Busca Vetorial Exata e Aproximada

WITH [...] AS reference_vector
SELECT [...]
FROM table
WHERE [...] -- uma cláusula WHERE é opcional
ORDER BY <DistanceFunction>(vectors, reference_vector)
LIMIT <N>

CREATE TABLE tab(id Int32, vec Array(Float32)) ENGINE = MergeTree ORDER BY id;

INSERT INTO tab VALUES (0, [1.0, 0.0]), (1, [1.1, 0.0]), (2, [1.2, 0.0]), (3, [1.3, 0.0]), (4, [1.4, 0.0]), (5, [1.5, 0.0]), (6, [0.0, 2.0]), (7, [0.0, 2.1]), (8, [0.0, 2.2]), (9, [0.0, 2.3]), (10, [0.0, 2.4]), (11, [0.0, 2.5]);

WITH [0., 2.] AS reference_vec
SELECT id, vec
FROM tab
ORDER BY L2Distance(vec, reference_vec) ASC
LIMIT 3;

   ┌─id─┬─vec─────┐
1. │  6 │ [0,2]   │
2. │  7 │ [0,2.1] │
3. │  8 │ [0,2.2] │
   └────┴─────────┘

CREATE TABLE table
(
  [...],
  vectors Array(Float*),
  INDEX <index_name> vectors TYPE vector_similarity(<type>, <distance_function>, <dimensions>) [GRANULARITY <N>]
)
ENGINE = MergeTree
ORDER BY [...]

ALTER TABLE table ADD INDEX <index_name> vectors TYPE vector_similarity(<type>, <distance_function>, <dimensions>) [GRANULARITY <N>];

ALTER TABLE table MATERIALIZE INDEX <index_name> SETTINGS mutations_sync = 2;

CREATE TABLE table
(
  [...],
  vectors Array(Float*),
  INDEX index_name vectors TYPE vector_similarity('hnsw', <distance_function>, <dimensions>[, <quantization>, <hnsw_max_connections_per_layer>, <hnsw_candidate_list_size_for_construction>]) [GRANULARITY N]
)
ENGINE = MergeTree
ORDER BY [...]

Storage consumption = Number of vectors * Dimension * Size of column data type

Storage consumption = 1 milhão * 1536 * 4 (para Float32) = 6,1 GB

Memória para vetores no índice (mv) = Número de vetores * Dimensão * Tamanho do tipo de dado quantizado
Memória para o grafo em memória (mg) = Número de vetores * hnsw_max_connections_per_layer * Bytes_per_node_id (= 4) * Layer_node_repetition_factor (= 2)

Consumo de memória: mv + mg

Memória para vetores no índice (mv) = 1 milhão * 1536 * 2 (para BFloat16) = 3072 MB
Memória para o grafo em memória (mg) = 1 milhão * 64 * 2 * 4 = 512 MB

Consumo de memória = 3072 + 512 = 3584 MB

WITH [...] AS reference_vector
SELECT [...]
FROM table
WHERE [...] -- uma cláusula WHERE é opcional
ORDER BY <DistanceFunction>(vectors, reference_vector)
LIMIT <N>

EXPLAIN indexes = 1
WITH [0.462, 0.084, ..., -0.110] AS reference_vec
SELECT id, vec
FROM tab
ORDER BY L2Distance(vec, reference_vec) ASC
LIMIT 10;

┌─explain─────────────────────────────────────────────────────────────────────────────────────────┐
 1. │ Expression (Project names)                                                                      │
 2. │   Limit (preliminary LIMIT (without OFFSET))                                                    │
 3. │     Sorting (Sorting for ORDER BY)                                                              │
 4. │       Expression ((Before ORDER BY + (Projection + Change column names to column identifiers))) │
 5. │         ReadFromMergeTree (default.tab)                                                         │
 6. │         Indexes:                                                                                │
 7. │           PrimaryKey                                                                            │
 8. │             Condition: true                                                                     │
 9. │             Parts: 1/1                                                                          │
10. │             Granules: 575/575                                                                   │
11. │           Skip                                                                                  │
12. │             Name: idx                                                                           │
13. │             Description: vector_similarity GRANULARITY 100000000                                │
14. │             Parts: 1/1                                                                          │
15. │             Granules: 10/575                                                                    │
    └─────────────────────────────────────────────────────────────────────────────────────────────────┘

WITH [0., 2.] AS reference_vec
SELECT id, vec
FROM tab
WHERE year = 2025
ORDER BY L2Distance(vec, reference_vec) ASC
LIMIT 3;

SELECT bookid, author, title
FROM books
WHERE price < 2.00
ORDER BY cosineDistance(book_vector, getEmbedding('Books on ancient Asian empires'))
LIMIT 10

SELECT bookid, author, title
FROM books
WHERE price < 2.00
ORDER BY cosineDistance(book_vector, getEmbedding('Books on ancient Asian empires'))
LIMIT 10
SETTING vector_search_index_fetch_multiplier = 3.0;

EXPLAIN header = 1
WITH [0., 2.] AS reference_vec
SELECT id
FROM tab
ORDER BY L2Distance(vec, reference_vec) ASC
LIMIT 3
SETTINGS vector_search_with_rescoring = 0

