Vector Search

TuringDB has a built-in vector index that lets you run k-nearest-neighbor searches over embedding vectors. You bring your own embeddings, from any model or provider, and TuringDB handles the indexing, storage, and fast retrieval. Each vector is associated with a numerical ID. That ID can be a node property, an edge property, or a foreign key referencing data in an external system. This keeps the index lightweight and flexible: the vector store doesn’t need to know what your data looks like.

Vector indexes live at the TuringDB root level, independent of graphs and versioning. A single vector index can serve searches across multiple graphs and commits.

Create a vector index

A vector index is defined by a name, a dimension, and a distance metric. Syntax:

CREATE VECTOR INDEX <name> WITH DIMENSION <dim> METRIC <metric>

Parameter	Description
`<name>`	Identifier for the vector index
`<dim>`	Dimension of the embedding vectors (positive integer)
`<metric>`	Distance metric: `EUCLID` (Euclidean distance) or `COSINE` (cosine similarity)

Cypher
Python SDK

CREATE VECTOR INDEX doc_embeddings WITH DIMENSION 768 METRIC COSINE

client.query("CREATE VECTOR INDEX doc_embeddings WITH DIMENSION 768 METRIC COSINE")

Load embeddings

Once the index exists, load your pre-computed embeddings from a file. Each row in the file maps a numerical ID to a vector. The file path is relative to your TuringDB data directory (~/.turing/data by default).

The TuringDB data directory defaults to ~/.turing/data. You can change it at startup with the -turing-dir flag.

Syntax:

LOAD VECTOR FROM "<filepath>" IN <index_name>

Parameter	Description
`<filepath>`	Path to the embeddings file, relative to the TuringDB data directory
`<index_name>`	Name of the target vector index

Cypher
Python SDK

LOAD VECTOR FROM "document_vectors.csv" IN doc_embeddings

client.query('LOAD VECTOR FROM "document_vectors.csv" IN doc_embeddings')

Search

VECTOR SEARCH finds the k nearest neighbors of a query vector and yields their IDs. It is a read statement, so you can chain it with MATCH to pull back the actual graph data. Syntax:

VECTOR SEARCH IN <index_name> FOR <k> (<vector>) YIELD <variable>

Parameter	Description
`<index_name>`	Name of the vector index to search
`<k>`	Number of nearest neighbors to return (positive integer)
`<vector>`	Query vector as a list literal of float values
`<variable>`	Variable name to hold the result IDs

Standalone search

Cypher
Python SDK

VECTOR SEARCH IN doc_embeddings FOR 5 (0.12, 0.45, 0.78, 0.33) YIELD ids
RETURN ids

df = client.query("""
VECTOR SEARCH IN doc_embeddings FOR 5 (0.12, 0.45, 0.78, 0.33) YIELD ids
RETURN ids
""")
print(df)

Combining with MATCH

This is where it gets interesting. Chain VECTOR SEARCH with a MATCH clause to join the nearest-neighbor IDs back to your graph:

Cypher
Python SDK

VECTOR SEARCH IN doc_embeddings FOR 10 (0.12, 0.45, 0.78, 0.33) YIELD ids
MATCH (d:Document) WHERE d.id = ids
RETURN d.title, d.summary

df = client.query("""
VECTOR SEARCH IN doc_embeddings FOR 10 (0.12, 0.45, 0.78, 0.33) YIELD ids
MATCH (d:Document) WHERE d.id = ids
RETURN d.title, d.summary
""")
print(df)

The ids variable works exactly like a variable introduced by CALL ... YIELD, so any subsequent MATCH clause can reference it.

Manage indexes

List all vector indexes

SHOW VECTOR INDEXES

Delete a vector index

DELETE VECTOR INDEX doc_embeddings

This removes the index and frees the associated resources.

