API Reference¶

Information on the search methods (and their parameters) and spaces, can be found on this page.

nmslib.init¶

This function acts act the main entry point into NMS lib. This function should be called first before calling any other method.

nmslib.init(space: str='cosinesimil', space_params: object=None, method: str='hnsw', data_type: nmslib.DataType=DataType.DENSE_VECTOR, dtype: nmslib.DistType=DistType.FLOAT) → object¶

This function initializes a new NMSLIB index

Parameters:	space (str optional) – The metric space to create for this index space_params (dict optional) – Parameters for configuring the space method (str optional) – The index method to use data_type (nmslib.DataType optional) – The type of data to index (dense/sparse/string vectors)
Returns:
Return type:	A new NMSLIB Index.

class nmslib.DistType¶

FLOAT¶

INT¶

class nmslib.DataType¶

DENSE_VECTOR¶

OBJECT_AS_STRING¶

SPARSE_VECTOR¶

nmslib.FloatIndex¶

nmslib.dist.FloatIndex

class nmslib.dist.FloatIndex¶

addDataPoint(self: nmslib.dist.FloatIndex, id: int, data: object) → int¶

Adds a single datapoint to the index

Parameters:	id (int) – The id of the object to add data (object) – The object to add to the index.
Returns:	The position the item was added at
Return type:	int

addDataPointBatch(self: nmslib.dist.FloatIndex, data: object, ids: object=None) → int¶

Adds multiple datapoints to the index

Parameters:	data (object) – The objects to add to the index. ids (array_like optional) – The ids of the object being inserted. If not set will default to the row id of each object in the dataset
Returns:	The number of items added
Return type:	int

createIndex(self: nmslib.dist.FloatIndex, index_params: object=None, print_progress: bool=False) → None¶

Creates the index, and makes it available for querying

Parameters:	index_params (dict optional) – Dictionary of optional parameters to use in indexing print_progress (bool optional) – Whether or not to display progress bar when creating index

getDistance(self: nmslib.dist.FloatIndex, arg0: int, arg1: int) → float¶

knnQuery(self: nmslib.dist.FloatIndex, vector: object, k: int=10) → object¶

Finds the approximate K nearest neighbours of a vector in the index

Parameters:

vector (array_like) – A 1D vector to query for.
k (int optional) – The number of neighbours to return

Returns:

ids (array_like.) – A 1D vector of the ids of each nearest neighbour.
distances (array_like.) – A 1D vector of the distance to each nearest neigbhour.

knnQueryBatch(self: nmslib.dist.FloatIndex, queries: object, k: int=10, num_threads: int=0) → object¶

Performs multiple queries on the index, distributing the work over a thread pool

Parameters:	input (list) – A list of queries to query for k (int optional) – The number of neighbours to return num_threads (int optional) – The number of threads to use
Returns:	A list of tuples of (ids, distances)
Return type:	list

loadIndex(self: nmslib.dist.FloatIndex, filename: str, load_data: bool=False) → None¶

Loads the index from disk

Parameters:	filename (str) – The filename to read from load_data (bool optional) – Whether or not to load previously saved data.

saveIndex(self: nmslib.dist.FloatIndex, filename: str, save_data: bool=False) → None¶

Saves the index and/or data to disk

Parameters:	filename (str) – The filename to save to save_data (bool optional) – Whether or not to save data

setQueryTimeParams(self: nmslib.dist.FloatIndex, params: object=None) → None¶

Sets parameters used in knnQuery.

Parameters:	params (dict) – A dictionary of params to use in querying. Setting params to None will reset

nmslib.IntIndex¶

nmslib.dist.IntIndex

class nmslib.dist.IntIndex¶

addDataPoint(self: nmslib.dist.IntIndex, id: int, data: object) → int¶

Adds a single datapoint to the index

Parameters:	id (int) – The id of the object to add data (object) – The object to add to the index.
Returns:	The position the item was added at
Return type:	int

addDataPointBatch(self: nmslib.dist.IntIndex, data: object, ids: object=None) → int¶

Adds multiple datapoints to the index

Parameters:	data (object) – The objects to add to the index. ids (array_like optional) – The ids of the object being inserted. If not set will default to the row id of each object in the dataset
Returns:	The number of items added
Return type:	int

createIndex(self: nmslib.dist.IntIndex, index_params: object=None, print_progress: bool=False) → None¶

Creates the index, and makes it available for querying

Parameters:	index_params (dict optional) – Dictionary of optional parameters to use in indexing print_progress (bool optional) – Whether or not to display progress bar when creating index

getDistance(self: nmslib.dist.IntIndex, arg0: int, arg1: int) → int¶

knnQuery(self: nmslib.dist.IntIndex, vector: object, k: int=10) → object¶

Finds the approximate K nearest neighbours of a vector in the index

Parameters:

vector (array_like) – A 1D vector to query for.
k (int optional) – The number of neighbours to return

Returns:

ids (array_like.) – A 1D vector of the ids of each nearest neighbour.
distances (array_like.) – A 1D vector of the distance to each nearest neigbhour.

knnQueryBatch(self: nmslib.dist.IntIndex, queries: object, k: int=10, num_threads: int=0) → object¶

Performs multiple queries on the index, distributing the work over a thread pool

Parameters:	input (list) – A list of queries to query for k (int optional) – The number of neighbours to return num_threads (int optional) – The number of threads to use
Returns:	A list of tuples of (ids, distances)
Return type:	list

loadIndex(self: nmslib.dist.IntIndex, filename: str, load_data: bool=False) → None¶

Loads the index from disk

Parameters:	filename (str) – The filename to read from load_data (bool optional) – Whether or not to load previously saved data.

saveIndex(self: nmslib.dist.IntIndex, filename: str, save_data: bool=False) → None¶

Saves the index and/or data to disk

Parameters:	filename (str) – The filename to save to save_data (bool optional) – Whether or not to save data

setQueryTimeParams(self: nmslib.dist.IntIndex, params: object=None) → None¶

Sets parameters used in knnQuery.

Parameters:	params (dict) – A dictionary of params to use in querying. Setting params to None will reset