MongoDB Atlas Vector Search

Overview

The MongoDB - Atlas Vector Search Snap enables you to efficiently perform advanced vector-based queries. This Snap is useful for Similarity searches, Approximate Nearest Neighbor (ANN) queries, and Range queries on vector data stored in MongoDB Atlas. Learn more about vector search queries.

MongoDB Atlas Vector Search Snap in pipeline

Note: You can run MongoDB Atlas Vector Search Snap in Groundplex or Cloudplex via the MongoDB Atlas service.

Supported Accounts

Configure MongoDB Accounts

This is a Read-type Snap.

Prerequisites

Create a MongoDB Atlas account.
Create a vector search index in MongoDB Atlas.
Configure MongoDB Atlas cluster and allow your IP in network security settings.

Limitations

The listSearchIndexes command can only be run on a deployment hosted on MongoDB Atlas, and requires an Atlas cluster tier of at least M10. Hence, for the suggestions list to populate in the Search index field, you must deploy the MongoDB Atlas cluster with at least an M10 tier. However, you can add the index manually, even if the Snap displays an error when populating the suggestions.

Known issues

None.

Snap views


Type	Description	Examples of upstream and downstream Snaps
Input	Requires the MongoDB aggregate command to perform vector search.	Join JSON Generator
Output	Retrieves query results from the collection as specified in the settings.	Mapper Copy
Learn more about Error handling.

Snap settings

Note: Learn about the common controls in the Snap settings dialog.


Field/Field set	Description
Label `String`	Required.Specify a unique name for the Snap. Modify this to be more appropriate, especially if there are more than one of the same Snap in the pipeline. Default value: MongoDB - Atlas Vector Search Example: `Atlas Vector Search`
Database name `String/Expression/ Suggestion`	Specify the database where the query is executed. If you do not specify, the database configured in the account is used. Default value: N/A Example: `chunking_strategy`
Collection name `String/Expression/ Suggestion`	Required. Specify or select a MongoDB collection name to find indexes for vector search. Default value: N/A Example: `chunk_metadata`
Search index `String/Expression/ Suggestion`	Required. Specify the name of the vector search index to query. Alternatively, click the Suggestions icon to select the vector search index. Note: For search index suggestions to work, you must deploy the MongoDB Atlas cluster with at least an M10 tier. Otherwise, you can add the index manually, even if the Snap displays an error when you click the Suggestion icon. Default value: N/A Example: `vector_index`
Vector field `String/Expression/ Suggestion`	Required. Specify the name of the vector field that you want to search. Default value: N/A Example: `embedding`
Number of candidates `Integer/Expression`	Required. Specify the number of candidates for vector search. Note: The Number of candidates value must be higher than the Limit value to increase the accuracy of the results. Maximum value `10000` Minimum value: `0` Default value: `100` Example: `50`
Limit `Integer/Expression`	Required. Specify the number of results to return for each query. Maximum value `10000` Minimum value: `0` Default value: `4` Example: `10`
Advanced settings	Optional configuration for batch processing, timezone offset, and retry behavior.
Batch size `Integer`	Required. Specify the number of documents to return per batch. Default value: `0` Example: `10`
Timezone hours offset `Integer`	Specify the time zone hour offset to apply to all DateTime fields. Note: The values for this field must be in the range of `-12` through `14`. Default value: `0` Example: `12`
Timezone minutes offset `Integer`	Specify the time zone minutes offset to apply to all DateTime fields. Note: The values for this field must be in the range of `0` through `59`. Default value: `0` Example: `1`
Number of retries `Integer/Expression`	Specify the maximum number of attempts to be made to receive a response. The request is terminated if the attempts do not result in a response. Note: If the Number of retries value is set to `0` (the default value), the retry option is disabled, and the Snap does not initiate a retry. The pipeline will not attempt to retry the operation in case of a failure—any failure encountered during the database operation will immediately result in the pipeline failing without any retry attempts to recover from the errors. If the Snap fails on all retries, it routes the last occurred exception to the error view. Default value: `0` Example: `4`
Retry interval (seconds) `Integer/Expression`	Specify the time interval between two retry requests. Default value: `1` Example: `5`
Snap execution `Dropdown list`	Choose one of the three modes in which the Snap executes. Available options are: `Validate & Execute`: Performs limited execution of the Snap and generates a data preview during pipeline validation. Subsequently, performs full execution of the Snap (unlimited records) during pipeline runtime. `Execute only`: Performs full execution of the Snap during pipeline execution without generating preview data. `Disabled`: Disables the Snap and all Snaps that are downstream from it. Default value: `Validate & Execute` Example: `Execute only`

Temporary files

During execution, data processing on Snaplex nodes occurs principally in-memory as streaming and is unencrypted. When processing larger datasets that exceed the available compute memory, the Snap writes unencrypted pipeline data to local storage to optimize the performance. These temporary files are deleted when the pipeline execution completes. You can configure the temporary data's location in the Global properties table of the Snaplex node properties, which can also help avoid pipeline errors because of the unavailability of space. Learn more about Temporary Folder in Configuration Options.

Troubleshooting


Error	Reason	Resolution
`The vector search operation failed.`	The specified vector might be invalid.	Specify a valid vector and retry.
`Unable to process the input document.`	The input document might be invalid.	Specify valid input and retry.
`The collection provided cannot be found.`	The specified collection is not available.	Specify a valid collection and retry.
`The index is not specified.`	The index field is mandatory.	Specify a valid index and retry.
`The input document does not have a 'vector' specified in the structure.`	The input document must have a valid vector.	Add the vector to query the input document and retry.
`Unable to retrieve index fields` `No indexes were found querying database <name>` `Command failed with error 8000 (AtlasError)`	Unrecognized pipeline stage <name-of-the-pipeline>.	Ensure the cluster runs on MongoDB Atlas and has a cluster tier of at least M10. Additionally, confirm that the user has the required permissions to access the search index.