# API

This document covers the Elastiknn API, including: settings, REST API payloads, all approximate similarity models, some common patterns, and some nice-to-know implementation details.

## Elasticsearch Settings

Elastiknn offers the following settings:

Setting | Scope | Required | Default | Description |
---|---|---|---|---|

`elastiknn.jdk-incubator-vector.enabled` |
Cluster | false | false | If true, Elastiknn will use SIMD vector operations from the jdk.incubator.vector module. Note that you likely also need to pass `--add-modules=jdk.incubator.vector` to the JVM, e.g., using the `ES_JAVA_OPTS` environment variable. |

## Vectors

You need to specify vectors when indexing documents and when running queries. In both cases you use the same JSON structure to define vectors. Each vector also has a shorthand alternative, which can be convenient when using tools that don’t support nested documents. The examples below show how to specify vectors when indexing them. The format for specifying vectors in queries is covered later.

### elastiknn_dense_float_vector

This assumes you’ve defined a mapping where `my_vec`

has type `elastiknn_dense_float_vector`

.

```
POST /my-index/_doc
{
"my_vec": {
"values": [0.1, 0.2, 0.3, ...] # 1
}
}
```

```
POST /my-index/_doc
{
"my_vec": [0.1, 0.2, 0.3, ...] # 2
}
```

# | Description |
---|---|

1 | JSON list of all floating point values in your vector. The length should match the `dims` in your mapping. |

2 | Shorthand alternative to #1. |

### elastiknn_sparse_bool_vector

This assumes you’ve defined a mapping where `my_vec`

has type `elastiknn_sparse_bool_vector`

.

```
POST /my-index/_doc
{
"my_vec": {
"true_indices": [1, 3, 5, ...], # 1
"total_indices": 100, # 2
}
}
```

```
POST /my-index/_doc
{
"my_vec": [[1, 3, 5, ...], 100] # 3
}
```

# | Description |
---|---|

1 | JSON list of the indices which are `true` in your vector. |

2 | The total number of indices in your vector. This should match the `dims` in your mapping. |

3 | Shorthand alternative to #1 and #2. A two-item list where the first item is the `true_indices` and the second is the `total_indices` . |

## Mappings

Before indexing vectors, you first define a mapping specifying a vector datatype, an indexing model, and the model’s parameters. This determines which queries are supported for the indexed vectors.

### General Structure

The general mapping structure looks like this:

```
PUT /my-index/_mapping
{
"properties": { # 1
"my_vec": { # 2
"type": "elastiknn_sparse_bool_vector", # 3
"elastiknn": { # 4
"dims": 100, # 5
"model": "exact", # 6
... # 7
}
}
}
}
```

# | Description |
---|---|

1 | Dictionary of document fields. Same as the official PUT Mapping API. |

2 | Name of the field containing your vector. This is arbitrary and can be nested under other fields. |

3 | Type of vector you want to store. See datatypes below. |

4 | Dictionary of elastiknn settings. |

5 | Dimensionality of your vector. All vectors stored at this field (`my_vec` ) must have the same dimensionality. |

6 | Model type. This and the model parameters will determine what kind of searches you can run. See more on models below. |

7 | Additional model parameters. See models below. |

### elastiknn_sparse_bool_vector Datatype

This type is optimized for vectors where each index is either `true`

or `false`

and the majority of indices are `false`

.
For example, you might represent a bag-of-words encoding of a document, where each index corresponds to a word in a vocabulary and any single document contains a very small fraction of all words.
Internally, Elastiknn saves space by only storing the true indices.

```
PUT /my-index/_mapping
{
"properties": {
"my_vec": {
"type": "elastiknn_sparse_bool_vector", # 1
"elastiknn": {
"dims": 25000, # 2
... # 3
}
}
}
}
```

# | Description |
---|---|

1 | Type name. |

2 | Dimensionality of the vector. This is the total number of possible indices. |

