// @generated by protoc-gen-es v2.5.0 with parameter "target=js+dts,import_extension=js,json_types=true"
// @generated from file arg_services/nlp/v1/nlp.proto (package arg_services.nlp.v1, syntax proto3)
/* eslint-disable */

// Service for offloading computationally complex NLP tasks.

import type { GenEnum, GenFile, GenMessage, GenService } from "@bufbuild/protobuf/codegenv2";
import type { JsonObject, Message } from "@bufbuild/protobuf";
import type { StructJson } from "@bufbuild/protobuf/wkt";

/**
 * Describes the file arg_services/nlp/v1/nlp.proto.
 */
export declare const file_arg_services_nlp_v1_nlp: GenFile;

/**
 * Common message for configuring spacy.
 *
 * @generated from message arg_services.nlp.v1.NlpConfig
 */
export declare type NlpConfig = Message<"arg_services.nlp.v1.NlpConfig"> & {
  /**
   * Any language supported by spacy (e.g., `en`).
   * [Reference](https://spacy.io/usage/models#languages).
   *
   * @generated from field: string language = 1;
   */
  language: string;

  /**
   * Name of the trained spacy pipeline (e.g., `en_core_web_lg`).
   * If empty, a blank spacy model will be used (e.g., if you only need embeddings and provide custom `embedding_models`.
   * [Example: English models](https://spacy.io/models/en).
   *
   * @generated from field: string spacy_model = 2;
   */
  spacyModel: string;

  /**
   * List of embeddings to use for computing word/sentence vectors.
   * If given, these embeddings will **override** the embeddings of the specified `spacy_model`.
   * Multiple models are concatenated to each other, increasing the length of the resulting vector.
   *
   * @generated from field: repeated arg_services.nlp.v1.EmbeddingModel embedding_models = 3;
   */
  embeddingModels: EmbeddingModel[];

  /**
   * Mathematical function to determine a similarity score given two strings.
   *
   * @generated from field: arg_services.nlp.v1.SimilarityMethod similarity_method = 4;
   */
  similarityMethod: SimilarityMethod;
};

/**
 * Common message for configuring spacy.
 *
 * @generated from message arg_services.nlp.v1.NlpConfig
 */
export declare type NlpConfigJson = {
  /**
   * Any language supported by spacy (e.g., `en`).
   * [Reference](https://spacy.io/usage/models#languages).
   *
   * @generated from field: string language = 1;
   */
  language?: string;

  /**
   * Name of the trained spacy pipeline (e.g., `en_core_web_lg`).
   * If empty, a blank spacy model will be used (e.g., if you only need embeddings and provide custom `embedding_models`.
   * [Example: English models](https://spacy.io/models/en).
   *
   * @generated from field: string spacy_model = 2;
   */
  spacyModel?: string;

  /**
   * List of embeddings to use for computing word/sentence vectors.
   * If given, these embeddings will **override** the embeddings of the specified `spacy_model`.
   * Multiple models are concatenated to each other, increasing the length of the resulting vector.
   *
   * @generated from field: repeated arg_services.nlp.v1.EmbeddingModel embedding_models = 3;
   */
  embeddingModels?: EmbeddingModelJson[];

  /**
   * Mathematical function to determine a similarity score given two strings.
   *
   * @generated from field: arg_services.nlp.v1.SimilarityMethod similarity_method = 4;
   */
  similarityMethod?: SimilarityMethodJson;
};

/**
 * Describes the message arg_services.nlp.v1.NlpConfig.
 * Use `create(NlpConfigSchema)` to create a new message.
 */
export declare const NlpConfigSchema: GenMessage<NlpConfig, {jsonType: NlpConfigJson}>;

/**
 * @generated from message arg_services.nlp.v1.SimilaritiesRequest
 */
export declare type SimilaritiesRequest = Message<"arg_services.nlp.v1.SimilaritiesRequest"> & {
  /**
   * Spacy config.
   *
   * @generated from field: arg_services.nlp.v1.NlpConfig config = 1;
   */
  config?: NlpConfig;

  /**
   * List of string pairs to compare.
   *
   * @generated from field: repeated arg_services.nlp.v1.TextTuple text_tuples = 2;
   */
  textTuples: TextTuple[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: JsonObject;
};

/**
 * @generated from message arg_services.nlp.v1.SimilaritiesRequest
 */
export declare type SimilaritiesRequestJson = {
  /**
   * Spacy config.
   *
   * @generated from field: arg_services.nlp.v1.NlpConfig config = 1;
   */
  config?: NlpConfigJson;

