> ## Documentation Index > Fetch the complete documentation index at: https://mintlify.com/qdrant/qdrant/llms.txt > Use this file to discover all available pages before exploring further. # Search API > REST API endpoints for vector similarity search, recommendations, and query operations in Qdrant Qdrant provides powerful search capabilities including vector similarity search, recommendation systems, and flexible query APIs. These endpoints help you find similar vectors and implement semantic search. ## Search Points Retrieve the closest points based on vector similarity and optional filtering conditions. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/search' \ -H 'Content-Type: application/json' \ -d '{ "vector": [0.2, 0.1, 0.9, 0.7], "limit": 3 }' ``` **Path Parameters** Name of the collection to search in **Query Parameters** Read consistency level: `majority`, `quorum`, or `all` Request timeout in seconds **Request Body** Vector to search for. Can be a dense vector array, sparse vector, or named vector. Maximum number of results to return. Default is 10. Offset for pagination. Default is 0. Filter conditions to apply All conditions must match At least one condition must match All conditions must NOT match Search parameters HNSW graph exploration factor. Higher values = more accurate but slower. If true, perform exact search instead of approximate Quantization search parameters Only return results with a score above this threshold Whether to return payload. Can be `true`, `false`, array of field names, or include/exclude object. Whether to return vectors. Can be `true`, `false`, or array of vector names. **Response** Array of scored points Point ID Similarity score Point payload (if requested) Point vector (if requested) ## Batch Search Perform multiple search queries in a single request for better performance. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/search/batch' \ -H 'Content-Type: application/json' \ -d '{ "searches": [ { "vector": [0.2, 0.1, 0.9, 0.7], "limit": 3 }, { "vector": [0.5, 0.3, 0.2, 0.3], "limit": 3, "filter": { "must": [{"key": "city", "match": {"value": "London"}}] } } ] }' ``` **Path Parameters** Name of the collection to search in **Query Parameters** Read consistency level Request timeout in seconds **Request Body** Array of search requests. Each search has the same parameters as the single search endpoint. **Response** Array of result arrays, one for each search query ## Search Groups Search for points grouped by a specific payload field. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/search/groups' \ -H 'Content-Type: application/json' \ -d '{ "vector": [0.2, 0.1, 0.9, 0.7], "group_by": "city", "limit": 3, "group_size": 2 }' ``` **Path Parameters** Name of the collection to search in **Query Parameters** Read consistency level Request timeout in seconds **Request Body** Vector to search for Payload field to group results by Maximum number of groups to return Maximum number of points per group Filter conditions Whether to return payload Whether to return vectors Join payload from another collection **Response** Array of groups Group identifier (value of the group\_by field) Points in this group ## Recommend Points Get recommendations based on positive and negative example points. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/recommend' \ -H 'Content-Type: application/json' \ -d '{ "positive": [1, 2], "negative": [3], "limit": 5 }' ``` **Path Parameters** Name of the collection **Query Parameters** Read consistency level Request timeout in seconds **Request Body** Array of point IDs or vectors to use as positive examples Array of point IDs or vectors to use as negative examples Maximum number of results to return Offset for pagination Filter conditions Vector name to use for recommendation (for named vectors) Lookup positive/negative examples from another collection Minimum score threshold Whether to return payload Whether to return vectors **Response** Array of scored points with id, score, payload, and vector ## Recommend Batch Perform multiple recommendation queries in a single request. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/recommend/batch' \ -H 'Content-Type: application/json' \ -d '{ "searches": [ { "positive": [1, 2], "limit": 3 }, { "positive": [5, 6], "negative": [3], "limit": 5 } ] }' ``` **Path Parameters** Name of the collection **Query Parameters** Read consistency level Request timeout in seconds **Request Body** Array of recommendation requests **Response** Array of result arrays, one for each recommendation query ## Recommend Groups Get recommendations grouped by a payload field. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/recommend/groups' \ -H 'Content-Type: application/json' \ -d '{ "positive": [1, 2], "group_by": "category", "limit": 3, "group_size": 2 }' ``` Parameters are similar to search groups, but with positive/negative examples instead of a query vector. ## Query Points Unified query API that supports multiple query types including nearest neighbors, discovery, and context-based search. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/query' \ -H 'Content-Type: application/json' \ -d '{ "query": [0.2, 0.1, 0.9, 0.7], "limit": 10 }' ``` **Path Parameters** Name of the collection **Query Parameters** Read consistency level Request timeout in seconds **Request Body** Query specification. Can be: * Dense vector array for nearest neighbor search * Object with `nearest` for vector search * Object with `recommend` for recommendation * Object with `discover` for discovery search * Object with `context` for context-based search * Object with `order_by` for payload-based ordering Maximum number of results Offset for pagination Filter conditions Minimum score threshold Whether to return payload Whether to return vectors **Response** Array of scored points ## Query Batch Perform multiple query operations in a single request. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/query/batch' \ -H 'Content-Type: application/json' \ -d '{ "searches": [ { "query": [0.2, 0.1, 0.9, 0.7], "limit": 3 }, { "query": {"recommend": {"positive": [1, 2]}}, "limit": 5 } ] }' ``` **Path Parameters** Name of the collection **Query Parameters** Read consistency level Request timeout in seconds **Request Body** Array of query requests **Response** Array of query results ## Query Groups Perform query operations with results grouped by a payload field. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/query/groups' \ -H 'Content-Type: application/json' \ -d '{ "query": [0.2, 0.1, 0.9, 0.7], "group_by": "category", "limit": 3, "group_size": 2 }' ``` **Path Parameters** Name of the collection **Query Parameters** Read consistency level Request timeout in seconds **Request Body** Query specification (same as query points) Payload field to group by Maximum number of groups Maximum points per group Filter conditions Whether to return payload Whether to return vectors **Response** Array of groups with grouped query results ## Search Matrix (Pairs) Compute similarity scores between all pairs of points. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/search/matrix/pairs' \ -H 'Content-Type: application/json' \ -d '{ "sample": 100, "limit": 10 }' ``` **Path Parameters** Name of the collection **Request Body** Number of points to sample for matrix computation Number of neighbors per point Filter to select points **Response** Array of point pairs with similarity scores ## Search Matrix (Offsets) Compute similarity matrix in offset format. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/search/matrix/offsets' \ -H 'Content-Type: application/json' \ -d '{ "sample": 100, "limit": 10 }' ``` Parameters are identical to search matrix pairs, but returns results in a compressed offset format suitable for large matrices. ## Count Points Count points in a collection that match a given filter. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/count' \ -H 'Content-Type: application/json' \ -d '{ "filter": { "must": [ { "key": "city", "match": { "value": "London" } } ] }, "exact": true }' ``` **Path Parameters** Name of the collection to count in **Query Parameters** Read consistency level: `majority`, `quorum`, or `all` Request timeout in seconds **Request Body** Filter conditions to apply. Only points matching this filter will be counted. If true, provide exact count. If false, may provide approximate count for better performance. **Response** Number of points matching the filter ## Facet Count points for each unique value of a payload key, optionally with filtering. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/facet' \ -H 'Content-Type: application/json' \ -d '{ "key": "city", "limit": 10 }' ``` **Path Parameters** Name of the collection to facet in **Query Parameters** Read consistency level: `majority`, `quorum`, or `all` Request timeout in seconds **Request Body** Payload key to facet by Maximum number of unique values to return. Default is 10. Filter conditions to apply before faceting If true, provide exact counts. If false, may use approximations for better performance. **Response** Array of facet results The unique value of the payload key Number of points with this value **Example: Count products by category** ```bash theme={null} curl -X POST 'http://localhost:6333/collections/products/facet' \ -H 'Content-Type: application/json' \ -d '{ "key": "category", "limit": 20, "filter": { "must": [ { "key": "in_stock", "match": { "value": true } } ] } }' ``` ## Discover Points Look for points that satisfy context pairs (positive/negative examples) and optionally approach a target. ```bash theme={null} curl -X POST 'http://localhost:6333/collections/{collection_name}/points/discover' \ -H 'Content-Type: application/json' \ -d '{ "context": [ { "positive": 100, "negative": 718 }, { "positive": 200, "negative": 300 } ], "target": [0.2, 0.1, 0.9, 0.7], "limit": 10 }' ``` **Path Parameters** Name of the collection to search in **Query Parameters** Read consistency level: `majority`, `quorum`, or `all` Request timeout in seconds **Request Body** Array of positive-negative example pairs. Each pair consists of a `positive` point ID and a `negative` point ID. Results will be points that are more similar to positives than negatives. Optional target vector to approach. Can be a dense vector or named vector. Maximum number of results to return. Default is 10. Filter conditions to apply Whether to return payload Whether to return vectors **Response** Array of scored points. The score combines context ranking and distance to target. Point ID Combined score. Integer part represents context rank, decimal part represents distance to target. Point payload (if requested) **Use Case: Analogical Search** Find items that have a similar relationship to the target as the positive/negative pairs demonstrate. ```bash theme={null} # Find images similar to "night" as "day" is to "beach" curl -X POST 'http://localhost:6333/collections/images/points/discover' \ -H 'Content-Type: application/json' \ -d '{ "context": [ { "positive": 1, # "day" image "negative": 2 # "beach" image } ], "target": 3, # "night" image "limit": 5 }' ```