3 | Aditional model parameters. See models below. |

### elastiknn_dense_float_vector Datatype

This type is optimized for vectors where each index is a floating point number, all of the indices are populated, and the dimensionality usually doesn’t exceed ~1000. For example, you might store a word embedding or an image vector. Internally, Elastiknn uses Java Floats to store the values.

```
PUT /my-index/_mapping
{
"properties": {
"my_vec": {
"type": "elastiknn_dense_float_vector", # 1
"elastiknn": {
"dims": 100, # 2
... # 3
}
}
}
}
```

# | Description |
---|---|

1 | Type name. |

2 | Dimensionality of the vector. This shouldn’t exceed single-digit thousands. If it does, consider doing some sort of dimensionality reduction. |

3 | Aditional model parameters. See models below. |

### Exact Mapping

The exact model will allow you to run exact searches.
These don’t leverage any indexing constructs and have `O(n^2)`

runtime, where `n`

is the total number of documents.

You don’t need to supply any `"model": "..."`

value or any model parameters to use this model.

```
PUT /my-index/_mapping
{
"properties": {
"my_vec": {
"type": "elastiknn_(dense_float | sparse_bool)_vector", # 1
"elastiknn": {
"dims": 100, # 2
}
}
}
}
```

# | Description |
---|---|

1 | Vector datatype. Both dense float and sparse bool are supported |

2 | Vector dimensionality. |

### Jaccard LSH Mapping

Uses the Minhash algorithm to hash and store sparse bool vectors such that they support approximate Jaccard similarity queries.

The implementation is influenced by Chapter 3 of Mining Massive Datasets., the Spark MinHash implementation, the tdebatty/java-LSH Github project, and the Minhash for Dummies blog post.

```
PUT /my-index/_mapping
{
"properties": {
"my_vec": {
"type": "elastiknn_sparse_bool_vector", # 1
"elastiknn": {
"dims": 25000, # 2
"model": "lsh", # 3
"similarity": "jaccard", # 4
"L": 99, # 5
"k": 1 # 6
}
}
}
}
```

# | Description |
---|---|

1 | Vector datatype. Must be sparse bool vector. |

2 | Vector dimensionality. |

3 | Model type. |

4 | Similarity. |

5 | Number of hash tables. Generally, increasing this value increases recall. |

6 | Number of hash functions combined to form a single hash value. Generally, increasing this value increases precision. |

### Hamming LSH Mapping

Uses the Bit-Sampling algorithm to hash and store sparse bool vectors such that they support approximate Hamming similarity queries.

Only difference from the canonical bit-sampling method is that it samples and combines `k`

bits to form a single hash value.
For example, if you set `L = 100, k = 3`

, it samples `100 * 3 = 300`

bits from the vector and concatenates sets of 3
bits to form each hash value, for a total of 100 hash values.

```
PUT /my-index/_mapping
{
"properties": {
"my_vec": {
"type": "elastiknn_sparse_bool_vector", # 1
"elastiknn": {
"dims": 25000, # 2
"model": "lsh", # 3
"similarity": "hamming", # 4
"L": 99, # 5
"k": 2
}
}
}
}
```

# | Description |
---|---|

1 | Vector datatype. Must be sparse bool vector. |

2 | Vector dimensionality. |

3 | Model type. |

4 | Similarity. |

5 | Number of hash tables. Generally, increasing this value increases recall. |

6 | Number of hash functions combined to form a single hash value. Generally, increasing this value increases precision. |

### Cosine LSH Mapping

Uses the Random Projection algorithm
to hash and store dense float vectors such that they support approximate Cosine similarity queries.^{1}

The implementation is influenced by Chapter 3 of Mining Massive Datasets.

```
PUT /my-index/_mapping
{
"properties": {
"my_vec": {
"type": "elastiknn_dense_float_vector", # 1
"elastiknn": {
"dims": 100, # 2
"model": "lsh", # 3
"similarity": "cosine", # 4
"L": 99, # 5
"k": 1 # 6
}
}
}
}
```

# | Description |
---|---|

1 | Vector datatype. Must be dense float vector. |

2 | Vector dimensionality. |

3 | Model type. |

4 | Similarity. |

5 | Number of hash tables. Generally, increasing this value increases recall. |

6 | Number of hash functions combined to form a single hash value. Generally, increasing this value increases precision. |

### L2 LSH Mapping

Uses the Stable Distributions method to hash and store dense float vectors such that they support approximate L2 (Euclidean) similarity queries.

