Skip to content

MongoDB Vector Search Integration

MongoDB is the leading open source database for unstructured data, and we’ve made it easy to use MongoDB Atlas’ vector search capabilities on your computer vision data directly from FiftyOne!

Follow these simple instructions to configure a MongoDB Atlas cluster and get started using MongoDB Atlas + FiftyOne.

FiftyOne provides an API to create MongoDB Atlas vector search indexes, 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 MongoDB similarity indexes!

image-similarity

Basic recipe

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

  1. Configure a MongoDB Atlas cluster

  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 MongoDB similarity index for the samples or object patches in a dataset by setting the parameter backend="mongodb" and specifying a brain_key of your choice

  5. Use this MongoDB 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 configure a MongoDB Atlas 7.0 or later cluster and provide its connection string to run this example:

export FIFTYONE_DATABASE_NAME=fiftyone
export FIFTYONE_DATABASE_URI='mongodb+srv://$USERNAME:$PASSWORD@fiftyone.XXXXXX.mongodb.net/?retryWrites=true&w=majority'

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
mongodb_index = fob.compute_similarity(
    dataset,
    embeddings="embeddings",  # the field in which to store the embeddings
    brain_key="mongodb_index",
    backend="mongodb",
)

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

# Wait for the index to be ready for querying...
assert mongodb_index.ready

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

# Step 5 (optional): Cleanup

# Delete the MongoDB vector search index
mongodb_index.cleanup()

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

Note

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

Setup

In order to use MongoDB vector search, you must connect your FiftyOne installation to MongoDB Atlas, which you can do by navigating to https://cloud.mongodb.com, creating an account, and following the instructions there to configure your cluster.

Note

You must be running MongoDB Atlas 7.0 or later in order to programmatically create vector search indexes ( source).

As of this writing, Atlas’ shared tier (M0, M2, M5) is running MongoDB 6. In order to use MongoDB 7, you must upgrade to an M10 cluster, which starts at $0.08/hour.

Configuring your connection string

You can connect FiftyOne to your MongoDB Atlas cluster by simply providing its connection string:

export FIFTYONE_DATABASE_NAME=fiftyone
export FIFTYONE_DATABASE_URI='mongodb+srv://$USERNAME:$PASSWORD@fiftyone.XXXXXX.mongodb.net/?retryWrites=true&w=majority'

Using the MongoDB backend

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

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

import fiftyone.brain as fob

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

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

export FIFTYONE_BRAIN_DEFAULT_SIMILARITY_BACKEND=mongodb

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

{
    "default_similarity_backend": "mongodb"
}

MongoDB config parameters

The MongoDB backend supports a variety of query parameters that can be used to customize your similarity queries. These parameters include:

  • index_name ( None): the name of the MongoDB vector search index to use or create. If not specified, a new unique name is generated automatically

  • metric ( “cosine”): the distance/similarity metric to use when creating a new index. The supported values are ("cosine", "dotproduct", "euclidean")

For detailed information on these parameters, see the MongoDB 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": {
        "mongodb": {
            "index_name": "your-index",
            "metric": "cosine"
        }
    }
}

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

mongodb_index = fob.compute_similarity(
    ...
    backend="mongodb",
    brain_key="mongodb_index",
    index_name="your-index",
    metric="cosine",
)

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 MongoDB vector search index, which you can do as follows:

# Delete the MongoDB vector search index
mongodb_index = dataset.load_brain_results(brain_key)
mongodb_index.cleanup()

Examples

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

Note

All of the examples below assume you have configured your MongoDB Atlas cluster as described in this section.

