Tutorial 2: Integrating Django Vector Database

Welcome to Tutorial 2! In the previous tutorial, we setup our Blog Post project that we will use for the rest of the tutorial to demonstrate the feature of django-vectordb. In this tutorial, we will integrate django-vectordb into our Blog Site. By the end of this tutorial, we will:

• Integrate django-vectordb into our project.
• Update our Post model to allow django-vectordb to automatically extract the text for vector search, as well as any metadata we want to filter on.
• Update the vector database with all the blog Posts that are currently on our site.
• Configure our site to automatically sync with the vector database whenever we create, update or delete posts.
• Configure our site to display related blog recommendations using the django-vectordb.

Let's get started!

Adding VectorDB to the Project

If you haven't installed vectordb already, install it with the following command:

pip install "django-vectordb[standard]"

Next lets add django-vectordb to the INSTALLED_APPS. Open tutorial/settings.py and add vectordb to the installed apps.

tutorial/settings.py

# settings.py
INSTALLED_APPS = [
    "django.contrib.admin",
    "django.contrib.auth",
    "django.contrib.contenttypes",
    "django.contrib.sessions",
    "django.contrib.messages",
    "django.contrib.staticfiles",
    "blog",
    "vectordb", # <--
]

Finally, lets run the migrations to create the vectordb table.

./manage.py migrate

Update Post Model for VectorDB to Automatically Extract Text and Metadata

Next lets add to methods to our model to allow vectordb to properly index our Posts.

First we need to pick the text we want to search by. For this tutorial we use both the title and the description of each blog. We tell vectordb about the text we want to use through the get_vectordb_text method. Second, we need to select the metadata we want to use for filtering. For this example, we will use the user_id, username and created_at of each blog post. We will break down the create_at into year, month and day to demonstrate nested filtering. We tell django-vectordb about the metadata we want to use through the get_vectordb_metadata method. If we don,t provide this method, vectordb will use the django serializer to serialize the entire model. This maybe suboptimal for large models.

# blog/models.py
from django.db import models
from django.contrib.auth import get_user_model

User = get_user_model()


class Post(models.Model):
    title = models.CharField(max_length=100)
    description = models.TextField()
    user = models.ForeignKey(User, on_delete=models.CASCADE)
    created_at = models.DateTimeField(auto_now_add=True)

    def get_vectordb_text(self):
        return self.title + " " + self.description

    def get_vectordb_metadata(self):
        return {
            "user_id": self.user.id,
            "username": self.user.username,
            "created_at": {
                "year": self.created_at.year,
                "month": self.created_at.month,
                "day": self.created_at.day,
            },
        }

Update VectorDB with Existing Posts

Django vector database provides a management command to update the vector database with existing models. It takes two arguments the app_name and the model_name, which in this case are blog and Post. Lets run this command to update the vector database with all the existing posts.

./manage.py vectordb_sync <app_name> <model_name>

./manage.py vectordb_sync blog Post

Info

Django vector database also provides a management command for reseting the vector database. Note that this is distractive and irreversible.

./manage.py vectordb_reset

Summary

In this tutorial, we successfully integrated the Django vector database into our blog post project. Additionally, we synced our existing blog posts with the Django vector database using the vectordb_sync management command. In the upcoming tutorial, we will set up our blog post app to automatically synchronize with the Django vector database whenever we create, update, or delete blog posts.