Skip to content

Qdrant Integration

Qdrant is one of the most popular vector search engines available, and we’ve made it easy to use Qdrant’s vector search capabilities on your computer vision data directly from FiftyOne!

Follow these simple instructions to configure your Qdrant server and get started using Qdrant + FiftyOne.

FiftyOne provides an API to create Qdrant collections, upload vectors, and run similarity queries, both programmatically in Python and via point-and-click in the App.

Note

Did you know? You can search by natural language using Qdrant similarity indexes!

image-similarity

Basic recipe

The basic workflow to use Qdrant to create a similarity index on your FiftyOne datasets and use this to query your data is as follows:

  1. Start a Qdrant service locally

  2. Load a dataset into FiftyOne

  3. Compute embedding vectors for samples or patches in your dataset, or select a model to use to generate embeddings

  4. Use the compute_similarity() method to generate a Qdrant similarity index for the samples or object patches in a dataset by setting the parameter backend="qdrant" and specifying a brain_key of your choice

  5. Use this Qdrant similarity index to query your data with sort_by_similarity()

  6. If desired, delete the index

The example below demonstrates this workflow.

Note

You must launch a Qdrant server and install the Qdrant Python client to run this example:

docker pull qdrant/qdrant
docker run -p 6333:6333 qdrant/qdrant

pip install qdrant-client

Note that, if you are using a custom Qdrant server, you can store your credentials as described in this section to avoid entering them manually each time you interact with your Qdrant index.

First let’s load a dataset into FiftyOne and compute embeddings for the samples:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

# Step 1: Load your data into FiftyOne
dataset = foz.load_zoo_dataset("quickstart")

# Steps 2 and 3: Compute embeddings and create a similarity index
qdrant_index = fob.compute_similarity(
    dataset,
    brain_key="qdrant_index",
    backend="qdrant",
)

Once the similarity index has been generated, we can query our data in FiftyOne by specifying the brain_key:

# Step 4: Query your data
query = dataset.first().id  # query by sample ID
view = dataset.sort_by_similarity(
    query,
    brain_key="qdrant_index",
    k=10,  # limit to 10 most similar samples
)

# Step 5 (optional): Cleanup

# Delete the Qdrant collection
qdrant_index.cleanup()

# Delete run record from FiftyOne
dataset.delete_brain_run("qdrant_index")

Note

Skip to this section for a variety of common Qdrant query patterns.

Setup

The easiest way to get started with Qdrant is to install locally via Docker:

docker pull qdrant/qdrant
docker run -p 6333:6333 qdrant/qdrant

Installing the Qdrant client

In order to use the Qdrant backend, you must also install the Qdrant Python client:

pip install qdrant-client

Using the Qdrant backend

By default, calling compute_similarity() or sort_by_similarity() will use an sklearn backend.

To use the Qdrant backend, simply set the optional backend parameter of compute_similarity() to "qdrant":

import fiftyone.brain as fob

fob.compute_similarity(..., backend="qdrant", ...)

Alternatively, you can permanently configure FiftyOne to use the Qdrant backend by setting the following environment variable:

export FIFTYONE_BRAIN_DEFAULT_SIMILARITY_BACKEND=qdrant

or by setting the default_similarity_backend parameter of your brain config located at ~/.fiftyone/brain_config.json:

{
    "default_similarity_backend": "qdrant"
}

Authentication

If you are using a custom Qdrant server, you can provide your credentials in a variety of ways.

Environment variables (recommended)

The recommended way to configure your Qdrant credentials is to store them in the environment variables shown below, which are automatically accessed by FiftyOne whenever a connection to Qdrant is made.

export FIFTYONE_BRAIN_SIMILARITY_QDRANT_URL=localhost:6333
export FIFTYONE_BRAIN_SIMILARITY_QDRANT_API_KEY=XXXXXXXX
export FIFTYONE_BRAIN_SIMILARITY_QDRANT_GRPC_PORT=6334
export FIFTYONE_BRAIN_SIMILARITY_QDRANT_PREFER_GRPC=false

The API_KEY, GRPC_PORT, and PREFER_GRPC environment variables are optional.

FiftyOne Brain config

You can also store your credentials in your brain config located at ~/.fiftyone/brain_config.json:

{
    "similarity_backends": {
        "qdrant": {
            "url": "http://localhost:6333",
            "api_key": "XXXXXXXX",
            "grpc_port": 6334,
            "prefer_grpc": false
        }
    }
}

Note that this file will not exist until you create it.