The implementation is influenced by Chapter 3 of Mining Massive Datasets.

```
PUT /my-index/_mapping
{
"properties": {
"my_vec": {
"type": "elastiknn_dense_float_vector", # 1
"elastiknn": {
"dims": 100, # 2
"model": "lsh", # 3
"similarity": "l2", # 4
"L": 99, # 5
"k": 1, # 6
"w": 3 # 7
}
}
}
}
```

# | Description |
---|---|

1 | Vector datatype. Must be dense float vector. |

2 | Vector dimensionality. |

3 | Model type. |

4 | Similarity. |

5 | Number of hash tables. Generally, increasing this value increases recall. |

6 | |

7 | Integer bucket width. This determines how close two vectors have to be, when projected onto a third common vector, in order for the two vectors to share a hash value. Typical values are low single-digit integers. |

### Permutation LSH Mapping

Uses the model described in Large-Scale Image Retrieval with Elasticsearch by Amato, et. al..

This model describes a vector by the `k`

indices (*positions in the vector*) with the greatest absolute values.
The intuition is that each index corresponds to some latent concept, and indices with high absolute values carry more
information about their respective concepts than those with low absolute values.
The research for this method has focused mainly on Cosine similarity,^{1} though the implementation also supports L1 and L2.

**An example**

The vector `[10, -2, 0, 99, 0.1, -8, 42, -13, 6, 0.1]`

with `k = 4`

is represented by indices `[4, 7, -8, 1]`

.
Indices are 1-indexed and indices for negative values are negated (hence the -8).
Indices can optionally be repeated based on their ranking.
In this example, the indices would be repeated `[4, 4, 4, 4, 7, 7, 7, -8, -8, 1]`

.
Index 4 has the highest absolute value, so it’s repeated `k - 0 = 4`

times.
Index 7 has the second highest absolute value, so it’s repeated `k - 1 = 3`

times, and so on.
The search algorithm computes the score as the size of the intersection of the stored vector’s representation and the
query vector’s representation.
So for a query vector represented by `[2, 2, 2, 2, 7, 7, 7, 4, 4, 5]`

, the intersection is `[7, 7, 7, 4, 4]`

, producing
a score of 5.
In some experiments, repetition has actually decreased recall, so it’s advised that you try with and without repetition.

```
PUT /my-index/_mapping
{
"properties": {
"my_vec": {
"type": "elastiknn_dense_float_vector", # 1
"elastiknn": {
"dims": 100, # 2
"model": "permutation_lsh", # 3
"k": 10, # 4
"repeating": true # 5
}
}
}
}
```

# | Description |
---|---|

1 | Vector datatype. Must be dense float vector. |

2 | Vector dimensionality. |

3 | Model type. |

4 | The number of top indices to pick. |

5 | Whether to repeat the indices proportionally to their rank. See the notes on repeating above. |

## Nearest Neighbor Queries

Elastiknn provides a query called `elastiknn_nearest_neighbors`

, which can be used in a `GET /_search`

request just like
standard Elasticsearch queries, as well as in combination with standard Elasticsearch queries.

### General Structure

The general query structure looks like this:

```
GET /my-index/_search
{
"query": {
"elastiknn_nearest_neighbors": { # 1
"field": "my_vec", # 2
"vec": { # 3
"values": [0.1, 0.2, 0.3, ...],
},
"model": "exact", # 4
"similarity": "cosine", # 5
... # 6
}
}
}
```

# | Description |
---|---|

1 | Query type provided by Elastiknn. |

2 | Name of the field containing your vectors. |

3 | Query vector. In this example it’s a literal vector, but can also be a reference to an indexed vector. |

4 | Model type. Exact search used in this example. More models covered below. |

5 | One of the five similarity functions used to score indexed vectors. |

6 | Additional query parameters for different model/similarity combinations. Covered more below. |

### Compatibility of Vector Types and Similarities

Jaccard and Hamming similarity only work with sparse bool vectors.
Cosine,^{1} L1, and L2 similarity only work with dense float vectors.
The following documentation assume this restriction is known.