Query id: a2a9d0c8-a525-45c1-96ca-c5a11fa66f47

    ┌─explain─────────────────────────────────────────────────────────────────────────────────────────────────┐
 1. │ Expression (Project names)                                                                              │
 2. │ Header: id Int32                                                                                        │
 3. │   Limit (preliminary LIMIT (without OFFSET))                                                            │
 4. │   Header: L2Distance(__table1.vec, _CAST([0., 2.]_Array(Float64), 'Array(Float64)'_String)) Float64     │
 5. │           __table1.id Int32                                                                             │
 6. │     Sorting (Sorting for ORDER BY)                                                                      │
 7. │     Header: L2Distance(__table1.vec, _CAST([0., 2.]_Array(Float64), 'Array(Float64)'_String)) Float64   │
 8. │             __table1.id Int32                                                                           │
 9. │       Expression ((Before ORDER BY + (Projection + Change column names to column identifiers)))         │
10. │       Header: L2Distance(__table1.vec, _CAST([0., 2.]_Array(Float64), 'Array(Float64)'_String)) Float64 │
11. │               __table1.id Int32                                                                         │
12. │         ReadFromMergeTree (default.tab)                                                                 │
13. │         Header: id Int32                                                                                │
14. │                 _distance Float32                                                                       │
    └─────────────────────────────────────────────────────────────────────────────────────────────────────────┘

CREATE TABLE tab(id Int32, vec Array(Float32) CODEC(NONE), INDEX idx vec TYPE vector_similarity('hnsw', 'L2Distance', 2)) ENGINE = MergeTree ORDER BY id;

2026-02-03 07:39:10.351635 [1386] f0ac5c85-1b1c-4f35-8848-87a1d1aa00ba : VectorSimilarityIndex Start loading vector similarity index

<...>

2026-02-03 07:40:25.217603 [1386] f0ac5c85-1b1c-4f35-8848-87a1d1aa00ba : VectorSimilarityIndex Loaded vector similarity index: max_level = 2, connectivity = 64, size = 1808111, capacity = 1808111, memory_usage = 8.00 GiB, bytes_per_vector = 4096, scalar_words = 1024, nodes = 1808111, edges = 51356964, max_edges = 233395072

SELECT metric, value
FROM system.metrics
WHERE metric = 'VectorSimilarityIndexCacheBytes'

SYSTEM FLUSH LOGS query_log;

SELECT ProfileEvents['VectorSimilarityIndexCacheHits'], ProfileEvents['VectorSimilarityIndexCacheMisses']
FROM system.query_log
WHERE type = 'QueryFinish' AND query_id = '<...>'
ORDER BY event_time_microseconds;

search_v = openai_client.embeddings.create(input = "[Good Books]", model='text-embedding-3-large', dimensions=1536).data[0].embedding

params = {'search_v': search_v}
result = chclient.query(
   "SELECT id FROM items
    ORDER BY cosineDistance(vector, %(search_v)s)
    LIMIT 10",
    parameters = params)

search_v = openai_client.embeddings.create(input = "[Good Books]", model='text-embedding-3-large', dimensions=1536).data[0].embedding

params = {'$search_v_binary$': np.array(search_v, dtype=np.float32).tobytes()}
result = chclient.query(
   "SELECT id FROM items
    ORDER BY cosineDistance(vector, reinterpret($search_v_binary$, 'Array(Float32)'))
    LIMIT 10"
    parameters = params)

SELECT database, table, name, formatReadableSize(data_compressed_bytes)
FROM system.data_skipping_indices
WHERE type = 'vector_similarity';

┌─database─┬─table─┬─name─┬─formatReadab⋯ssed_bytes)─┐
│ default  │ tab   │ idx  │ 348.00 MB                │
└──────────┴───────┴──────┴──────────────────────────┘

column_name QBit(element_type, dimension)

CREATE TABLE fruit_animal (
    word String,
    vec QBit(Float64, 5)
) ENGINE = MergeTree
ORDER BY word;

INSERT INTO fruit_animal VALUES
    ('apple', [-0.99105519, 1.28887844, -0.43526649, -0.98520696, 0.66154391]),
    ('banana', [-0.69372815, 0.25587061, -0.88226235, -2.54593015, 0.05300475]),
    ('orange', [0.93338752, 2.06571317, -0.54612565, -1.51625717, 0.69775337]),
    ('dog', [0.72138876, 1.55757105, 2.10953259, -0.33961248, -0.62217325]),
    ('cat', [-0.56611276, 0.52267331, 1.27839863, -0.59809804, -1.26721048]),
    ('horse', [-0.61435682, 0.48542571, 1.21091247, -0.62530446, -1.33082533]);

SELECT
    word,
    L2DistanceTransposed(vec, [-0.88693672, 1.31532824, -0.51182908, -0.99652702, 0.59907770], 64) AS distance
FROM fruit_animal
ORDER BY distance;

   ┌─word───┬────────────distance─┐
1. │ apple  │ 0.14639757188169716 │
2. │ banana │   1.998961369007679 │
3. │ orange │   2.039041552613732 │
4. │ cat    │   2.752802631487914 │
5. │ horse  │  2.7555776805484813 │
6. │ dog    │   3.382295083120104 │
   └────────┴─────────────────────┘

SELECT
    word,
    L2DistanceTransposed(vec, [-0.88693672, 1.31532824, -0.51182908, -0.99652702, 0.59907770], 12) AS distance
FROM fruit_animal
ORDER BY distance;

   ┌─word───┬───────────distance─┐
1. │ apple  │  0.757668703053566 │
2. │ orange │ 1.5499475034938677 │
3. │ banana │ 1.6168396735102937 │
4. │ cat    │  2.429752230904804 │
5. │ horse  │  2.524650475528617 │
6. │ dog    │   3.17766975527459 │
   └────────┴────────────────────┘