const keywordToMatch = keyword.toLowerCase();
+        const found = response.includes(keywordToMatch);
+        return {
+            keyword,
+            found,
+            evaluationApproachResult: {
+                score: found ? 1.0 : 0.0,
+                approachUsed: EvaluationApproach.EXACT,
+            },
+        };
+    });
+}
+//# sourceMappingURL=exact.js.map

package/dist/collection/lib/evaluation/evaluators/exact/exact.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"exact.js","sourceRoot":"","sources":["../../../../../src/lib/evaluation/evaluators/exact/exact.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,kBAAkB,EAAE,MAAM,iBAAiB,CAAC;AAErD,MAAM,CAAC,KAAK,UAAU,iBAAiB,CACrC,OAA0B;IAE1B,MAAM,EAAE,UAAU,EAAE,eAAe,EAAE,cAAc,EAAE,GAAG,OAAO,CAAC;IAEhE,6DAA6D;IAC7D,MAAM,gBAAgB,GAAG,eAAe;QACtC,CAAC,CAAC,eAAe;aACZ,KAAK,CAAC,QAAQ,CAAC;aACf,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;aAClB,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC;QAC9B,CAAC,CAAC,EAAE,CAAC;IAEP,MAAM,cAAc,GAAG,gBAAgB,CAAC,gBAAgB,EAAE,cAAc,CAAC,CAAC;IAE1E,sDAAsD;IACtD,MAAM,UAAU,GAAG,cAAc,CAAC,MAAM,CAAC;IACzC,MAAM,UAAU,GAAG,cAAc,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC;IAC9D,MAAM,MAAM,GAAG,UAAU,KAAK,UAAU,CAAC;IAEzC,OAAO;QACL,UAAU;QACV,MAAM;QACN,cAAc;QACd,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;QACnC,oBAAoB,EAAE,OAAO,CAAC,oBAAoB;QAClD,wBAAwB,EAAE;YACxB,KAAK,EAAE,UAAU,GAAG,CAAC,CAAC,CAAC,CAAC,UAAU,GAAG,UAAU,CAAC,CAAC,CAAC,CAAC;YACnD,YAAY,EAAE,kBAAkB,CAAC,KAAK;SACvC;KACF,CAAC;AACJ,CAAC;AAED,SAAS,gBAAgB,CACvB,gBAA0B,EAC1B,cAAsB;IAEtB,oCAAoC;IACpC,MAAM,QAAQ,GAAG,cAAc,CAAC,WAAW,EAAE,CAAC;IAE9C,OAAO,gBAAgB,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE;QACpC,MAAM,cAAc,GAAG,OAAO,CAAC,WAAW,EAAE,CAAC;QAC7C,MAAM,KAAK,GAAG,QAAQ,CAAC,QAAQ,CAAC,cAAc,CAAC,CAAC;QAEhD,OAAO;YACL,OAAO;YACP,KAAK;YACL,wBAAwB,EAAE;gBACxB,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,GAAG;gBACxB,YAAY,EAAE,kBAAkB,CAAC,KAAK;aACvC;SACF,CAAC;IACJ,CAAC,CAAC,CAAC;AACL,CAAC","sourcesContent":["import { EvaluationRequest, EvaluationResult, KeywordMatch } from '../../types';\nimport { EvaluationApproach } from '../../constants';\n\nexport async function performEvaluation(\n  request: EvaluationRequest,\n): Promise<EvaluationResult> {\n  const { testCaseId, expectedOutcome, actualResponse } = request;\n\n  // Split expectedOutcome by newlines to create keywords array\n  const expectedKeywords = expectedOutcome\n    ? expectedOutcome\n        .split(/[\\n,]+/)\n        .map(k => k.trim())\n        .filter(k => k.length > 0)\n    : [];\n\n  const keywordMatches = evaluateKeywords(expectedKeywords, actualResponse);\n\n  // Test passes only if ALL expected keywords are found\n  const totalItems = keywordMatches.length;\n  const foundItems = keywordMatches.filter(m => m.found).length;\n  const passed = foundItems === totalItems;\n\n  return {\n    testCaseId,\n    passed,\n    keywordMatches,\n    timestamp: new Date().toISOString(),\n    evaluationParameters: request.evaluationParameters,\n    evaluationApproachResult: {\n      score: totalItems > 0 ? foundItems / totalItems : 1,\n      approachUsed: EvaluationApproach.EXACT,\n    },\n  };\n}\n\nfunction evaluateKeywords(\n  expectedKeywords: string[],\n  actualResponse: string,\n): KeywordMatch[] {\n  // Case-insensitive keyword matching\n  const response = actualResponse.toLowerCase();\n\n  return expectedKeywords.map(keyword => {\n    const keywordToMatch = keyword.toLowerCase();\n    const found = response.includes(keywordToMatch);\n\n    return {\n      keyword,\n      found,\n      evaluationApproachResult: {\n        score: found ? 1.0 : 0.0,\n        approachUsed: EvaluationApproach.EXACT,\n      },\n    };\n  });\n}\n"]}