Create a similarity index

In order to create a new MongoDB 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 = "mongodb_index"

# Option 1: Compute embeddings on the fly from model name
fob.compute_similarity(
    dataset,
    model=model_name,
    embeddings="embeddings",  # the field in which to store the embeddings
    backend="mongodb",
    brain_key=brain_key,
)

# Option 2: Compute embeddings on the fly from model instance
fob.compute_similarity(
    dataset,
    model=model,
    embeddings="embeddings",  # the field in which to store the embeddings
    backend="mongodb",
    brain_key=brain_key,
)

# Option 3: Pass precomputed embeddings as a numpy array
embeddings = dataset.compute_embeddings(model)
fob.compute_similarity(
    dataset,
    embeddings=embeddings,
    embeddings_field="embeddings",  # the field in which to store the embeddings
    backend="mongodb",
    brain_key=brain_key,
)

# Option 4: Pass precomputed embeddings by field name
# Note that MongoDB vector indexes require list fields
embeddings = dataset.compute_embeddings(model)
dataset.set_values("embeddings", embeddings.tolist())
fob.compute_similarity(
    dataset,
    embeddings="embeddings",  # the field that contains the embeddings
    backend="mongodb",
    brain_key=brain_key,
)

Note

You can customize the MongoDB index by passing any supported parameters as extra kwargs.

Create a patch similarity index

Warning

The MongoDB backend does not yet support indexing object patches, so the code below will not yet run. Check back soon!

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",
    embeddings="embeddings",  # the attribute in which to store the embeddings
    backend="mongodb",
    brain_key="mongodb_patches",
)

Note

You can customize the MongoDB index by passing any supported parameters as extra kwargs.

Connect to an existing index

If you have already created a MongoDB index storing the embedding vectors for the samples or patches in your dataset, you can connect to it by passing the index_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
    index_name="your-index",            # the existing MongoDB index
    brain_key="mongodb_index",
    backend="mongodb",
)

Add/remove embeddings from an index

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

These methods can come in handy if you modify your FiftyOne dataset and need to update the Mongodb 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")

mongodb_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    embeddings="embeddings",  # the field in which to store the embeddings
    brain_key="mongodb_index",
    backend="mongodb",
)
print(mongodb_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
mongodb_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)
mongodb_index.add_to_index(embeddings, sample_ids)

print(mongodb_index.total_index_size)  # 210

Retrieve embeddings from an index

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

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

dataset = foz.load_zoo_dataset("quickstart")

mongodb_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    embeddings="embeddings",  # the field in which to store the embeddings
    brain_key="mongodb_index",
    backend="mongodb",
)

# Retrieve embeddings for the entire dataset
ids = dataset.values("id")
embeddings, sample_ids, _ = mongodb_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, _ = mongodb_index.get_embeddings(sample_ids=ids)
print(embeddings.shape)  # (10, 512)
print(sample_ids.shape)  # (10,)

Querying a MongoDB index

You can query a MongoDB 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")

mongodb_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    embeddings="embeddings",  # the field in which to store the embeddings
    brain_key="mongodb_index",
    backend="mongodb",
)

# Wait for the index to be ready for querying...
assert mongodb_index.ready

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

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

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

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

Note

Currently, when performing a similarity search on a view with the MongoDB backend, the full index is queried and the resulting samples are restricted to the desired view. This may result in fewer samples than requested being returned by the search.

Checking if an index is ready

You can use the ready property of a MongoDB index to check whether a newly created vector search index is ready for querying:

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

dataset = foz.load_zoo_dataset("quickstart")

mongodb_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    embeddings="embeddings",  # the field in which to store the embeddings
    brain_key="mongodb_index",
    backend="mongodb",
)

# Wait for the index to be ready for querying...
assert mongodb_index.ready

Advanced usage

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

Here’s an example of creating a similarity index backed by a customized MongoDB index. Just for fun, we’ll specify a custom index 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 MongoDB index
mongodb_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    embeddings_field="embeddings",  # the field in which to store the embeddings
    embeddings=False,               # add embeddings later
    brain_key="mongodb_index",
    backend="mongodb",
    index_name="custom-quickstart-index",
    metric="dotproduct",
)

# Add embeddings for a subset of the dataset
view = dataset[:20]
embeddings, sample_ids, _ = mongodb_index.compute_embeddings(view)
mongodb_index.add_to_index(embeddings, sample_ids)

print(mongodb_index.total_index_size)  # 20
print(mongodb_index.config.index_name)  # custom-quickstart-index
print(mongodb_index.config.metric)  # dotproduct