  /**
   * List of string pairs to compare.
   *
   * @generated from field: repeated arg_services.nlp.v1.TextTuple text_tuples = 2;
   */
  textTuples?: TextTupleJson[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: StructJson;
};

/**
 * Describes the message arg_services.nlp.v1.SimilaritiesRequest.
 * Use `create(SimilaritiesRequestSchema)` to create a new message.
 */
export declare const SimilaritiesRequestSchema: GenMessage<SimilaritiesRequest, {jsonType: SimilaritiesRequestJson}>;

/**
 * @generated from message arg_services.nlp.v1.SimilaritiesResponse
 */
export declare type SimilaritiesResponse = Message<"arg_services.nlp.v1.SimilaritiesResponse"> & {
  /**
   * List of similarities ordered just like the original `text_tuples`.
   *
   * @generated from field: repeated double similarities = 1;
   */
  similarities: number[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: JsonObject;
};

/**
 * @generated from message arg_services.nlp.v1.SimilaritiesResponse
 */
export declare type SimilaritiesResponseJson = {
  /**
   * List of similarities ordered just like the original `text_tuples`.
   *
   * @generated from field: repeated double similarities = 1;
   */
  similarities?: (number | "NaN" | "Infinity" | "-Infinity")[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: StructJson;
};

/**
 * Describes the message arg_services.nlp.v1.SimilaritiesResponse.
 * Use `create(SimilaritiesResponseSchema)` to create a new message.
 */
export declare const SimilaritiesResponseSchema: GenMessage<SimilaritiesResponse, {jsonType: SimilaritiesResponseJson}>;

/**
 * Store a pair of strings.
 *
 * @generated from message arg_services.nlp.v1.TextTuple
 */
export declare type TextTuple = Message<"arg_services.nlp.v1.TextTuple"> & {
  /**
   * @generated from field: string text1 = 1;
   */
  text1: string;

  /**
   * @generated from field: string text2 = 2;
   */
  text2: string;
};

/**
 * Store a pair of strings.
 *
 * @generated from message arg_services.nlp.v1.TextTuple
 */
export declare type TextTupleJson = {
  /**
   * @generated from field: string text1 = 1;
   */
  text1?: string;

  /**
   * @generated from field: string text2 = 2;
   */
  text2?: string;
};

/**
 * Describes the message arg_services.nlp.v1.TextTuple.
 * Use `create(TextTupleSchema)` to create a new message.
 */
export declare const TextTupleSchema: GenMessage<TextTuple, {jsonType: TextTupleJson}>;

/**
 * Wrapper message to encode a list of strings that can also be `null`.
 *
 * @generated from message arg_services.nlp.v1.Strings
 */
export declare type Strings = Message<"arg_services.nlp.v1.Strings"> & {
  /**
   * @generated from field: repeated string values = 1;
   */
  values: string[];
};

/**
 * Wrapper message to encode a list of strings that can also be `null`.
 *
 * @generated from message arg_services.nlp.v1.Strings
 */
export declare type StringsJson = {
  /**
   * @generated from field: repeated string values = 1;
   */
  values?: string[];
};

/**
 * Describes the message arg_services.nlp.v1.Strings.
 * Use `create(StringsSchema)` to create a new message.
 */
export declare const StringsSchema: GenMessage<Strings, {jsonType: StringsJson}>;

/**
 * @generated from message arg_services.nlp.v1.DocBinRequest
 */
export declare type DocBinRequest = Message<"arg_services.nlp.v1.DocBinRequest"> & {
  /**
   * Spacy config.
   *
   * @generated from field: arg_services.nlp.v1.NlpConfig config = 1;
   */
  config?: NlpConfig;

  /**
   * List of strings to be processed.
   *
   * @generated from field: repeated string texts = 2;
   */
  texts: string[];

  /**
   * Attributes that shall be included in the DocBin object.
   * Defaults to `("ORTH", "TAG", "HEAD", "DEP", "ENT_IOB", "ENT_TYPE", "ENT_KB_ID", "LEMMA", "MORPH", "POS")`.
   * Possible values: `("IS_ALPHA", "IS_ASCII", "IS_DIGIT", "IS_LOWER", "IS_PUNCT", "IS_SPACE", "IS_TITLE", "IS_UPPER", "LIKE_URL", "LIKE_NUM", "LIKE_EMAIL", "IS_STOP", "IS_OOV_DEPRECATED", "IS_BRACKET", "IS_QUOTE", "IS_LEFT_PUNCT", "IS_RIGHT_PUNCT", "IS_CURRENCY", "ID", "ORTH", "LOWER", "NORM", "SHAPE", "PREFIX", "SUFFIX", "LENGTH", "CLUSTER", "LEMMA", "POS", "TAG", "DEP", "ENT_IOB", "ENT_TYPE", "ENT_ID", "ENT_KB_ID", "HEAD", "SENT_START", "SENT_END", "SPACY", "PROB", "LANG", "MORPH", "IDX")`.
   * [Documentation](https://spacy.io/api/token#attributes).
   *
   * @generated from field: optional arg_services.nlp.v1.Strings attributes = 3;
   */
  attributes?: Strings;