These restrictions aren’t inherent to the types and algorithms, i.e., you could in theory run cosine similarity on sparse vectors. The restriction merely reflects the most common patterns and simplifies the implementation.

### Similarity Scoring

Elasticsearch queries must return a non-negative floating-point score.
For Elastiknn, the score for an indexed vector represents its similarity to the query vector.
However, not all similarity functions increase as similarity increases.
For example, a perfect similarity for the L1 and L2 functions is 0.
Such functions really represent *distance* without a well-defined mapping from distance to similarity.
In these cases Elastiknn applies a transformation to invert the score such that more similar vectors have higher scores.
The exact transformations are described below.

Similarity | Transformation to Elasticsearch Score | Min Value | Max Value |
---|---|---|---|

Jaccard | N/A | 0 | 1.0 |

Hamming | N/A | 0 | 1.0 |

Cosine^{1} |
`cosine similarity + 1` |
0 | 2 |

L1 | `1 / (1 + l1 distance)` |
0 | 1 |

L2 | `1 / (1 + l2 distance)` |
0 | 1 |

If you’re using the `elastiknn_nearest_neighbors`

query with other queries, and the score values are inconvenient (e.g. huge values like 1e6), consider wrapping the query in a Script Score Query, where you can access and transform the `_score`

value.

### Query Vector

The query vector is either a literal vector or a pointer to an indexed vector.

```
GET /my-index/_search
{
"query": {
"elastiknn_nearest_neighbors": {
...
"vec": { # 1
"true_indices": [1, 3, 5, ...],
"total_indices": 1000
},
...
"vec": { # 2
"values": [0.1, 0.2, 0.3, ...]
},
...
"vec": { # 3
"index": "my-other-index",
"field": "my_vec",
"id": "abc123"
},
}
}
}
```

# | Description |
---|---|

1 | Literal sparse bool query vector. |

2 | Literal dense float query vector. |

3 | Indexed query vector. This assumes you have another index called `my-other-index` with a document with id `abc123` that contains a valid vector in field `my_vec` . |

### Exact Query

Computes the exact similarity of a query vector against all indexed vectors. The algorithm is not efficient compared to approximate search, but the implementation has been extensively profiled and optimized.

```
GET /my-index/_search
{
"query": {
"elastiknn_nearest_neighbors": {
"field": "my_vec",
"vec": { # 1
"values": [0.1, 0.2, 0.3, ...],
},
"model": "exact", # 2
"similarity": "(cosine | l1 | l2)", # 3
}
}
}
```

# | Description |
---|---|

1 | Query vector. Must match the datatype of `my_vec` or be a pointer to an indexed vector that matches the type. |

2 | Model name. |

3 | Similarity function. Must be compatible with the vector type. |

### LSH Search Strategy

All LSH search models follow roughly the same strategy. They first retrieve approximate neighbors based on common hash terms and then compute the exact similarity for a subset of the best approximate candidates. The exact steps are as follows:

- Hash the query vector using model parameters that were specified in the indexed vector’s mapping.
- Use the hash values to construct and execute a query that finds other vectors with the same hash values. The query is a modification of Lucene’s TermInSetQuery.
- Take the top vectors with the most matching hashes and compute their exact similarity to the query vector.
The
`candidates`

parameter controls the number of exact similarity computations. Specifically, we compute exact similarity for the topcandidate vectors in each segment. As a reminder, each Elasticsearch index has >= 1 shards, and each shard has >= 1 segments. That means if you set`candidates`

`"candiates": 200`

for an index with 2 shards, each with 3 segments, then you’ll compute the exact similarity for`2 * 3 * 200 = 1200`

vectors.`candidates`

must be set to a number greater or equal to the number of Elasticsearch results you want to get. Higher values generally mean higher recall and higher latency.

