Qdrant

The Qdrant piece in ZBrain Flow provides integration with Qdrant's vector database, allowing you to store, manage, and search vector embeddings directly from your workflows. This powerful connector enables you to work with high-dimensional vectors for similarity search and retrieval without specialized knowledge of vector databases. With Qdrant integration, you can build automations that store AI-generated embeddings, perform semantic searches, manage vector collections, and leverage vector similarity for advanced data matching and retrieval. This piece is ideal for organizations looking to incorporate vector search capabilities into their automated processes for applications like semantic document search, recommendation systems, and AI-powered matching workflows.

How to Use Qdrant in ZBrain Flow?

Select Qdrant as Your Connection

• Click on the '+' button in the Flow and search for Qdrant.

• Select Qdrant.

• Decide on the action you need, then select it. ZBrain Flow provides several options:

  • Add points to collection – Store vector embeddings and metadata in your database.

  • Get Collection List – View all vector collections in your Qdrant instance.

  • Get Collection Infos – Retrieve all details about a specific collection.

  • Delete Collection – Remove a collection from your database.

  • Delete Points – Remove specific vectors from a collection.

  • Get Points – Retrieve vectors and their metadata from a collection.

  • Search Points – Find vectors most similar to a query vector.

How to Connect to Your Qdrant Vector Database

Before using any Qdrant actions in ZBrain Flow, you'll need to set up a connection to your Qdrant instance. This is a one-time setup that will allow you to access your vector database. The steps involved in connecting to your Qdrant instance are as follows:

• From the Qdrant action you’ve selected, click on the connection dropdown and select 'Create connection'.

• In the popup window, you'll have to enter the following details:

  • 'Connection Name' - Enter a descriptive name to identify this Qdrant connection

  • 'Server Address' - Paste the URL of the Qdrant instance

  • 'API KEY' - Enter the API key you created or configured

• Click 'Save' to store this connection

How to Add Points to a Collection in Qdrant

Step 1: Configure Collection

• In the 'Collection Name' field, enter the name of the collection where you want to store vectors.

Step 2: Provide Vector Data

• In the 'Embeddings' field, enter your vector embeddings.

Step 3: Set Vector Identifiers (Optional)

• Click 'Add Item' in the 'Embeddings Ids' section to specify custom IDs for your vectors. If left empty, Qdrant will generate IDs automatically.

Step 4: Configure Vector Similarity Settings

• From the 'Calculation Method of distance' dropdown, select how similarity should be measured. The default is 'Cosine', which works well for most embedding models. Other options include 'Euclidean' and 'Dot', depending on your embedding model's properties.

Step 5: Add Metadata (Optional)

• In the 'Additional Payload' field, enter JSON data to associate with your vectors.

Step 6: Select Storage Type

• From the 'Storage' dropdown, select where your vectors will be stored. 'On Disk' is more suitable for large collections. 'On Memory' provides faster access but requires more RAM.

How to Get the Collection List

To get the collection list, you need to connect to your Qdrant instance.

How to Get Collection Infos

Step 1: Specify Collection

• First, connect to your Qdrant instance.

• In the 'Collection Name' field, enter the name of the collection you want to analyze. This should be an existing collection in your Qdrant database.

How to Delete a Collection

For this action, you can follow the same steps as in the "How to Get Collection Infos" section above.

After configuration, instead of retrieving information, this action will permanently delete the specified collection from your Qdrant instance, removing all vectors and associated metadata.

How to Delete Points from Collections

Step 1: Specify Collection

• In the 'Collection Name' field, enter the name of the collection containing the vectors you want to delete.

Step 2: Define Selection Method

• From the 'Choose Points By' dropdown, select the method to use to get the points. The available options are:

  • 'Ids' – Retrieve points by specifying their unique identifiers.

  • 'Filtering' – Use conditions to filter and fetch points based on specific criteria.

Step 3: Identifying Points to Delete (Based on Selected Method)

If you have selected 'Ids' as your method:

• In the 'Ids' field, provide the list of vector IDs you want to remove.

If you have selected 'Filtering' as your method:

• Click 'Add Item' in the 'Must Have' section to define metadata conditions that vectors must match to be deleted.

• Optionally, click 'Add Item' in the 'Must Not Have' section to define conditions that will exempt vectors from deletion.

How to Get Points from Your Collection

For this action, you can follow the same steps as in the "How to Delete Points from Collections" section above. After configuration, instead of deleting vectors, this action will retrieve the specified vectors from your Qdrant collection.

How to Search Points

Step 1: Specify Collection

• In the 'Collection Name' field, enter the name of the collection you want to search within.

Step 2: Provide Query Vector

• In the 'Embedding' field, enter the vector embedding you want to search for. This should be an array of numbers with the same dimensionality as your collection vectors.

Step 3: Apply Metadata Filters (Optional)

• Click 'Add Item' in the 'Must Have' section to define conditions that vectors must match to be included in results.

• Click 'Add Item' in the 'Must Not Have' section to define conditions that will exclude vectors from results.

Step 4: Configure Negative Search (Optional)

• In the 'Negative Vector' field, optionally provide a vector embedding you want results to be dissimilar from. This helps refine search by pushing results away from certain characteristics.

Step 5: Set Results Limit

• In the 'Limit Result' field, specify the maximum number of similar vectors to return. The default is 20, but you can adjust this based on your needs.