Python

In the subsequent examples, assume the following variables are defined

python

DATABASE_URL = "postgresql://postgres:postgres@localhost:5432/postgres"
embedding = [1, 2, 3]
query = "My text input"

psycopg2

Install psycopg2

bash

pip install psycopg2-binary

Connect to database and store / query vectors

python

import psycopg2

conn = psycopg2.connect(DATABASE_URL)
cur = conn.cursor()

# Insert a vector
cur.execute("INSERT INTO books (book_embedding) VALUES (%s)", (embedding,))

# Find nearest rows to a vector
cur.execute(f"SELECT * FROM books ORDER BY book_embedding <-> %s LIMIT 5", (embedding,))

# Find nearest rows to a vector generated from text
cur.execute(f"SELECT * FROM books ORDER BY book_embedding <-> text_embedding('BAAI/bge-small-en', %s) LIMIT 5", (query,))

cur.close()
conn.close()

psycopg3

Enable the extension

python

conn.execute('CREATE EXTENSION IF NOT EXISTS lantern')

Create a table

python

conn.execute('CREATE TABLE items (id bigserial PRIMARY KEY, embedding REAL[3])')

Insert a vector

python

embedding = [1, 2, 3]
conn.execute('INSERT INTO items (embedding) VALUES (%s)', (embedding,))

Get the nearest neighbors to a vector

python

conn.execute('SELECT * FROM items ORDER BY embedding <-> %s LIMIT 5', (embedding,)).fetchall()

asyncpg

Install asyncpg

bash

pip install asyncpg

Connect to database and store / query vectors

python

import asyncpg

conn = await asyncpg.connect(DATABASE_URL)

# Insert a vector
await conn.execute("INSERT INTO books (book_embedding) VALUES ($1)", embedding)

# Find nearest rows to a vector
await conn.fetch(f"SELECT * FROM books ORDER BY book_embedding <-> $1 LIMIT 5", embedding)

# Find nearest rows to a vector generated from text
await conn.fetch(f"SELECT * FROM books ORDER BY book_embedding <-> text_embedding('BAAI/bge-small-en', $1) LIMIT 5", query)

await conn.close()

Lantern Python Client

See the Github repo for documentation and examples.

Lantern Pinecone Client

See the Github repo for documentation and examples.

Edit this page