  /**
   * List of pipeline components that shall be enabled/disabled when processing documents.
   * If only certain attributed (e.g. POS tags) are relevant, one can enhance the performance by selecting components.
   * **Important**: There may be custom spacy components that are required for some of the functionality of the NLP service.
   * In the reference implementation, this applies to the components `embedding_models` and `similarity_method`.
   *
   * @generated from oneof arg_services.nlp.v1.DocBinRequest.pipes
   */
  pipes: {
    /**
     * @generated from field: arg_services.nlp.v1.Strings enabled_pipes = 4;
     */
    value: Strings;
    case: "enabledPipes";
  } | {
    /**
     * @generated from field: arg_services.nlp.v1.Strings disabled_pipes = 5;
     */
    value: Strings;
    case: "disabledPipes";
  } | { case: undefined; value?: undefined };

  /**
   * List of vectors that shall be saved in the `DocBin` object.
   * The computation is time-consuming, so you should only specify the embeddings you actually use!
   *
   * @generated from field: repeated arg_services.nlp.v1.EmbeddingLevel embedding_levels = 6;
   */
  embeddingLevels: EmbeddingLevel[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: JsonObject;
};

/**
 * @generated from message arg_services.nlp.v1.DocBinRequest
 */
export declare type DocBinRequestJson = {
  /**
   * Spacy config.
   *
   * @generated from field: arg_services.nlp.v1.NlpConfig config = 1;
   */
  config?: NlpConfigJson;

  /**
   * List of strings to be processed.
   *
   * @generated from field: repeated string texts = 2;
   */
  texts?: string[];

  /**
   * Attributes that shall be included in the DocBin object.
   * Defaults to `("ORTH", "TAG", "HEAD", "DEP", "ENT_IOB", "ENT_TYPE", "ENT_KB_ID", "LEMMA", "MORPH", "POS")`.
   * Possible values: `("IS_ALPHA", "IS_ASCII", "IS_DIGIT", "IS_LOWER", "IS_PUNCT", "IS_SPACE", "IS_TITLE", "IS_UPPER", "LIKE_URL", "LIKE_NUM", "LIKE_EMAIL", "IS_STOP", "IS_OOV_DEPRECATED", "IS_BRACKET", "IS_QUOTE", "IS_LEFT_PUNCT", "IS_RIGHT_PUNCT", "IS_CURRENCY", "ID", "ORTH", "LOWER", "NORM", "SHAPE", "PREFIX", "SUFFIX", "LENGTH", "CLUSTER", "LEMMA", "POS", "TAG", "DEP", "ENT_IOB", "ENT_TYPE", "ENT_ID", "ENT_KB_ID", "HEAD", "SENT_START", "SENT_END", "SPACY", "PROB", "LANG", "MORPH", "IDX")`.
   * [Documentation](https://spacy.io/api/token#attributes).
   *
   * @generated from field: optional arg_services.nlp.v1.Strings attributes = 3;
   */
  attributes?: StringsJson;

  /**
   * @generated from field: arg_services.nlp.v1.Strings enabled_pipes = 4;
   */
  enabledPipes?: StringsJson;

  /**
   * @generated from field: arg_services.nlp.v1.Strings disabled_pipes = 5;
   */
  disabledPipes?: StringsJson;

  /**
   * List of vectors that shall be saved in the `DocBin` object.
   * The computation is time-consuming, so you should only specify the embeddings you actually use!
   *
   * @generated from field: repeated arg_services.nlp.v1.EmbeddingLevel embedding_levels = 6;
   */
  embeddingLevels?: EmbeddingLevelJson[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: StructJson;
};

/**
 * Describes the message arg_services.nlp.v1.DocBinRequest.
 * Use `create(DocBinRequestSchema)` to create a new message.
 */
export declare const DocBinRequestSchema: GenMessage<DocBinRequest, {jsonType: DocBinRequestJson}>;

/**
 * @generated from message arg_services.nlp.v1.DocBinResponse
 */
export declare type DocBinResponse = Message<"arg_services.nlp.v1.DocBinResponse"> & {
  /**
   * Serialized [`DocBin`](https://spacy.io/api/docbin) object
   *
   * @generated from field: bytes docbin = 1;
   */
  docbin: Uint8Array;