Keyword arguments

You can manually provide credentials as keyword arguments each time you call methods like compute_similarity() that require connections to Qdrant:

import fiftyone.brain as fob

qdrant_index = fob.compute_similarity(
    ...
    backend="qdrant",
    brain_key="qdrant_index",
    url="http://localhost:6333",
    api_key="XXXXXXXX",
    grpc_port=6334,
    prefer_grpc=False
)

Note that, when using this strategy, you must manually provide the credentials when loading an index later via load_brain_results():

qdrant_index = dataset.load_brain_results(
    "qdrant_index",
    url="http://localhost:6333",
    api_key="XXXXXXXX",
    grpc_port=6334,
    prefer_grpc=False
)

Qdrant config parameters

The Qdrant backend supports a variety of query parameters that can be used to customize your similarity queries. These parameters broadly fall into four categories:

  1. Basic vector database parameters

  2. Hierarchical navigable small world (HNSW) parameters

  3. Write-ahead-log (WAL) parameters

  4. Performance/optimizers parameters

For detailed information on these parameters, see the Qdrant documentation.

You can specify these parameters via any of the strategies described in the previous section. Here’s an example of a brain config that includes all of the available parameters:

{
    "similarity_backends": {
        "qdrant": {
            "metric": "cosine",
            "replication_factor": null,
            "shard_number": null,
            "write_consistency_factor": null,
            "hnsw_config": {
                "m": 16,
                "ef_construct": 100,
                "full_scan_threshold": 10000,
                "max_indexing_threads": null,
                "on_disk": null,
                "payload_m": null
            },
            "optimizers_config": {
                "deleted_threshold": 0.2,
                "vacuum_min_vector_number": 1000,
                "default_segment_number": 0,
                "max_segment_size": null,
                "memmap_threshold": null,
                "indexing_threshold": 20000,
                "flush_interval_sec": 5,
                "max_optimization_threads": 1
            },
            "wal_config": {
                "wal_capacity_mb": 32,
                "wal_segments_ahead": 0
            }
        }
    }
}

However, typically these parameters are directly passed to compute_similarity() to configure a specific new index:

qdrant_index = fob.compute_similarity(
    ...
    backend="qdrant",
    brain_key="qdrant_index",
    collection_name="your-collection-name",
    metric="cosine",
    replication_factor=1,
)

Managing brain runs

FiftyOne provides a variety of methods that you can use to manage brain runs.

For example, you can call list_brain_runs() to see the available brain keys on a dataset:

import fiftyone.brain as fob

# List all brain runs
dataset.list_brain_runs()

# Only list similarity runs
dataset.list_brain_runs(type=fob.Similarity)

# Only list specific similarity runs
dataset.list_brain_runs(
    type=fob.Similarity,
    patches_field="ground_truth",
    supports_prompts=True,
)

Or, you can use get_brain_info() to retrieve information about the configuration of a brain run:

info = dataset.get_brain_info(brain_key)
print(info)

Use load_brain_results() to load the SimilarityIndex instance for a brain run.

You can use rename_brain_run() to rename the brain key associated with an existing similarity results run:

dataset.rename_brain_run(brain_key, new_brain_key)

Finally, you can use delete_brain_run() to delete the record of a similarity index computation from your FiftyOne dataset:

dataset.delete_brain_run(brain_key)

Note

Calling delete_brain_run() only deletes the record of the brain run from your FiftyOne dataset; it will not delete any associated Qdrant collection, which you can do as follows:

# Delete the Qdrant collection
qdrant_index = dataset.load_brain_results(brain_key)
qdrant_index.cleanup()

Examples

This section demonstrates how to perform some common vector search workflows on a FiftyOne dataset using the Qdrant backend.

Note

All of the examples below assume you have configured your Qdrant server as described in this section.

Create a similarity index

In order to create a new Qdrant similarity index, you need to specify either the embeddings or model argument to compute_similarity(). Here’s a few possibilities:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")
model_name = "clip-vit-base32-torch"
model = foz.load_zoo_model(model_name)
brain_key = "qdrant_index"

# Option 1: Compute embeddings on the fly from model name
fob.compute_similarity(
    dataset,
    model=model_name,
    backend="qdrant",
    brain_key=brain_key,
)

# Option 2: Compute embeddings on the fly from model instance
fob.compute_similarity(
    dataset,
    model=model,
    backend="qdrant",
    brain_key=brain_key,
)

# Option 3: Pass precomputed embeddings as a numpy array
embeddings = dataset.compute_embeddings(model)
fob.compute_similarity(
    dataset,
    embeddings=embeddings,
    backend="qdrant",
    brain_key=brain_key,
)

# Option 4: Pass precomputed embeddings by field name
dataset.compute_embeddings(model, embeddings_field="embeddings")
fob.compute_similarity(
    dataset,
    embeddings="embeddings",
    backend="qdrant",
    brain_key=brain_key,
)

Note

You can customize the Qdrant collection by passing any supported parameters as extra kwargs.

Create a patch similarity index

You can also create a similarity index for object patches within your dataset by including the patches_field argument to compute_similarity():

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

fob.compute_similarity(
    dataset,
    patches_field="ground_truth",
    model="clip-vit-base32-torch",
    backend="qdrant",
    brain_key="qdrant_patches",
)

Note

You can customize the Qdrant collection by passing any supported parameters as extra kwargs.