package/dist/collection/lib/evaluation/evaluators/rouge1-evaluator.js

@@ -0,0 +1,88 @@
+import * as rouge from "js-rouge";
+import { DEFAULT_ROUGE_PASS_SCORE, EvaluationApproach } from "../constants";
+/**
+ * Evaluates a single keyword against the candidate text using ROUGE-1.
+ *
+ * @param {string} keyword - The expected keyword to evaluate
+ * @param {string} candidate - The actual response text
+ * @param {number} rougeThreshold - The minimum ROUGE-1 score required to pass
+ * @returns {KeywordMatch} The evaluation result for this keyword
+ */
+function evaluateKeyword(keyword, candidate, rougeThreshold) {
+    let rouge1Score = 0;
+    try {
+        if (keyword.trim().length > 0 && candidate.length > 0) {
+            const rouge1 = rouge.n(candidate, keyword.trim(), { n: 1 });
+            rouge1Score = isNaN(rouge1) ? 0 : rouge1;
+        }
+        else {
+            console.warn(`ROUGE-1 not computed for keyword "${keyword}": Keyword or Candidate is missing.`);
+        }
+    }
+    catch (err) {
+        console.error(`ROUGE-1 computation failed for keyword "${keyword}":`, err);
+    }
+    const keywordPassed = rouge1Score >= rougeThreshold;
+    const keywordApproachResult = {
+        score: rouge1Score,
+        approachUsed: EvaluationApproach.ROUGE_1,
+    };
+    return {
+        keyword: keyword,
+        found: keywordPassed,
+        evaluationApproachResult: keywordApproachResult,
+    };
+}
+/**
+ * Computes the ROUGE-1 score for a single keyword against the candidate text.
+ *
+ * ROUGE-1 measures the overlap of unigrams (single words) between the candidate
+ * and reference text. A score of 1.0 indicates perfect overlap.
+ *
+ * @example
+ * const match = evaluateSingleKeyword(
+ *   "The quick brown fox",
+ *   "quick fox",
+ *   0.5
+ * );
+ * // Returns: { keyword: "quick fox", found: true, score: 0.67, ... }
+ * //general idea , here we are doing it. by word to word comparison
+ */
+export async function performRouge1Evaluation(request) {
+    const { testCaseId, actualResponse, expectedOutcome, evaluationParameters } = request;
+    // Split expectedOutcome by newlines to create keywords array
+    const expectedKeywords = expectedOutcome
+        ? expectedOutcome
+            .split(/[\n,]+/)
+            .map(k => k.trim())
+            .filter(k => k.length > 0)
+        : [];
+    const candidate = (actualResponse || '').trim();
+    const rougeThreshold = evaluationParameters.threshold ?? DEFAULT_ROUGE_PASS_SCORE;
+    let keywordsPassed = 0;
+    const totalKeywords = expectedKeywords.length;
+    const keywordMatches = expectedKeywords.map(keyword => {
+        const match = evaluateKeyword(keyword, candidate, rougeThreshold);
+        if (match.found) {
+            keywordsPassed++;
+        }
+        return match;
+    });
+    const overallPassed = keywordsPassed === totalKeywords;
+    const overallApproachResult = {
+        score: keywordsPassed / totalKeywords,