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: JsonObject;
};

/**
 * @generated from message arg_services.nlp.v1.DocBinResponse
 */
export declare type DocBinResponseJson = {
  /**
   * Serialized [`DocBin`](https://spacy.io/api/docbin) object
   *
   * @generated from field: bytes docbin = 1;
   */
  docbin?: string;

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: StructJson;
};

/**
 * Describes the message arg_services.nlp.v1.DocBinResponse.
 * Use `create(DocBinResponseSchema)` to create a new message.
 */
export declare const DocBinResponseSchema: GenMessage<DocBinResponse, {jsonType: DocBinResponseJson}>;

/**
 * @generated from message arg_services.nlp.v1.VectorsRequest
 */
export declare type VectorsRequest = Message<"arg_services.nlp.v1.VectorsRequest"> & {
  /**
   * Spacy config.
   *
   * @generated from field: arg_services.nlp.v1.NlpConfig config = 1;
   */
  config?: NlpConfig;

  /**
   * List of strings that shall be embedded (i.e., converted to vectors).
   *
   * @generated from field: repeated string texts = 2;
   */
  texts: string[];

  /**
   * List of vectors that shall be returned.
   * The computation is time-consuming, so you should only specify the embeddings you actually use!
   *
   * @generated from field: repeated arg_services.nlp.v1.EmbeddingLevel embedding_levels = 3;
   */
  embeddingLevels: EmbeddingLevel[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: JsonObject;
};

/**
 * @generated from message arg_services.nlp.v1.VectorsRequest
 */
export declare type VectorsRequestJson = {
  /**
   * Spacy config.
   *
   * @generated from field: arg_services.nlp.v1.NlpConfig config = 1;
   */
  config?: NlpConfigJson;

  /**
   * List of strings that shall be embedded (i.e., converted to vectors).
   *
   * @generated from field: repeated string texts = 2;
   */
  texts?: string[];

  /**
   * List of vectors that shall be returned.
   * The computation is time-consuming, so you should only specify the embeddings you actually use!
   *
   * @generated from field: repeated arg_services.nlp.v1.EmbeddingLevel embedding_levels = 3;
   */
  embeddingLevels?: EmbeddingLevelJson[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: StructJson;
};

/**
 * Describes the message arg_services.nlp.v1.VectorsRequest.
 * Use `create(VectorsRequestSchema)` to create a new message.
 */
export declare const VectorsRequestSchema: GenMessage<VectorsRequest, {jsonType: VectorsRequestJson}>;

/**
 * @generated from message arg_services.nlp.v1.VectorsResponse
 */
export declare type VectorsResponse = Message<"arg_services.nlp.v1.VectorsResponse"> & {
  /**
   * List of vectors whose order corresponds to the one of `texts`.
   *
   * @generated from field: repeated arg_services.nlp.v1.VectorResponse vectors = 1;
   */
  vectors: VectorResponse[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: JsonObject;
};

/**
 * @generated from message arg_services.nlp.v1.VectorsResponse
 */
export declare type VectorsResponseJson = {
  /**
   * List of vectors whose order corresponds to the one of `texts`.
   *
   * @generated from field: repeated arg_services.nlp.v1.VectorResponse vectors = 1;
   */
  vectors?: VectorResponseJson[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: StructJson;
};

/**
 * Describes the message arg_services.nlp.v1.VectorsResponse.
 * Use `create(VectorsResponseSchema)` to create a new message.
 */
export declare const VectorsResponseSchema: GenMessage<VectorsResponse, {jsonType: VectorsResponseJson}>;

/**
 * Container object that includes vectors for all levels specified in `embedding_levels`.
 *
 * @generated from message arg_services.nlp.v1.VectorResponse
 */
export declare type VectorResponse = Message<"arg_services.nlp.v1.VectorResponse"> & {
  /**
   * One vector for the whole string.
   *
   * @generated from field: arg_services.nlp.v1.Vector document = 1;
   */
  document?: Vector;

  /**
   * Vectors for all tokens in the string.
   *
   * @generated from field: repeated arg_services.nlp.v1.Vector tokens = 2;
   */
  tokens: Vector[];

  /**
   * Vectors for all sentences found in the string.
   *
   * @generated from field: repeated arg_services.nlp.v1.Vector sentences = 3;
   */
  sentences: Vector[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: JsonObject;
};

/**
 * Container object that includes vectors for all levels specified in `embedding_levels`.
 *
 * @generated from message arg_services.nlp.v1.VectorResponse
 */
export declare type VectorResponseJson = {
  /**
   * One vector for the whole string.
   *
   * @generated from field: arg_services.nlp.v1.Vector document = 1;
   */
  document?: VectorJson;