Complete workflow

Cypher
Python SDK

// 1. Create a vector index for product embeddings
CREATE VECTOR INDEX product_vectors WITH DIMENSION 384 METRIC COSINE

// 2. Load embeddings generated by your model
LOAD VECTOR FROM "product_embeddings.csv" IN product_vectors

// 3. Find the 10 most similar products to a query embedding
VECTOR SEARCH IN product_vectors FOR 10 (0.15, 0.82, 0.44, 0.91) YIELD ids
MATCH (p:Product) WHERE p.id = ids
RETURN p.name, p.price, p.category

// 4. Inspect existing indexes
SHOW VECTOR INDEXES

// 5. Clean up
DELETE VECTOR INDEX product_vectors

from turingdb import TuringDB

client = TuringDB(host="http://localhost:6666")

# 1. Create a vector index for product embeddings
client.query("CREATE VECTOR INDEX product_vectors WITH DIMENSION 384 METRIC COSINE")

# 2. Load embeddings generated by your model
client.query('LOAD VECTOR FROM "product_embeddings.csv" IN product_vectors')

# 3. Find the 10 most similar products to a query embedding
df = client.query("""
VECTOR SEARCH IN product_vectors FOR 10 (0.15, 0.82, 0.44, 0.91) YIELD ids
MATCH (p:Product) WHERE p.id = ids
RETURN p.name, p.price, p.category
""")
print(df)

# 4. Inspect existing indexes
client.query("SHOW VECTOR INDEXES")

# 5. Clean up
client.query("DELETE VECTOR INDEX product_vectors")

Embeddings as node properties

The vector index above is a standalone, root-level structure keyed by numeric IDs. Separately, you can store an embedding directly as a node or edge property (type Embedding) and compare embeddings inline with the cosine_similarity and euclidean_distance functions — no index required. Embedding literals use parentheses (...), not square brackets (which are list literals):

// Set an embedding property (inside a change)
MATCH (n:Document {name: 'intro'}) SET n.emb = (0.12, 0.45, 0.78, 0.33)

// Compare embeddings inline
MATCH (n:Document) RETURN n.name, cosine_similarity(n.emb, (0.12, 0.45, 0.78, 0.33))
MATCH (n:Document) RETURN n.name, euclidean_distance(n.emb, (0.12, 0.45, 0.78, 0.33))

Bulk-loading embeddings from Parquet — `LOAD EMBEDDING FROM`

To attach embeddings to existing nodes in bulk, load them from a Parquet file. This is a write, so run it inside a change. Syntax:

LOAD EMBEDDING FROM "<filepath>" AS <property_name>

The Parquet file (relative to the TuringDB data directory) must have exactly two columns:

Column	Type	Description
`node_id`	`INT64`	Internal TuringDB node ID to attach the embedding to
`embedding`	`FIXED_LEN_BYTE_ARRAY` (`dim × 4` bytes)	Little-endian float32 vector; dimension inferred from width

change = client.new_change()
client.checkout(change=change)
client.query('LOAD EMBEDDING FROM "embeddings.parquet" AS emb')
client.query("COMMIT")
client.query("CHANGE SUBMIT")
client.checkout()

The named property is created with type Embedding. The load fails if any node_id is missing from the graph, or if the property name already exists with a non-embedding type.

When importing a graph from JSONL, you can mark embedding properties at load time instead — see LOAD JSONL ... WITH EMBEDDINGS.

Get Started

Concepts

Benchmarks

Query Language

Importing data

Graph Development

Graph Algorithms

Tutorials

Python SDK

Troubleshooting

Vector Search

Create a vector index

Load embeddings

Search

Standalone search

Combining with MATCH

Manage indexes

List all vector indexes

Delete a vector index

Complete workflow

Embeddings as node properties

Bulk-loading embeddings from Parquet — `LOAD EMBEDDING FROM`

​Create a vector index

​Load embeddings

​Search

​Standalone search

​Combining with MATCH

​Manage indexes

​List all vector indexes

​Delete a vector index

​Complete workflow

​Embeddings as node properties

​Bulk-loading embeddings from Parquet — LOAD EMBEDDING FROM

Create a vector index

Load embeddings

Search

Standalone search

Combining with MATCH

Manage indexes

List all vector indexes

Delete a vector index

Complete workflow

Embeddings as node properties

Bulk-loading embeddings from Parquet — `LOAD EMBEDDING FROM`