+        approachUsed: EvaluationApproach.ROUGE_1,
+    };
+    return {
+        testCaseId: testCaseId,
+        passed: overallPassed,
+        keywordMatches: keywordMatches,
+        timestamp: new Date().toISOString(),
+        evaluationParameters: {
+            ...evaluationParameters,
+            threshold: rougeThreshold,
+        },
+        evaluationApproachResult: overallApproachResult,
+    };
+}
+//# sourceMappingURL=rouge1-evaluator.js.map

package/dist/collection/lib/evaluation/evaluators/rouge1-evaluator.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"rouge1-evaluator.js","sourceRoot":"","sources":["../../../../src/lib/evaluation/evaluators/rouge1-evaluator.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,KAAK,MAAM,UAAU,CAAC;AAGlC,OAAO,EAAE,wBAAwB,EAAE,kBAAkB,EAAE,MAAM,cAAc,CAAC;AAE5E;;;;;;;GAOG;AACH,SAAS,eAAe,CACtB,OAAe,EACf,SAAiB,EACjB,cAAsB;IAEtB,IAAI,WAAW,GAAG,CAAC,CAAC;IAEpB,IAAI,CAAC;QACH,IAAI,OAAO,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC,IAAI,SAAS,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YACtD,MAAM,MAAM,GAAG,KAAK,CAAC,CAAC,CAAC,SAAS,EAAE,OAAO,CAAC,IAAI,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC;YAC5D,WAAW,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC;QAC3C,CAAC;aAAM,CAAC;YACN,OAAO,CAAC,IAAI,CACV,qCAAqC,OAAO,qCAAqC,CAClF,CAAC;QACJ,CAAC;IACH,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,OAAO,CAAC,KAAK,CAAC,2CAA2C,OAAO,IAAI,EAAE,GAAG,CAAC,CAAC;IAC7E,CAAC;IAED,MAAM,aAAa,GAAG,WAAW,IAAI,cAAc,CAAC;IAEpD,MAAM,qBAAqB,GAA6B;QACtD,KAAK,EAAE,WAAW;QAClB,YAAY,EAAE,kBAAkB,CAAC,OAAO;KACzC,CAAC;IAEF,OAAO;QACL,OAAO,EAAE,OAAO;QAChB,KAAK,EAAE,aAAa;QACpB,wBAAwB,EAAE,qBAAqB;KAChD,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;GAcG;AAEH,MAAM,CAAC,KAAK,UAAU,uBAAuB,CAC3C,OAA0B;IAE1B,MAAM,EAAE,UAAU,EAAE,cAAc,EAAE,eAAe,EAAE,oBAAoB,EAAE,GACzE,OAAO,CAAC;IAEV,6DAA6D;IAC7D,MAAM,gBAAgB,GAAG,eAAe;QACtC,CAAC,CAAC,eAAe;aACZ,KAAK,CAAC,QAAQ,CAAC;aACf,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;aAClB,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC;QAC9B,CAAC,CAAC,EAAE,CAAC;IAEP,MAAM,SAAS,GAAG,CAAC,cAAc,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;IAChD,MAAM,cAAc,GAClB,oBAAoB,CAAC,SAAS,IAAI,wBAAwB,CAAC;IAE7D,IAAI,cAAc,GAAG,CAAC,CAAC;IACvB,MAAM,aAAa,GAAG,gBAAgB,CAAC,MAAM,CAAC;IAE9C,MAAM,cAAc,GAAmB,gBAAgB,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE;QACpE,MAAM,KAAK,GAAG,eAAe,CAAC,OAAO,EAAE,SAAS,EAAE,cAAc,CAAC,CAAC;QAElE,IAAI,KAAK,CAAC,KAAK,EAAE,CAAC;YAChB,cAAc,EAAE,CAAC;QACnB,CAAC;QAED,OAAO,KAAK,CAAC;IACf,CAAC,CAAC,CAAC;IAEH,MAAM,aAAa,GAAG,cAAc,KAAK,aAAa,CAAC;IAEvD,MAAM,qBAAqB,GAA6B;QACtD,KAAK,EAAE,cAAc,GAAG,aAAa;QACrC,YAAY,EAAE,kBAAkB,CAAC,OAAO;KACzC,CAAC;IAEF,OAAO;QACL,UAAU,EAAE,UAAU;QACtB,MAAM,EAAE,aAAa;QACrB,cAAc,EAAE,cAAc;QAC9B,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;QACnC,oBAAoB,EAAE;YACpB,GAAG,oBAAoB;YACvB,SAAS,EAAE,cAAc;SAC1B;QACD,wBAAwB,EAAE,qBAAqB;KAChD,CAAC;AACJ,CAAC","sourcesContent":["import * as rouge from 'js-rouge';\nimport { EvaluationApproachResult } from '../../../types/evaluation';\nimport { EvaluationRequest, EvaluationResult, KeywordMatch } from '../types';\nimport { DEFAULT_ROUGE_PASS_SCORE, EvaluationApproach } from '../constants';\n\n/**\n * Evaluates a single keyword against the candidate text using ROUGE-1.