  /**
   * Vectors for all tokens in the string.
   *
   * @generated from field: repeated arg_services.nlp.v1.Vector tokens = 2;
   */
  tokens?: VectorJson[];

  /**
   * Vectors for all sentences found in the string.
   *
   * @generated from field: repeated arg_services.nlp.v1.Vector sentences = 3;
   */
  sentences?: VectorJson[];

  /**
   * Implementation-specific information can be encoded here
   *
   * @generated from field: google.protobuf.Struct extras = 15;
   */
  extras?: StructJson;
};

/**
 * Describes the message arg_services.nlp.v1.VectorResponse.
 * Use `create(VectorResponseSchema)` to create a new message.
 */
export declare const VectorResponseSchema: GenMessage<VectorResponse, {jsonType: VectorResponseJson}>;

/**
 * Container for storing a vector as a list of floats.
 *
 * @generated from message arg_services.nlp.v1.Vector
 */
export declare type Vector = Message<"arg_services.nlp.v1.Vector"> & {
  /**
   * @generated from field: repeated double vector = 1;
   */
  vector: number[];
};

/**
 * Container for storing a vector as a list of floats.
 *
 * @generated from message arg_services.nlp.v1.Vector
 */
export declare type VectorJson = {
  /**
   * @generated from field: repeated double vector = 1;
   */
  vector?: (number | "NaN" | "Infinity" | "-Infinity")[];
};

/**
 * Describes the message arg_services.nlp.v1.Vector.
 * Use `create(VectorSchema)` to create a new message.
 */
export declare const VectorSchema: GenMessage<Vector, {jsonType: VectorJson}>;

/**
 * Specification of one model that is used to generate embeddings for strings.
 *
 * @generated from message arg_services.nlp.v1.EmbeddingModel
 */
export declare type EmbeddingModel = Message<"arg_services.nlp.v1.EmbeddingModel"> & {
  /**
   * Each embedding has to be implemented, thus this enum is used to select the correct one.
   *
   * @generated from field: arg_services.nlp.v1.EmbeddingType model_type = 1;
   */
  modelType: EmbeddingType;

  /**
   * You have to specify the name of the model that should be used by the selected impelemtation (i.e., `model_type`).
   * We provide links to exemplary models for each implementation in the documentation of `EmbeddingType`.
   *
   * @generated from field: string model_name = 2;
   */
  modelName: string;

  /**
   * In case the selected model is not capable of directly creating sentence embeddings, you have to select a pooling strategy.
   * You can either use a standard function (`pooling_type`) or compute the power mean (`pmean`).
   *
   * @generated from oneof arg_services.nlp.v1.EmbeddingModel.pooling
   */
  pooling: {
    /**
     * Standard pooling functions like mean, min, max.
     *
     * @generated from field: arg_services.nlp.v1.Pooling pooling_type = 3;
     */
    value: Pooling;
    case: "poolingType";
  } | {
    /**
     * Power mean (or generalized mean).
     * This method allows you to alter the computation of the mean representation.
     * Special cases include arithmetic mean (p = 1), geometric mean (p = 0), harmonic mean (p = -1), minimum (p = -∞), maximum (p = ∞).
     * [Wikipedia](https://en.wikipedia.org/wiki/Generalized_mean).
     * [Paper](https://arxiv.org/abs/1803.01400).
     *
     * @generated from field: double pmean = 4;
     */
    value: number;
    case: "pmean";
  } | { case: undefined; value?: undefined };
};

/**
 * Specification of one model that is used to generate embeddings for strings.
 *
 * @generated from message arg_services.nlp.v1.EmbeddingModel
 */
export declare type EmbeddingModelJson = {
  /**
   * Each embedding has to be implemented, thus this enum is used to select the correct one.
   *
   * @generated from field: arg_services.nlp.v1.EmbeddingType model_type = 1;
   */
  modelType?: EmbeddingTypeJson;

  /**
   * You have to specify the name of the model that should be used by the selected impelemtation (i.e., `model_type`).
   * We provide links to exemplary models for each implementation in the documentation of `EmbeddingType`.
   *
   * @generated from field: string model_name = 2;
   */
  modelName?: string;

  /**
   * Standard pooling functions like mean, min, max.
   *
   * @generated from field: arg_services.nlp.v1.Pooling pooling_type = 3;
   */
  poolingType?: PoolingJson;

  /**
   * Power mean (or generalized mean).
   * This method allows you to alter the computation of the mean representation.
   * Special cases include arithmetic mean (p = 1), geometric mean (p = 0), harmonic mean (p = -1), minimum (p = -∞), maximum (p = ∞).
   * [Wikipedia](https://en.wikipedia.org/wiki/Generalized_mean).
   * [Paper](https://arxiv.org/abs/1803.01400).
   *
   * @generated from field: double pmean = 4;
   */
  pmean?: number | "NaN" | "Infinity" | "-Infinity";
};

