How to beginner · 3 min read

How to store OpenAI embeddings in a database

Q: How to store OpenAI embeddings in a database

Use the OpenAI SDK to generate embeddings with models like text-embedding-3-large. Convert the embedding vector to a storable format (e.g., JSON or binary) and save it in a database column designed for vectors or arrays, such as FLOAT[] in PostgreSQL or BLOB in SQLite.

Quick answer

Use the OpenAI SDK to generate embeddings with models like text-embedding-3-large. Convert the embedding vector to a storable format (e.g., JSON or binary) and save it in a database column designed for vectors or arrays, such as FLOAT[] in PostgreSQL or BLOB in SQLite.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install openai>=1.0
Basic knowledge of SQL and a database (e.g., SQLite, PostgreSQL)

Setup

Install the openai Python package and set your API key as an environment variable for secure access.

bash

pip install openai>=1.0

Step by step

This example shows how to generate an embedding using OpenAI's text-embedding-3-large model and store it in a SQLite database as a JSON string.

python

import os
import json
import sqlite3
from openai import OpenAI

# Initialize OpenAI client
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

# Connect to SQLite database (or create it)
conn = sqlite3.connect("embeddings.db")
cursor = conn.cursor()

# Create table to store embeddings
cursor.execute('''
CREATE TABLE IF NOT EXISTS embeddings (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    text TEXT NOT NULL,
    embedding TEXT NOT NULL
)
''')

# Text to embed
text_to_embed = "OpenAI provides powerful embedding models."

# Generate embedding
response = client.embeddings.create(
    model="text-embedding-3-large",
    input=text_to_embed
)
embedding_vector = response.data[0].embedding

# Convert embedding vector to JSON string for storage
embedding_json = json.dumps(embedding_vector)

# Insert into database
cursor.execute(
    "INSERT INTO embeddings (text, embedding) VALUES (?, ?)",
    (text_to_embed, embedding_json)
)
conn.commit()

# Query and print stored embedding
cursor.execute("SELECT id, text, embedding FROM embeddings")
rows = cursor.fetchall()
for row in rows:
    stored_id, stored_text, stored_embedding_json = row
    stored_embedding = json.loads(stored_embedding_json)
    print(f"ID: {stored_id}, Text: {stored_text}, Embedding vector length: {len(stored_embedding)}")

conn.close()

output

ID: 1, Text: OpenAI provides powerful embedding models., Embedding vector length: 1536

Common variations

Use PostgreSQL with FLOAT8[] column type for native vector storage.
Store embeddings as binary blobs for efficiency.
Use async OpenAI client calls for high throughput.
Use different embedding models like text-embedding-3-small for smaller vectors.

Troubleshooting

If you get API key errors, verify OPENAI_API_KEY is set correctly in your environment.
For database insertion errors, ensure your table schema matches the data types.
If embeddings seem empty or incorrect, check the model name and API response structure.

✅

Key Takeaways

Use the OpenAI SDK's embeddings endpoint with a current model like text-embedding-3-large.
Store embeddings as JSON strings or native array types depending on your database capabilities.
Always secure your API key via environment variables and never hardcode it.
SQLite is good for prototyping; use PostgreSQL or specialized vector DBs for production.
Validate your database schema matches the embedding data format to avoid insertion errors.

Verified 2026-04 · text-embedding-3-large

Verify ↗