How to beginner · 3 min read

How to create a Weaviate schema

Q: How to create a Weaviate schema

Use the official weaviate-client Python library to create a schema by defining classes and their properties in a JSON-like dictionary, then call client.schema.create_class(). This sets up your vector database schema for storing and querying data.

Quick answer

Use the official weaviate-client Python library to create a schema by defining classes and their properties in a JSON-like dictionary, then call client.schema.create_class(). This sets up your vector database schema for storing and querying data.

PREREQUISITES

Python 3.8+
pip install weaviate-client
Running Weaviate instance (local or cloud)
Weaviate URL and API key if applicable

Setup

Install the official weaviate-client Python package and set environment variables for your Weaviate URL and API key if needed.

bash

pip install weaviate-client

Step by step

Define a schema class with properties and create it using the Weaviate client. This example creates a Article class with title and content properties.

python

import os
import weaviate

# Connect to Weaviate instance
client = weaviate.Client(
    url=os.environ.get('WEAVIATE_URL', 'http://localhost:8080'),
    auth_client_secret=weaviate.AuthApiKey(api_key=os.environ.get('WEAVIATE_API_KEY')) if os.environ.get('WEAVIATE_API_KEY') else None
)

# Define schema class
article_class = {
    "class": "Article",
    "description": "A class to store articles",
    "properties": [
        {
            "name": "title",
            "dataType": ["text"],
            "description": "Title of the article"
        },
        {
            "name": "content",
            "dataType": ["text"],
            "description": "Content of the article"
        }
    ]
}

# Create the class in Weaviate schema
client.schema.create_class(article_class)

# Verify schema creation
schema = client.schema.get()
print(schema)

output

{'classes': [{'class': 'Article', 'description': 'A class to store articles', 'properties': [{'name': 'title', 'dataType': ['text'], 'description': 'Title of the article'}, {'name': 'content', 'dataType': ['text'], 'description': 'Content of the article'}]}]}

Common variations

You can create multiple classes, use different data types like int, number, date, or geoCoordinates. For async usage, use asyncio with the weaviate-client async support. You can also update or delete schema classes with client.schema.update_class() and client.schema.delete_class().

python

import asyncio
import weaviate

async def create_schema_async():
    client = weaviate.AsyncClient(
        url=os.environ.get('WEAVIATE_URL', 'http://localhost:8080'),
        auth_client_secret=weaviate.AuthApiKey(api_key=os.environ.get('WEAVIATE_API_KEY')) if os.environ.get('WEAVIATE_API_KEY') else None
    )

    new_class = {
        "class": "Book",
        "description": "A class for books",
        "properties": [
            {"name": "author", "dataType": ["text"]},
            {"name": "publishedYear", "dataType": ["int"]}
        ]
    }

    await client.schema.create_class(new_class)
    schema = await client.schema.get()
    print(schema)

asyncio.run(create_schema_async())

output

{'classes': [{'class': 'Article', ...}, {'class': 'Book', 'description': 'A class for books', 'properties': [{'name': 'author', 'dataType': ['text']}, {'name': 'publishedYear', 'dataType': ['int']}]}]}

Troubleshooting

If you get a connection error, verify your Weaviate URL and that the instance is running.
If schema creation fails with a 409 conflict, the class already exists; use update_class() or delete it first.
Ensure your API key is correct if authentication is enabled.

✅

Key Takeaways

Use the official weaviate-client Python library to manage schemas programmatically.
Define classes with properties specifying name, dataType, and description.
Handle authentication and connection via environment variables for security.
Use async client for asynchronous schema operations if needed.
Check for existing classes to avoid conflicts during schema creation.

Verified 2026-04

Verify ↗