### Jaccard LSH Query

Retrieve sparse bool vectors based on approximate Jaccard similarity.

```
GET /my-index/_search
{
"query": {
"elastiknn_nearest_neighbors": {
"field": "my_vec", # 1
"vec": { # 2
"true_indices": [1, 3, 5, ...],
"total_indices": 100
},
"model": "lsh", # 3
"similarity": "jaccard", # 4
"candidates": 50 # 5
}
}
}
```

# | Description |
---|---|

1 | Indexed field. Must use `lsh` mapping model with `jaccard` similarity. |

2 | Query vector. Must be literal sparse bool or a pointer to an indexed sparse bool vector. |

3 | Model name. |

4 | Similarity function. |

5 | Number of candidates per segment. See the section on LSH Search Strategy. |

### Hamming LSH Query

Retrieve sparse bool vectors based on approximate Hamming similarity.

```
GET /my-index/_search
{
"query": {
"elastiknn_nearest_neighbors": {
"field": "my_vec", # 1
"vec": { # 2
"true_indices": [1, 3, 5, ...],
"total_indices": 100
},
"model": "lsh", # 3
"similarity": "hamming", # 4
"candidates": 50 # 5
}
}
}
```

# | Description |
---|---|

1 | Indexed field. Must use `lsh` mapping model with `hamming` similarity. |

2 | Query vector. Must be literal sparse bool or a pointer to an indexed sparse bool vector. |

3 | Model name. |

4 | Similarity function. |

5 | Number of candidates per segment. See the section on LSH Search Strategy. |

6 | Set to true to use the more-like-this heuristic to pick a subset of hashes. Generally faster but still experimental. |

### Cosine LSH Query

Retrieve dense float vectors based on approximate Cosine similarity.^{1}

```
GET /my-index/_search
{
"query": {
"elastiknn_nearest_neighbors": {
"field": "my_vec", # 1
"vec": { # 2
"values": [0.1, 0.2, 0.3, ...]
},
"model": "lsh", # 3
"similarity": "cosine", # 4
"candidates": 50 # 5
}
}
}
```

# | Description |
---|---|

1 | Indexed field. Must use `lsh` mapping model with `cosine` similarity. |

2 | Query vector. Must be literal dense float or a pointer to an indexed dense float vector. |

3 | Model name. |

4 | Similarity function. |

5 | Number of candidates per segment. See the section on LSH Search Strategy. |

6 | Set to true to use the more-like-this heuristic to pick a subset of hashes. Generally faster but still experimental. |

### L1 LSH Query

Not yet implemented.

### L2 LSH Query

Retrieve dense float vectors based on approximate L2 similarity.

```
GET /my-index/_search
{
"query": {
"elastiknn_nearest_neighbors": {
"field": "my_vec", # 1
"vec": { # 2
"values": [0.1, 0.2, 0.3, ...]
},
"model": "lsh", # 3
"similarity": "l2", # 4
"candidates": 50, # 5
"probes": 2 # 6
}
}
}
```

# | Description |
---|---|

1 | Indexed field. Must use `lsh` mapping model with `l2` similarity. |

2 | Query vector. Must be literal dense float or a pointer to an indexed dense float vector. |

3 | Model name. |

4 | Similarity function. |

5 | Number of candidates per segment. See the section on LSH Search Strategy. |

6 | Number of probes for using the multiprobe search technique. Default value is zero. Max value is `3^k` . Generally, increasing `probes` will increase recall, will allow you to use a smaller value for `L` with comparable recall, but introduces some additional computation at query time. See the L2 LSH mapping section for more on `L` and `k` . |

### Permutation LSH Query

Retrieve dense float vectors based on the permutation LSH algorithm. See the permutation LSH mapping for more about the algorithm.

```
GET /my-index/_search
{
"query": {
"elastiknn_nearest_neighbors": {
"field": "my_vec", # 1
"vec": { # 2
"values": [0.1, 0.2, 0.3, ...]
},
"model": "permutation_lsh", # 3
"similarity": "cosine", # 4
"candidates": 50 # 5
}
}
}
```