\n *\n * @param {string} keyword - The expected keyword to evaluate\n * @param {string} candidate - The actual response text\n * @param {number} rougeThreshold - The minimum ROUGE-1 score required to pass\n * @returns {KeywordMatch} The evaluation result for this keyword\n */\nfunction evaluateKeyword(\n  keyword: string,\n  candidate: string,\n  rougeThreshold: number,\n): KeywordMatch {\n  let rouge1Score = 0;\n\n  try {\n    if (keyword.trim().length > 0 && candidate.length > 0) {\n      const rouge1 = rouge.n(candidate, keyword.trim(), { n: 1 });\n      rouge1Score = isNaN(rouge1) ? 0 : rouge1;\n    } else {\n      console.warn(\n        `ROUGE-1 not computed for keyword \"${keyword}\": Keyword or Candidate is missing.`,\n      );\n    }\n  } catch (err) {\n    console.error(`ROUGE-1 computation failed for keyword \"${keyword}\":`, err);\n  }\n\n  const keywordPassed = rouge1Score >= rougeThreshold;\n\n  const keywordApproachResult: EvaluationApproachResult = {\n    score: rouge1Score,\n    approachUsed: EvaluationApproach.ROUGE_1,\n  };\n\n  return {\n    keyword: keyword,\n    found: keywordPassed,\n    evaluationApproachResult: keywordApproachResult,\n  };\n}\n\n/**\n * Computes the ROUGE-1 score for a single keyword against the candidate text.\n *\n * ROUGE-1 measures the overlap of unigrams (single words) between the candidate\n * and reference text. A score of 1.0 indicates perfect overlap.\n *\n * @example\n * const match = evaluateSingleKeyword(\n *   \"The quick brown fox\",\n *   \"quick fox\",\n *   0.5\n * );\n * // Returns: { keyword: \"quick fox\", found: true, score: 0.67, ... }\n * //general idea , here we are doing it. by word to word comparison\n */\n\nexport async function performRouge1Evaluation(\n  request: EvaluationRequest,\n): Promise<EvaluationResult> {\n  const { testCaseId, actualResponse, expectedOutcome, evaluationParameters } =\n    request;\n\n  // Split expectedOutcome by newlines to create keywords array\n  const expectedKeywords = expectedOutcome\n    ? expectedOutcome\n        .split(/[\\n,]+/)\n        .map(k => k.trim())\n        .filter(k => k.length > 0)\n    : [];\n\n  const candidate = (actualResponse || '').trim();\n  const rougeThreshold =\n    evaluationParameters.threshold ?? DEFAULT_ROUGE_PASS_SCORE;\n\n  let keywordsPassed = 0;\n  const totalKeywords = expectedKeywords.length;\n\n  const keywordMatches: KeywordMatch[] = expectedKeywords.map(keyword => {\n    const match = evaluateKeyword(keyword, candidate, rougeThreshold);\n\n    if (match.found) {\n      keywordsPassed++;\n    }\n\n    return match;\n  });\n\n  const overallPassed = keywordsPassed === totalKeywords;\n\n  const overallApproachResult: EvaluationApproachResult = {\n    score: keywordsPassed / totalKeywords,\n    approachUsed: EvaluationApproach.ROUGE_1,\n  };\n\n  return {\n    testCaseId: testCaseId,\n    passed: overallPassed,\n    keywordMatches: keywordMatches,\n    timestamp: new Date().toISOString(),\n    evaluationParameters: {\n      ...evaluationParameters,\n      threshold: rougeThreshold,\n    },\n    evaluationApproachResult: overallApproachResult,\n  };\n}\n"]}

package/dist/collection/lib/evaluation/evaluators/rougeL-evaluator.js

@@ -0,0 +1,82 @@
+import * as rouge from "js-rouge";
+import { DEFAULT_ROUGE_PASS_SCORE, EvaluationApproach } from "../constants";
+function evaluateKeyword(keyword, candidate, rougeThreshold) {