/**
 * Describes the message arg_services.nlp.v1.EmbeddingModel.
 * Use `create(EmbeddingModelSchema)` to create a new message.
 */
export declare const EmbeddingModelSchema: GenMessage<EmbeddingModel, {jsonType: EmbeddingModelJson}>;

/**
 * Possible methods to compute the similarity between two vectors.
 *
 * @generated from enum arg_services.nlp.v1.SimilarityMethod
 */
export enum SimilarityMethod {
  /**
   * If not given, the implementation defaults to cosine similarity.
   *
   * @generated from enum value: SIMILARITY_METHOD_UNSPECIFIED = 0;
   */
  UNSPECIFIED = 0,

  /**
   * Cosine similarity. [Wikipedia](https://en.wikipedia.org/wiki/Cosine_similarity).
   *
   * @generated from enum value: SIMILARITY_METHOD_COSINE = 1;
   */
  COSINE = 1,

  /**
   * DynaMax Jaccard. [Paper](https://arxiv.org/abs/1904.13264), [Code](https://github.com/babylonhealth/fuzzymax/blob/master/similarity/fuzzy.py).
   *
   * @generated from enum value: SIMILARITY_METHOD_DYNAMAX_JACCARD = 2;
   */
  DYNAMAX_JACCARD = 2,

  /**
   * MaxPool Jaccard. [Paper](https://arxiv.org/abs/1904.13264), [Code](https://github.com/babylonhealth/fuzzymax/blob/master/similarity/fuzzy.py).
   *
   * @generated from enum value: SIMILARITY_METHOD_MAXPOOL_JACCARD = 3;
   */
  MAXPOOL_JACCARD = 3,

  /**
   * DynaMax Dice. [Paper](https://arxiv.org/abs/1904.13264), [Code](https://github.com/babylonhealth/fuzzymax/blob/master/similarity/fuzzy.py).
   *
   * @generated from enum value: SIMILARITY_METHOD_DYNAMAX_DICE = 4;
   */
  DYNAMAX_DICE = 4,

  /**
   * DynaMax Otsuka. [Paper](https://arxiv.org/abs/1904.13264), [Code](https://github.com/babylonhealth/fuzzymax/blob/master/similarity/fuzzy.py).
   *
   * @generated from enum value: SIMILARITY_METHOD_DYNAMAX_OTSUKA = 5;
   */
  DYNAMAX_OTSUKA = 5,

  /**
   * Word Mover's Distance [Gensim Tutorial](https://radimrehurek.com/gensim/auto_examples/tutorials/run_wmd.html).
   *
   * @generated from enum value: SIMILARITY_METHOD_WMD = 6;
   */
  WMD = 6,

  /**
   * Levenshtein distance. [Wikipedia](https://en.wikipedia.org/wiki/Levenshtein_distance).
   *
   * @generated from enum value: SIMILARITY_METHOD_EDIT = 7;
   */
  EDIT = 7,

  /**
   * Jaccard similarity. [Wikipedia](https://en.wikipedia.org/wiki/Jaccard_index).
   *
   * @generated from enum value: SIMILARITY_METHOD_JACCARD = 8;
   */
  JACCARD = 8,

  /**
   * Angular distance. [Wikipedia](https://en.wikipedia.org/wiki/Angular_distance).
   *
   * @generated from enum value: SIMILARITY_METHOD_ANGULAR = 9;
   */
  ANGULAR = 9,

  /**
   * Manhattan distance. [Wikipedia](https://en.wikipedia.org/wiki/Taxicab_geometry).
   *
   * @generated from enum value: SIMILARITY_METHOD_MANHATTAN = 10;
   */
  MANHATTAN = 10,

  /**
   * Euclidean distance. [Wikipedia](https://en.wikipedia.org/wiki/Euclidean_distance).
   *
   * @generated from enum value: SIMILARITY_METHOD_EUCLIDEAN = 11;
   */
  EUCLIDEAN = 11,

  /**
   * Dot product. [Wikipedia](https://en.wikipedia.org/wiki/Dot_product).
   *
   * @generated from enum value: SIMILARITY_METHOD_DOT = 12;
   */
  DOT = 12,
}

