--- title: Indexing Vectors description: Vectors live alongside text and filters in the ParadeDB index canonical: https://www.paradedb.com/docs/reference/indexing/indexing-vectors --- This is a beta feature available in versions `0.25.0` and above. We strongly recommend `0.27.0` or later for substantial vector search improvements. See [How Vector Search Works](/concepts/vector/overview) for background. Make sure the [pgvector](https://github.com/pgvector/pgvector) extension is installed first. ParadeDB uses pgvector's vector types, but not its HNSW or IVF indexes. The ParadeDB index can index pgvector's `vector` type alongside your text and other columns. This lets you combine vector search with full text search and filters in a single index, which can significantly improve latency/recall for selective queries. ## Creating the Index The `mock_items` table comes with an `embedding` column of type `vector(8)` populated with sample embeddings. In this example, `embedding` is added to the ParadeDB index with cosine similarity as the distance function. ```sql SQL CREATE INDEX search_idx ON mock_items USING paradedb (id, description, category, embedding vector_cosine_ops); ``` ```ts Drizzle import { indexing } from "@paradedb/drizzle-paradedb"; // In the pgTable definition: (table) => [ indexing .paradedbIndex("search_idx") .on( table.id, table.description, table.category, indexing.vectorField(table.embedding, "cosine"), ), ]; ``` ```python Django from django.db import connection from paradedb.indexes import ParadeDBIndex with connection.schema_editor() as schema_editor: schema_editor.add_index( MockItem, ParadeDBIndex( fields={ "id": {}, "description": {}, "category": {}, "embedding": {"metric": "cosine"}, }, name="search_idx", ), ) ``` ```python SQLAlchemy from sqlalchemy import Index from paradedb.sqlalchemy import indexing idx = Index( "search_idx", indexing.ParadeDBField(MockItem.id), indexing.ParadeDBField(MockItem.description), indexing.ParadeDBField(MockItem.category), indexing.VectorField(MockItem.embedding, metric="cosine"), postgresql_using="paradedb", ) with engine.begin() as conn: idx.create(conn) ``` ```ruby Rails ActiveRecord::Base.connection.add_paradedb_index( :mock_items, fields: { id: {}, description: {}, category: {}, embedding: { metric: :cosine } }, name: :search_idx ) ``` ```cs EF Core modelBuilder.Entity() .HasParadeDbIndex("search_idx", e => e.Id) .HasField(e => e.Description) .HasField(e => e.Category) .HasField(e => e.Embedding, VectorMetric.Cosine); ``` The desired distance function is encoded into the index definition, and cannot be changed without reindexing. ```sql SQL embedding vector_l2_ops -- L2 (default) embedding vector_cosine_ops -- cosine embedding vector_ip_ops -- inner product ``` ```ts Drizzle indexing.vectorField(mockItems.embedding, "l2"); // L2 (default) indexing.vectorField(mockItems.embedding, "cosine"); // cosine indexing.vectorField(mockItems.embedding, "ip"); // inner product ``` ```python Django "embedding": {"metric": "l2"} # L2 (default) "embedding": {"metric": "cosine"} # cosine "embedding": {"metric": "ip"} # inner product ``` ```python SQLAlchemy indexing.VectorField(MockItem.embedding, metric="l2") # L2 (default) indexing.VectorField(MockItem.embedding, metric="cosine") # cosine indexing.VectorField(MockItem.embedding, metric="ip") # inner product ``` ```ruby Rails embedding: { metric: :l2 } # L2 (default) embedding: { metric: :cosine } # cosine embedding: { metric: :ip } # inner product ``` ```cs EF Core .HasField(e => e.Embedding, VectorMetric.L2) // L2 (default) .HasField(e => e.Embedding, VectorMetric.Cosine) // cosine .HasField(e => e.Embedding, VectorMetric.InnerProduct) // inner product ``` Only pgvector's `vector` type is supported. The `halfvec`, `sparsevec`, and `bit` types are not yet indexable. Each ParadeDB index currently supports one vector field, whose dimensions must be declared with `vector(N)`. If you track index build progress with [`pg_stat_progress_create_index`](https://www.postgresql.org/docs/current/progress-reporting.html#CREATE-INDEX-PROGRESS-REPORTING), you may notice progress appear to "stop" at intervals. This is expected: vectors are clustered with k-means at these points, which is computationally expensive. ## Index Options We recommend keeping the defaults, which work well for most workloads. The following options are available for development and debugging. ```sql SQL CREATE INDEX search_idx ON mock_items USING paradedb (id, description, category, embedding vector_cosine_ops) WITH (target_leaf_size=400, replication=2); ``` ```ts Drizzle import { indexing } from "@paradedb/drizzle-paradedb"; indexing .paradedbIndex("search_idx", { targetLeafSize: 400, }) .on( mockItems.id, mockItems.description, mockItems.category, indexing.vectorField(mockItems.embedding, "cosine"), ); ``` ```python Django from django.db import connection from paradedb.indexes import ParadeDBIndex with connection.schema_editor() as schema_editor: schema_editor.add_index( MockItem, ParadeDBIndex( fields={'id': {}, 'description': {}, 'category': {}, 'embedding': {'metric': 'cosine'}}, name="search_idx", target_leaf_size=400, ), ) ``` ```python SQLAlchemy from sqlalchemy import Index from paradedb.sqlalchemy import indexing idx = Index( "search_idx", indexing.ParadeDBField(MockItem.id), indexing.ParadeDBField(MockItem.description), indexing.ParadeDBField(MockItem.category), indexing.VectorField(MockItem.embedding, metric="cosine"), postgresql_using="paradedb", postgresql_with={'target_leaf_size': 400}, ) with engine.begin() as conn: idx.create(conn) ``` ```ruby Rails ActiveRecord::Base.connection.add_paradedb_index( :mock_items, fields: { id: {}, description: {}, category: {}, embedding: { metric: :cosine }, }, name: :search_idx, index_options: { target_leaf_size: 400 } ) ``` ```cs EF Core modelBuilder.Entity() .HasParadeDbIndex("search_idx", e => e.Id) .HasField(e => e.Description) .HasField(e => e.Category) .HasField(e => e.Embedding, VectorMetric.Cosine); ``` Fraction of indexed vectors sampled for training whenever a segment is clustered. `0` chooses the fraction automatically based on available memory. The target number of indexed vectors in a cluster. By default, cluster centroids are quantized. This makes them more compact and faster to search. ParadeDB uses boundary replication, which means that vectors sitting close to two clusters are replicated across both. Replication improves recall. A higher value replicates to more clusters, at the cost of longer index build times and more storage. ParadeDB uses two quantization layers by default, each receiving 1 bit per dimension. Each layer can use up to 4 bits. Set `quantization = 0` to disable quantization entirely. In our testing, the default `'[1,1]'` delivers the best performance.