Connect to an existing index

If you have already created a Qdrant collection storing the embedding vectors for the samples or patches in your dataset, you can connect to it by passing the collection_name to compute_similarity():

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",      # zoo model used (if applicable)
    embeddings=False,                   # don't compute embeddings
    collection_name="your-collection",  # the existing Qdrant collection
    brain_key="qdrant_index",
    backend="qdrant",
)

Add/remove embeddings from an index

You can use add_to_index() and remove_from_index() to add and remove embeddings from an existing Qdrant index.

These methods can come in handy if you modify your FiftyOne dataset and need to update the Qdrant index to reflect these changes:

import numpy as np

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

qdrant_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    brain_key="qdrant_index",
    backend="qdrant",
)
print(qdrant_index.total_index_size)  # 200

view = dataset.take(10)
ids = view.values("id")

# Delete 10 samples from a dataset
dataset.delete_samples(view)

# Delete the corresponding vectors from the index
qdrant_index.remove_from_index(sample_ids=ids)

# Add 20 samples to a dataset
samples = [fo.Sample(filepath="tmp%d.jpg" % i) for i in range(20)]
sample_ids = dataset.add_samples(samples)

# Add corresponding embeddings to the index
embeddings = np.random.rand(20, 512)
qdrant_index.add_to_index(embeddings, sample_ids)

print(qdrant_index.total_index_size)  # 210

Retrieve embeddings from an index

You can use get_embeddings() to retrieve embeddings from a Qdrant index by ID:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

qdrant_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    brain_key="qdrant_index",
    backend="qdrant",
)

# Retrieve embeddings for the entire dataset
ids = dataset.values("id")
embeddings, sample_ids, _ = qdrant_index.get_embeddings(sample_ids=ids)
print(embeddings.shape)  # (200, 512)
print(sample_ids.shape)  # (200,)

# Retrieve embeddings for a view
ids = dataset.take(10).values("id")
embeddings, sample_ids, _ = qdrant_index.get_embeddings(sample_ids=ids)
print(embeddings.shape)  # (10, 512)
print(sample_ids.shape)  # (10,)

Querying a Qdrant index

You can query a Qdrant index by appending a sort_by_similarity() stage to any dataset or view. The query can be any of the following:

  • An ID (sample or patch)

  • A query vector of same dimension as the index

  • A list of IDs (samples or patches)

  • A text prompt (if supported by the model)

import numpy as np

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    brain_key="qdrant_index",
    backend="qdrant",
)

# Query by vector
query = np.random.rand(512)  # matches the dimension of CLIP embeddings
view = dataset.sort_by_similarity(query, k=10, brain_key="qdrant_index")

# Query by sample ID
query = dataset.first().id
view = dataset.sort_by_similarity(query, k=10, brain_key="qdrant_index")

# Query by a list of IDs
query = [dataset.first().id, dataset.last().id]
view = dataset.sort_by_similarity(query, k=10, brain_key="qdrant_index")

# Query by text prompt
query = "a photo of a dog"
view = dataset.sort_by_similarity(query, k=10, brain_key="qdrant_index")

Note

Performing a similarity search on a DatasetView will only return results from the view; if the view contains samples that were not included in the index, they will never be included in the result.

This means that you can index an entire Dataset once and then perform searches on subsets of the dataset by constructing views that contain the images of interest.

Accessing the Qdrant client

You can use the client property of a Qdrant index to directly access the underlying Qdrant client instance and use its methods as desired:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

qdrant_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    brain_key="qdrant_index",
    backend="qdrant",
)

qdrant_client = qdrant_index.client
print(qdrant_client)
print(qdrant_client.get_collections())

Advanced usage

As previously mentioned, you can customize your Qdrant collections by providing optional parameters to compute_similarity().

In particular, the hnsw_config, wal_config, and optimizers_config parameters may impact the quality of your query results, as well as the time and memory required to perform approximate nearest neighbor searches. Additionally, you can specify parameters like replication_factor and shard_number to further tune performance.

Here’s an example of creating a similarity index backed by a customized Qdrant collection. Just for fun, we’ll specify a custom collection name, use dot product similarity, and populate the index for only a subset of our dataset:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

# Create a custom Qdrant index
qdrant_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    embeddings=False,  # we'll add embeddings below
    metric="dotproduct",
    brain_key="qdrant_index",
    backend="qdrant",
    collection_name="custom-quickstart-index",
    replication_factor=2,
    shard_number=2,
)

# Add embeddings for a subset of the dataset
view = dataset.take(10)
embeddings, sample_ids, _ = qdrant_index.compute_embeddings(view)
qdrant_index.add_to_index(embeddings, sample_ids)

qdrant_client = qdrant_index.client
print(qdrant_client.get_collections())