nitrite_vector 0.7.0

Vector / HNSW ANN indexing and RAG store for the Nitrite database
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138

//! RAG store: add/search/filter/min_score/delete over text + embedding + metadata.

mod common;

use common::temp_db;
use nitrite::doc;
use nitrite::filter::field;
use nitrite::nitrite::Nitrite;
use nitrite_vector::{Metric, RagStore, VectorIndexConfig};

// Keep the `Nitrite` handle alive: dropping it closes the underlying store.
fn store() -> (tempfile::TempDir, Nitrite, RagStore) {
    let (dir, db) = temp_db(VectorIndexConfig::new(3, Metric::Cosine));
    let store = RagStore::create(&db, "kb", Metric::Cosine).unwrap();
    (dir, db, store)
}

#[test]
fn search_returns_scored_hits_in_order() {
    let (_dir, _db, store) = store();
    store.add("apple", vec![1.0, 0.0, 0.0], doc! {"source": "fruit"}).unwrap();
    store.add("banana", vec![0.9, 0.1, 0.0], doc! {"source": "fruit"}).unwrap();
    store.add("car", vec![0.0, 1.0, 0.0], doc! {"source": "vehicle"}).unwrap();

    let hits = store.search(vec![1.0, 0.0, 0.0], 2).run().unwrap();
    assert_eq!(hits.len(), 2);
    assert_eq!(hits[0].text, "apple");
    assert_eq!(hits[1].text, "banana");
    // Scores are descending and the exact match is ~1.0 cosine similarity.
    assert!(hits[0].score >= hits[1].score);
    assert!((hits[0].score - 1.0).abs() < 1e-4);
}

#[test]
fn search_respects_metadata_filter() {
    let (_dir, _db, store) = store();
    store.add("apple", vec![1.0, 0.0, 0.0], doc! {"source": "fruit"}).unwrap();
    store.add("almost", vec![0.98, 0.02, 0.0], doc! {"source": "vehicle"}).unwrap();
    store.add("banana", vec![0.9, 0.1, 0.0], doc! {"source": "fruit"}).unwrap();

    // Nearest is "apple"/"almost", but only fruit is allowed through.
    let hits = store
        .search(vec![1.0, 0.0, 0.0], 2)
        .filter(field("source").eq("fruit"))
        .run()
        .unwrap();
    assert!(hits.iter().all(|h| {
        matches!(h.document.get("source"), Ok(nitrite::common::Value::String(s)) if s == "fruit")
    }));
    assert_eq!(hits[0].text, "apple");
}

#[test]
fn selective_filter_finds_matches_outside_the_oversample_window() {
    let (_dir, _db, store) = store();
    // 40 chunks that are all closer to the query than the one we want, so the
    // fixed `k * oversample` window (5 * 4 = 20) contains none of the matches.
    for i in 0..40 {
        let jitter = i as f32 * 0.001;
        store
            .add(format!("other-{i}"), vec![1.0 - jitter, jitter, 0.0], doc! {"entry": "other"})
            .unwrap();
    }
    store.add("wanted", vec![0.0, 0.0, 1.0], doc! {"entry": "target"}).unwrap();

    let hits = store
        .search(vec![1.0, 0.0, 0.0], 5)
        .filter(field("entry").eq("target"))
        .run()
        .unwrap();
    assert_eq!(hits.len(), 1, "selective filter must reach past the initial window");
    assert_eq!(hits[0].text, "wanted");
}

#[test]
fn filter_matching_nothing_returns_empty() {
    let (_dir, _db, store) = store();
    for i in 0..20 {
        store.add(format!("d{i}"), vec![1.0, i as f32 * 0.01, 0.0], doc! {"entry": "a"}).unwrap();
    }

    let hits = store
        .search(vec![1.0, 0.0, 0.0], 5)
        .filter(field("entry").eq("nope"))
        .run()
        .unwrap();
    assert!(hits.is_empty());
}

#[test]
fn filter_returns_all_matches_when_fewer_than_k() {
    let (_dir, _db, store) = store();
    for i in 0..30 {
        store.add(format!("o{i}"), vec![1.0, i as f32 * 0.001, 0.0], doc! {"entry": "other"}).unwrap();
    }
    store.add("t1", vec![0.0, 0.0, 1.0], doc! {"entry": "target"}).unwrap();
    store.add("t2", vec![0.0, 0.1, 1.0], doc! {"entry": "target"}).unwrap();

    let hits = store
        .search(vec![1.0, 0.0, 0.0], 10)
        .filter(field("entry").eq("target"))
        .run()
        .unwrap();
    assert_eq!(hits.len(), 2);
    assert!(hits[0].score >= hits[1].score);
}

#[test]
fn min_score_filters_out_dissimilar_hits() {
    let (_dir, _db, store) = store();
    store.add("close", vec![1.0, 0.0, 0.0], doc! {}).unwrap();
    store.add("orthogonal", vec![0.0, 1.0, 0.0], doc! {}).unwrap();

    // Cosine similarity of orthogonal vectors is 0, well below 0.5.
    let hits = store
        .search(vec![1.0, 0.0, 0.0], 5)
        .min_score(0.5)
        .run()
        .unwrap();
    assert_eq!(hits.len(), 1);
    assert_eq!(hits[0].text, "close");
}

#[test]
fn delete_removes_a_hit() {
    let (_dir, _db, store) = store();
    let apple = store.add("apple", vec![1.0, 0.0, 0.0], doc! {}).unwrap();
    store.add("banana", vec![0.9, 0.1, 0.0], doc! {}).unwrap();

    assert_eq!(store.len().unwrap(), 2);
    assert!(store.delete(&apple).unwrap());
    assert_eq!(store.len().unwrap(), 1);
    assert!(store.get(&apple).unwrap().is_none());

    let hits = store.search(vec![1.0, 0.0, 0.0], 5).run().unwrap();
    assert!(hits.iter().all(|h| h.text != "apple"));
    assert_eq!(hits[0].text, "banana");
}