# | Description |
---|---|

1 | Indexed field. Must use `permutation_lsh` mapping to use this query. |

2 | Query vector. Must be literal dense float or a pointer to an indexed dense float vector. |

3 | Model name. |

4 | Similarity function. Supports Cosine,^{1} L1, and L2. |

5 | Number of candidates per segment. See the section on LSH Search Strategy. |

### Model and Query Compatibility

Some models can support more than one type of query.
For example, sparse bool vectors indexed with the Jaccard LSH model support exact searches using both Jaccard and Hamming similarity.
The opposite is *not* true: vectors stored using the exact model do not support Jaccard LSH queries.

The tables below shows valid model/query combinations.
Rows are models and columns are queries.
The similarity functions are abbreviated (J: Jaccard, H: Hamming, C: Cosine,^{1} L1, L2).

#### elastiknn_sparse_bool_vector

Model / Query | Exact | Sparse Indexed | Jaccard LSH | Hamming LSH |
---|---|---|---|---|

Exact (i.e. no model specified) | ✔ (J, H) | x | x | x |

Sparse Indexed | ✔ (J, H) | ✔ (J, H) | x | x |

Jaccard LSH | ✔ (J, H) | x | ✔ | x |

Hamming LSH | ✔ (J, H) | x | x | ✔ |

#### elastiknn_dense_float_vector

Model / Query | Exact | Cosine LSH | L2 LSH | Permutation LSH |
---|---|---|---|---|

Exact (i.e. no model specified) | ✔ (C, L1, L2) | x | x | x |

Cosine LSH | ✔ (C, L1, L2) | ✔ | x | x |

L2 LSH | ✔ (C, L1, L2) | x | ✔ | x |

Permutation LSH | ✔ (C, L1, L2) | x | x | ✔ |

## Common Patterns

### Running Nearest Neighbors Query on a Filtered Subset of Documents

It’s common to filter for a subset of documents based on some property and *then* run the `elastiknn_nearest_neighbors`

query on that subset.
For example, if your docs contain a `color`

keyword, you might want to find all of the docs with `"color": "blue"`

, and only run `elastiknn_nearest_neighbors`

on that subset.
To do this, you can use the `elastiknn_nearest_neighbors`

query in a function score query
or in a query rescorer.
The function score query is usually simpler, but both are covered below.

#### Using a Function Score Query

```
GET /my-index/_search
{
"size": 10,
"query": {
"function_score": {
"query": {
"term": { "color": "blue" } # 1
},
"functions": [ # 2
{
"elastiknn_nearest_neighbors": { # 3
"field": "vec",
"similarity": "cosine",
"model": "exact",
"vec": {
"values": [0.1, 0.2, 0.3, ...]
}
}
}
]
}
}
}
```

# | Description |
---|---|

1 | Term query will limit the functions to only run on documents matching `"color": "blue"` . |

2 | List of functions which are applied to the matching documents. |

3 | `elastiknn_nearest_neighbors` query that is evaluated on matching documents. The query produces the similarity score, which, by default, is multiplied by the term query score. If you’d like to change this behavior, see the `score_mode` , `boost_mode` , and `weight` parameters in the Elasticsearch function score docs. |

**Caveats**

This does not yet support passing indexed vectors.

When using `"model": "lsh"`

, the `"candidates"`

parameter is ignored and vectors are not re-scored with the exact
similarity like they are with a `elastiknn_nearest_neighbors`

query.
Instead, the score is: `max similarity score * proportion of matching hashes`

.
This is a necessary consequence of the fact that score functions take a doc ID and must immediately return a score.