/**
 * Possible methods to compute the similarity between two vectors.
 *
 * @generated from enum arg_services.nlp.v1.SimilarityMethod
 */
export declare type SimilarityMethodJson = "SIMILARITY_METHOD_UNSPECIFIED" | "SIMILARITY_METHOD_COSINE" | "SIMILARITY_METHOD_DYNAMAX_JACCARD" | "SIMILARITY_METHOD_MAXPOOL_JACCARD" | "SIMILARITY_METHOD_DYNAMAX_DICE" | "SIMILARITY_METHOD_DYNAMAX_OTSUKA" | "SIMILARITY_METHOD_WMD" | "SIMILARITY_METHOD_EDIT" | "SIMILARITY_METHOD_JACCARD" | "SIMILARITY_METHOD_ANGULAR" | "SIMILARITY_METHOD_MANHATTAN" | "SIMILARITY_METHOD_EUCLIDEAN" | "SIMILARITY_METHOD_DOT";

/**
 * Describes the enum arg_services.nlp.v1.SimilarityMethod.
 */
export declare const SimilarityMethodSchema: GenEnum<SimilarityMethod, SimilarityMethodJson>;

/**
 * @generated from enum arg_services.nlp.v1.EmbeddingLevel
 */
export enum EmbeddingLevel {
  /**
   * In the default case, no vector is computed.
   *
   * @generated from enum value: EMBEDDING_LEVEL_UNSPECIFIED = 0;
   */
  UNSPECIFIED = 0,

  /**
   * Compute one vector for the whole string.
   *
   * @generated from enum value: EMBEDDING_LEVEL_DOCUMENT = 1;
   */
  DOCUMENT = 1,

  /**
   * Compute vectors for all tokens in the string.
   *
   * @generated from enum value: EMBEDDING_LEVEL_TOKENS = 2;
   */
  TOKENS = 2,

  /**
   * Compute vectors for all sentences found in the string.
   *
   * @generated from enum value: EMBEDDING_LEVEL_SENTENCES = 3;
   */
  SENTENCES = 3,
}

/**
 * @generated from enum arg_services.nlp.v1.EmbeddingLevel
 */
export declare type EmbeddingLevelJson = "EMBEDDING_LEVEL_UNSPECIFIED" | "EMBEDDING_LEVEL_DOCUMENT" | "EMBEDDING_LEVEL_TOKENS" | "EMBEDDING_LEVEL_SENTENCES";

/**
 * Describes the enum arg_services.nlp.v1.EmbeddingLevel.
 */
export declare const EmbeddingLevelSchema: GenEnum<EmbeddingLevel, EmbeddingLevelJson>;

/**
 * @generated from enum arg_services.nlp.v1.Pooling
 */
export enum Pooling {
  /**
   * IN the default case, the arithmetic mean should be used.
   *
   * @generated from enum value: POOLING_UNSPECIFIED = 0;
   */
  UNSPECIFIED = 0,

  /**
   * Arithmetic mean of all elements. [Wikipedia](https://en.wikipedia.org/wiki/Arithmetic_mean).
   *
   * @generated from enum value: POOLING_MEAN = 1;
   */
  MEAN = 1,

  /**
   * Maximum element of vector. [Wikipedia](https://en.wikipedia.org/wiki/Maximum).
   *
   * @generated from enum value: POOLING_MAX = 2;
   */
  MAX = 2,

  /**
   * Minimum element of vector. [Wikipedia](https://en.wikipedia.org/wiki/Minimum).
   *
   * @generated from enum value: POOLING_MIN = 3;
   */
  MIN = 3,

  /**
   * Sum of all elements.
   *
   * @generated from enum value: POOLING_SUM = 4;
   */
  SUM = 4,

  /**
   * First element of vector.
   *
   * @generated from enum value: POOLING_FIRST = 5;
   */
  FIRST = 5,

  /**
   * Last element of vector.
   *
   * @generated from enum value: POOLING_LAST = 6;
   */
  LAST = 6,

  /**
   * Median element of vector. [Wikipedia](https://en.wikipedia.org/wiki/Median).
   *
   * @generated from enum value: POOLING_MEDIAN = 7;
   */
  MEDIAN = 7,

  /**
   * Geometirc mean of all elements. [Wikipedia](https://en.wikipedia.org/wiki/Geometric_mean).
   *
   * @generated from enum value: POOLING_GMEAN = 8;
   */
  GMEAN = 8,

  /**
   * Harmonic mean of all elements. [Wikipedia](https://en.wikipedia.org/wiki/Harmonic_mean).
   *
   * @generated from enum value: POOLING_HMEAN = 9;
   */
  HMEAN = 9,
}

/**
 * @generated from enum arg_services.nlp.v1.Pooling
 */
export declare type PoolingJson = "POOLING_UNSPECIFIED" | "POOLING_MEAN" | "POOLING_MAX" | "POOLING_MIN" | "POOLING_SUM" | "POOLING_FIRST" | "POOLING_LAST" | "POOLING_MEDIAN" | "POOLING_GMEAN" | "POOLING_HMEAN";

