Actions26
- Collection Actions
- Payload Actions
- Point Actions
- Search Actions
- Vector Actions
Overview
This node operation allows querying a Qdrant collection to retrieve points grouped by a specified payload field. It is useful when you want to perform vector similarity searches and then organize the results into groups based on a particular attribute in the payload, such as category, type, or any other metadata field.
Common scenarios include:
- Searching for similar items (e.g., images, documents) and grouping them by their category or tag.
- Performing recommendation queries where results are clustered by user segments or product types.
- Analyzing search results with aggregation on specific fields to better understand distribution.
Example: Query a collection of product embeddings with a vector query, group results by "brand" field, and get top 3 products per brand that match the query vector.
Properties
| Name | Meaning |
|---|---|
| Collection Name | The target collection to query. Can be selected from a list of existing collections or specified by name. |
| Group By | The payload field name used to group the returned points. Must be a string or number field present in the payload. |
| Query | The query vector or complex query object to perform the search. Examples include raw vectors (arrays of numbers), IDs, or advanced query objects like recommend, fusion, sample, or formula queries. |
| Using | Specifies which named vector field in the collection to use for the query if multiple vector fields exist. |
| Filter | JSON object defining conditions to filter points before grouping and returning. Only points satisfying these conditions will be considered. |
| Params | Additional search parameters applied when there is no prefetch step. Used to fine-tune the search behavior. |
| Score Threshold | Minimum score threshold; only points with scores better than this value will be included in the results. |
| With Vector | Boolean flag indicating whether to include the vector data of each point in the response. |
| With Payload | Boolean flag indicating whether to include the payload data of each point in the response. |
| Lookup From | JSON specifying an alternate location or collection to look up IDs from during the query. |
| Group Size | Maximum number of points to return per group. Controls how many points are included for each distinct group value. |
| Limit | Maximum total number of results to return across all groups. |
| With Lookup | JSON configuration to enable looking for points in another collection using the group IDs obtained from the main query. |
| Prefetch | JSON describing sub-requests to perform before the main query. If provided, the main query operates on the results of these prefetches, allowing chained or dependent queries. |
| Shard Key | JSON specifying which shards to search within. If omitted, all shards are searched. Useful for sharded collections to limit scope. |
| Consistency | Read consistency level for the query operation. Options are Majority, Quorum, or All, defining how many replicas must respond for the read to be considered successful. |
| Timeout | Request timeout in seconds, overriding the global timeout setting for this specific request. |
Output
The output contains a JSON array of grouped points matching the query criteria. Each group corresponds to a unique value of the specified "Group By" field and includes up to "Group Size" points.
Each point in the result may include:
- Its unique ID.
- The payload data (if
With Payloadis true). - The vector data (if
With Vectoris true). - The similarity score relative to the query.
If binary data were involved (not indicated here), it would typically represent raw vector bytes or media content, but this operation focuses on JSON structured data.
Dependencies
- Requires connection to a Qdrant vector search service endpoint.
- Needs an API key or authentication token configured in n8n credentials to authorize requests.
- The node uses the Qdrant REST API to perform search and grouping operations.
- Properly configured collection(s) with vector fields and payload data are necessary.
Troubleshooting
- Empty Results: Ensure the collection name is correct and the query vector or object is valid. Also verify that the "Group By" field exists in the payload and is indexed.
- Invalid Query Format: The "Query" property expects a valid JSON array or object. Malformed JSON or unsupported query types will cause errors.
- Timeouts: If queries take too long, consider increasing the "Timeout" property or optimizing the collection indexing and shard usage.
- Score Threshold Too High: Setting a very high score threshold might filter out all results. Adjust accordingly.
- Authentication Errors: Verify that the API key credential is correctly set up and has permissions to access the target collection.
- Shard Key Misconfiguration: Specifying incorrect shard keys can lead to missing data or errors. Use shard keys only if you understand your collection's sharding setup.