+    let rougeLScore = 0;
+    try {
+        const trimmedKeyword = keyword.trim();
+        if (trimmedKeyword.length > 0 && candidate.length > 0) {
+            const referenceTokens = trimmedKeyword.toLowerCase().split(/\s+/);
+            const candidateTokens = candidate.toLowerCase().split(/\s+/);
+            if (referenceTokens.length === 1 &&
+                candidateTokens.includes(referenceTokens[0])) {
+                rougeLScore = 1;
+            }
+            else {
+                const lcsResult = rouge.lcs(candidateTokens, referenceTokens);
+                const lcsLength = typeof lcsResult === 'number' ? lcsResult : (lcsResult?.length ?? 0);
+                const recall = referenceTokens.length > 0 ? lcsLength / referenceTokens.length : 0;
+                const precision = candidateTokens.length > 0 ? lcsLength / candidateTokens.length : 0;
+                const denominator = precision + recall;
+                const f1Score = denominator > 0 ? (2 * precision * recall) / denominator : 0;
+                rougeLScore = f1Score;
+            }
+        }
+        else {
+            console.warn(`ROUGE-L not computed for keyword "${keyword}": Keyword or candidate missing.`);
+        }
+    }
+    catch (err) {
+        console.error(`ROUGE-L computation failed for keyword "${keyword}":`, err);
+    }
+    const keywordPassed = rougeLScore >= rougeThreshold;
+    const keywordApproachResult = {
+        score: rougeLScore,
+        approachUsed: EvaluationApproach.ROUGE_L,
+    };
+    return {
+        keyword,
+        found: keywordPassed,
+        evaluationApproachResult: keywordApproachResult,
+    };
+}
+export function performRougeLEvaluation(request) {
+    const { testCaseId, actualResponse, expectedOutcome, evaluationParameters } = request;
+    // Split expectedOutcome by newlines, commas, and periods to create keywords array
+    let expectedKeywords = expectedOutcome
+        ? expectedOutcome
+            .split(/[\n,.]+/)
+            .map(k => k.trim())
+            .filter(k => k.length > 0)
+        : [];
+    // If no keywords after filtering (e.g., whitespace-only input), treat the original input as a single keyword
+    if (expectedKeywords.length === 0 && expectedOutcome) {
+        expectedKeywords = [expectedOutcome];
+    }
+    const candidate = (actualResponse || '').trim();
+    const rougeThreshold = evaluationParameters.threshold ?? DEFAULT_ROUGE_PASS_SCORE;
+    let keywordsPassed = 0;
+    const totalKeywords = expectedKeywords.length;
+    const keywordMatches = expectedKeywords.map(keyword => {
+        const match = evaluateKeyword(keyword, candidate, rougeThreshold);
+        if (match.found)
+            keywordsPassed++;
+        return match;
+    });
+    const overallPassed = keywordsPassed === totalKeywords;
+    const overallApproachResult = {
+        score: totalKeywords > 0 ? keywordsPassed / totalKeywords : 1,
+        approachUsed: EvaluationApproach.ROUGE_L,
+    };
+    return {
+        testCaseId,
+        passed: overallPassed,
+        keywordMatches,
+        timestamp: new Date().toISOString(),
+        evaluationParameters: {
+            ...evaluationParameters,
+            threshold: rougeThreshold,
+        },
+        evaluationApproachResult: overallApproachResult,
+    };
+}
+//# sourceMappingURL=rougeL-evaluator.js.map