/**
 * Describes the enum arg_services.nlp.v1.Pooling.
 */
export declare const PoolingSchema: GenEnum<Pooling, PoolingJson>;

/**
 * @generated from enum arg_services.nlp.v1.EmbeddingType
 */
export enum EmbeddingType {
  /**
   * In the default case, no embedding is computed.
   *
   * @generated from enum value: EMBEDDING_TYPE_UNSPECIFIED = 0;
   */
  UNSPECIFIED = 0,

  /**
   * [Spacy](https://spacy.io/models).
   *
   * @generated from enum value: EMBEDDING_TYPE_SPACY = 1;
   */
  SPACY = 1,

  /**
   * [HuggingFace Transformers](https://huggingface.co/models).
   *
   * @generated from enum value: EMBEDDING_TYPE_TRANSFORMERS = 2;
   */
  TRANSFORMERS = 2,

  /**
   * [UKPLab Sentence Transformers](https://www.sbert.net/docs/pretrained_models.html)
   *
   * @generated from enum value: EMBEDDING_TYPE_SENTENCE_TRANSFORMERS = 3;
   */
  SENTENCE_TRANSFORMERS = 3,

  /**
   * Tensorflow Hub. Example: [Universal Sentence Encoder](https://tfhub.dev/google/universal-sentence-encoder/4).
   *
   * @generated from enum value: EMBEDDING_TYPE_TENSORFLOW_HUB = 4;
   */
  TENSORFLOW_HUB = 4,

  /**
   * [OpenAI](https://platform.openai.com/docs/models/embeddings).
   *
   * @generated from enum value: EMBEDDING_TYPE_OPENAI = 5;
   */
  OPENAI = 5,

  /**
   * [Ollama](https://ollama.com/blog/embedding-models)
   *
   * @generated from enum value: EMBEDDING_TYPE_OLLAMA = 6;
   */
  OLLAMA = 6,

  /**
   * [Cohere](https://docs.cohere.com/reference/embed)
   *
   * @generated from enum value: EMBEDDING_TYPE_COHERE = 7;
   */
  COHERE = 7,

  /**
   * [VoyageAI](https://docs.voyageai.com/docs/embeddings)
   *
   * @generated from enum value: EMBEDDING_TYPE_VOYAGEAI = 8;
   */
  VOYAGEAI = 8,
}

/**
 * @generated from enum arg_services.nlp.v1.EmbeddingType
 */
export declare type EmbeddingTypeJson = "EMBEDDING_TYPE_UNSPECIFIED" | "EMBEDDING_TYPE_SPACY" | "EMBEDDING_TYPE_TRANSFORMERS" | "EMBEDDING_TYPE_SENTENCE_TRANSFORMERS" | "EMBEDDING_TYPE_TENSORFLOW_HUB" | "EMBEDDING_TYPE_OPENAI" | "EMBEDDING_TYPE_OLLAMA" | "EMBEDDING_TYPE_COHERE" | "EMBEDDING_TYPE_VOYAGEAI";

/**
 * Describes the enum arg_services.nlp.v1.EmbeddingType.
 */
export declare const EmbeddingTypeSchema: GenEnum<EmbeddingType, EmbeddingTypeJson>;

/**
 * @generated from service arg_services.nlp.v1.NlpService
 */
export declare const NlpService: GenService<{
  /**
   * Compute embeddings (i.e., vectors) for strings.
   *
   * @generated from rpc arg_services.nlp.v1.NlpService.Vectors
   */
  vectors: {
    methodKind: "unary";
    input: typeof VectorsRequestSchema;
    output: typeof VectorsResponseSchema;
  },
  /**
   * Compute the similarity score between two strings.
   *
   * @generated from rpc arg_services.nlp.v1.NlpService.Similarities
   */
  similarities: {
    methodKind: "unary";
    input: typeof SimilaritiesRequestSchema;
    output: typeof SimilaritiesResponseSchema;
  },
  /**
   * Process strings by spacy and return them as [binary data](https://spacy.io/api/docbin).
   * Locally, spacy can restore this data **without** loading the underlying NLP models into the main memory.
   * Allows one to retrieve all computed attributes (e.g., POS tags, sentences), but can only be used by Python programs.
   *
   * @generated from rpc arg_services.nlp.v1.NlpService.DocBin
   */
  docBin: {
    methodKind: "unary";
    input: typeof DocBinRequestSchema;
    output: typeof DocBinResponseSchema;
  },
}>;