|
| 1 | +import { describe, expect, test } from "@jest/globals"; |
| 2 | +import { |
| 3 | + DEFAULT_CHROMA_BM25_STOPWORDS, |
| 4 | + ChromaBm25EmbeddingFunction, |
| 5 | + type ChromaBm25Config, |
| 6 | +} from "./index"; |
| 7 | + |
| 8 | +const isSorted = (arr: number[]): boolean => { |
| 9 | + for (let i = 1; i < arr.length; i += 1) { |
| 10 | + if (arr[i] < arr[i - 1]) { |
| 11 | + return false; |
| 12 | + } |
| 13 | + } |
| 14 | + return true; |
| 15 | +}; |
| 16 | + |
| 17 | +describe("ChromaBm25EmbeddingFunction", () => { |
| 18 | + const embedder = new ChromaBm25EmbeddingFunction(); |
| 19 | + |
| 20 | + test("matches comprehensive tokenization expectations", async () => { |
| 21 | + const [embedding] = await embedder.generate([ |
| 22 | + "Usain Bolt's top speed reached ~27.8 mph (44.72 km/h)", |
| 23 | + ]); |
| 24 | + |
| 25 | + const expectedIndices = [ |
| 26 | + 230246813, 395514983, 458027949, 488165615, 729632045, 734978415, |
| 27 | + 997512866, 1114505193, 1381820790, 1501587190, 1649421877, |
| 28 | + 1837285388, |
| 29 | + ]; |
| 30 | + const expectedValue = 1.6391153; |
| 31 | + |
| 32 | + expect(embedding.indices).toEqual(expectedIndices); |
| 33 | + embedding.values.forEach((value) => { |
| 34 | + expect(value).toBeCloseTo(expectedValue, 5); |
| 35 | + }); |
| 36 | + }); |
| 37 | + |
| 38 | + // mirrors rust test `test_bm25_stopwords_and_punctuation` to guarantee compatibility |
| 39 | + test("ensure Rust impl compatibilty", async () => { |
| 40 | + const [embedding] = await embedder.generate([ |
| 41 | + "The space-time continuum WARPS near massive objects...", |
| 42 | + ]); |
| 43 | + |
| 44 | + const expectedIndices = [ |
| 45 | + 90097469, 519064992, 737893654, 1110755108, 1950894484, 2031641008, |
| 46 | + 2058513491, |
| 47 | + ]; |
| 48 | + const expectedValue = 1.660867; |
| 49 | + |
| 50 | + expect(embedding.indices).toEqual(expectedIndices); |
| 51 | + embedding.values.forEach((value) => { |
| 52 | + expect(value).toBeCloseTo(expectedValue, 5); |
| 53 | + }); |
| 54 | + }); |
| 55 | + |
| 56 | + test("generates consistent embeddings for multiple documents", async () => { |
| 57 | + const texts = [ |
| 58 | + "Usain Bolt's top speed reached ~27.8 mph (44.72 km/h)", |
| 59 | + "The space-time continuum WARPS near massive objects...", |
| 60 | + "BM25 is great for sparse retrieval tasks", |
| 61 | + ]; |
| 62 | + |
| 63 | + const embeddings = await embedder.generate(texts); |
| 64 | + |
| 65 | + expect(embeddings).toHaveLength(texts.length); |
| 66 | + embeddings.forEach((embedding, index) => { |
| 67 | + expect(embedding.indices.length).toBeGreaterThan(0); |
| 68 | + expect(embedding.values.length).toBe(embedding.indices.length); |
| 69 | + expect(isSorted(embedding.indices)).toBe(true); |
| 70 | + |
| 71 | + embedding.values.forEach((value) => { |
| 72 | + expect(value).toBeGreaterThan(0); |
| 73 | + expect(Number.isFinite(value)).toBe(true); |
| 74 | + }); |
| 75 | + }); |
| 76 | + }); |
| 77 | + |
| 78 | + test("generateForQueries mirrors generate", async () => { |
| 79 | + const query = "retrieve BM25 docs"; |
| 80 | + const [queryEmbedding] = await embedder.generateForQueries([query]); |
| 81 | + const [docEmbedding] = await embedder.generate([query]); |
| 82 | + |
| 83 | + expect(queryEmbedding.indices).toEqual(docEmbedding.indices); |
| 84 | + expect(queryEmbedding.values).toEqual(docEmbedding.values); |
| 85 | + }); |
| 86 | + |
| 87 | + test("config round trip maintains settings", () => { |
| 88 | + const config = embedder.getConfig() as Required<ChromaBm25Config>; |
| 89 | + |
| 90 | + expect(config).toMatchObject({ |
| 91 | + k: 1.2, |
| 92 | + b: 0.75, |
| 93 | + avg_doc_length: 256, |
| 94 | + token_max_length: 40, |
| 95 | + }); |
| 96 | + expect(config.stopwords).toBeUndefined(); |
| 97 | + |
| 98 | + const custom = ChromaBm25EmbeddingFunction.buildFromConfig({ |
| 99 | + ...config, |
| 100 | + stopwords: DEFAULT_CHROMA_BM25_STOPWORDS.slice(0, 10), |
| 101 | + }); |
| 102 | + |
| 103 | + const rebuiltConfig = |
| 104 | + custom.getConfig() as Required<ChromaBm25Config>; |
| 105 | + expect(rebuiltConfig.k).toBeCloseTo(config.k); |
| 106 | + expect(rebuiltConfig.b).toBeCloseTo(config.b); |
| 107 | + expect(rebuiltConfig.avg_doc_length).toBeCloseTo(config.avg_doc_length); |
| 108 | + expect(rebuiltConfig.token_max_length).toBe(config.token_max_length); |
| 109 | + expect(rebuiltConfig.stopwords).toEqual( |
| 110 | + DEFAULT_CHROMA_BM25_STOPWORDS.slice(0, 10), |
| 111 | + ); |
| 112 | + }); |
| 113 | +}); |
0 commit comments