Vector Data Management

Welcome back! In the previous lesson, you learned how to perform search queries in a vector database, retrieving relevant information based on similarity scores. Today, we will focus on managing vector data more effectively by using operations that allow you to update, retrieve, and delete vector records. These operations are essential for maintaining and optimizing your vector database, ensuring that your data remains accurate and relevant. By the end of this lesson, you will be equipped with the skills to modify, access, and remove vector data efficiently.

In this section, we will introduce three essential operations for managing vector data in Qdrant: updating, retrieving, and deleting vectors. These operations allow you to modify existing data, access specific records, and remove data that is no longer needed. Understanding how to use these operations effectively will help you maintain a clean and efficient vector database.

• Updating: In Qdrant, you can update the metadata (payload) of a vector using the set_payload method. set_payload updates the provided payload keys and leaves other existing payload fields intact.
• Fetching: To retrieve specific vectors by their IDs, you can use the retrieve method, which returns the vectors and their associated metadata.
• Deleting: To remove vectors from your collection, you can use the delete method, specifying the IDs of the vectors you want to remove.

Let's look at how each of these operations works in practice.

Suppose you have a vector with the ID "rec3" that you want to update with new metadata. In Qdrant, you should use the UUID version of the string ID when calling the set_payload method.

import uuid

# Update the payload (metadata) for vector with ID "rec3"
client.set_payload(
    collection_name=collection_name,
    payload={"category": "nutrition", "new": True},
    points=[str(uuid.uuid5(uuid.NAMESPACE_DNS, "rec3"))]
)

In this example, we convert the string ID "rec3" to a UUID using uuid.uuid5(uuid.NAMESPACE_DNS, "rec3") and use it as the point ID for the update.

To fetch vector data from Qdrant, you should also use the UUID version of the string ID when calling the retrieve method.

import uuid

# Retrieve the vector with ID "rec3"
result = client.retrieve(
    collection_name=collection_name,
    ids=[str(uuid.uuid5(uuid.NAMESPACE_DNS, "rec3"))],
    with_vectors=True
)

print(result)

This code retrieves the vector with the UUID corresponding to "rec3" along with its values and metadata.

To delete vector data in Qdrant, use the UUID version of the string ID when specifying the point to delete.

import uuid
from qdrant_client.http.models import PointIdsList

# Delete the vector with ID "rec4"
client.delete(
    collection_name=collection_name,
    points_selector=PointIdsList(points=[str(uuid.uuid5(uuid.NAMESPACE_DNS, "rec4"))])
)

In this example, we delete the vector with the UUID corresponding to "rec4" from the collection.

Let's bring together the concepts of updating, fetching, and deleting vector data in Qdrant by implementing a comprehensive example. This example demonstrates how to update a vector, delete another, and verify these changes by retrieving the current state of the data.

The following snippet assumes that example_collection already exists and contains UUID-based points for the original IDs "rec3" and "rec4", as created in the earlier insertion examples.

from qdrant_client import QdrantClient, models
import uuid

# Connect to Qdrant
client = QdrantClient("localhost", port=6333)
collection_name = "example_collection"

# Update metadata for rec3
rec3_uuid = str(uuid.uuid5(uuid.NAMESPACE_DNS, "rec3"))
client.set_payload(
    collection_name=collection_name,
    payload={"category": "nutrition", "new": True},
    points=[rec3_uuid]
)
print("Updated rec3.")

# Delete record rec4
rec4_uuid = str(uuid.uuid5(uuid.NAMESPACE_DNS, "rec4"))
client.delete(
    collection_name=collection_name,
    points_selector=models.PointIdsList(points=[rec4_uuid])
)
print("Deleted rec4.")

# Verify update and deletion
rec3_after = client.retrieve(collection_name=collection_name, ids=[rec3_uuid])
rec4_after = client.retrieve(collection_name=collection_name, ids=[rec4_uuid])

print("rec3 after update:", rec3_after)
print("rec4 after delete (should be empty):", rec4_after)

# Cleanup
client.delete_collection(collection_name)
print(f"Collection '{collection_name}' deleted.")

In this example, we first update the vector with ID "rec3" by modifying its metadata. We then delete the vector with ID "rec4". Finally, we verify these operations using the retrieve method. This workflow ensures that your update and delete operations have been applied correctly.

In this lesson, you learned how to manage vector data in a vector database using Qdrant's update, retrieve, and delete operations. We explored how to modify existing data, access specific records, and remove data that is no longer needed. These operations are crucial for maintaining an efficient and relevant vector database. As you move forward, you'll have the opportunity to practice these concepts through exercises that reinforce what you've learned. Experiment with these methods and continue enhancing your skills in vector data management. Keep up the great work!