#### Using a Query Rescorer

```
GET /my-index/_search
{
"query": {
"term": { "color": "blue" } # 1
},
"rescore": {
"window_size": 10, # 2
"query": {
"rescore_query": {
"elastiknn_nearest_neighbors": { # 3
"field" : "vec",
"similarity" : "l2",
"model" : "exact",
"vec" : {
"values" : [0.1, 0.2, 0.3, ...]
}
}
},
"query_weight": 0, # 4
"rescore_query_weight": 1 # 5
}
}
}
```

# | Description |
---|---|

1 | Term query will limit to only run on documents containing `"color": "blue"` . |

2 | The window size controls how many of the docs that matched the term query will be considered for the nearest neighbors query. |

3 | `elastiknn_nearest_neighbors` query that evaluates L2 similarity for the “vec” field in any document containing `"color": "blue"` . |

4 | Ignore the score from the term query. |

5 | Use the score from the rescore query. |

**Caveats**

This does not yet support passing indexed vectors.

Elasticsearch has a configurable limit for the number of docs that are matched and passed to the `rescore`

query.
The default is 10,000.
You can modify the `index.max_rescore_window`

setting to get around this.

Given the default limit of 10k vectors passed to the nearest neighbors query, you can typically use exact queries. As a point of reference, exact queries on the Fashion-MNIST dataset (60k 784-dimensional vectors) run in about 250ms.

If you determine you need an approximate query for re-scoring, you should ensure that `candidates = window_size > size`

.
Ideally `candidates`

is 10x-100x larger than `size`

.
Also, consider that it’s possible for the approximate query to match fewer than `candidates`

vectors.
So you can end up with fewer than `size`

results in your search response.
This can happen because the nearest neighbors query is only given access to the `window_size`

vectors which matched the original query.
If you run into this, you should probably adjust your approximate model parameters for higher recall.
There are notes on each model about how the parameters affect recall and precision.

### Using Stored Fields for Faster Queries

This is a fairly well-known Elasticsearch optimization that applies nicely to some elastiknn use cases.
If you only need to retrieve a small subset of the document source (e.g. only the ID), you can store the relavant fields as `stored`

fields to get a meaningful speedup.
The Elastiknn scala client uses this optimization to store and retrieve document IDs, yielding a ~40% speedup for queries.
The setting is documented here
and discussed in detail in this Github issue.

## Nice to Know

Here are some other things worth knowing. Perhaps there will be a more cohesive way to present these in the future.

### Storing Model Parameters

The LSH models all use randomized parameters to hash vectors.
The simplest example is the bit-sampling model for Hamming similarity, parameterized by a list of randomly sampled indices.
A more complicated example is the stable distributions model for L2 similarity, parameterized by a set of random unit vectors
and a set of random bias scalars.
These parameters aren’t actually stored anywhere in Elasticsearch.
Rather, they are lazily re-computed from a fixed random seed (0) each time they are needed.
The advantage of this is that it avoids storing and synchronizing potentially large parameter blobs in the cluster.
The disadvantage is that it’s expensive to re-compute the randomized parameters.
So instead we keep a cache of models in each Elasticsearch node, keyed on the model hyperparameters (e.g. `L`

, `k`

, etc.).
The hyperparameters are stored inside the mappings where they are originally defined.

### Transforming and Indexing Vectors

Each vector is transformed (e.g. hashed) based on its mapping when the user makes an indexing request. All vectors store a binary doc values field containing a serialized version of the vector for exact queries, and vectors indexed using an LSH model index the hashes using a Lucene Term field. For example, for a sparse bool vector with a Jaccard LSH mapping, Elastiknn indexes the exact vector as a byte array in a doc values field and the vector’s hash values as a set of Lucene Terms.

### Parallelism

From Elasticsearch’s perspective, the `elastiknn_nearest_neighbors`

query is no different from any other query.
Elasticsearch receives a JSON query containing an `elastiknn_nearest_neighbors`

key, passes the JSON to a parser implemented by Elastiknn, the parser produces a Lucene query, and Elasticsearch executes that query on each shard in the index.

This means the simplest way to increase query parallelism is to add shards to your index. Obviously this has an upper limit, but the general performance implications of sharding are beyond the scope of this document.

The number of shards is a static setting provided when creating the index:

```
PUT /my-index
{
"settings": {
"index": {
"number_of_shards": 2
}
}
}
```

See the create index documentation for more details.