using System;
using System.ClientModel;
using System.ClientModel.Primitives;
using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Text.Json;
using System.Threading;
using System.Threading.Tasks;
namespace OpenAI.Embeddings;
// CUSTOM:
// - Renamed.
// - Suppressed constructor that takes endpoint parameter; endpoint is now a property in the options class.
// - Suppressed methods that only take the options parameter.
/// <summary> The service client for OpenAI embedding operations. </summary>
[CodeGenClient("Embeddings")]
[CodeGenSuppress("EmbeddingClient", typeof(ClientPipeline), typeof(ApiKeyCredential), typeof(Uri))]
[CodeGenSuppress("CreateEmbeddingAsync", typeof(EmbeddingGenerationOptions), typeof(CancellationToken))]
[CodeGenSuppress("CreateEmbedding", typeof(EmbeddingGenerationOptions), typeof(CancellationToken))]
public partial class EmbeddingClient
{
private readonly string _model;
// CUSTOM: Added as a convenience.
/// <summary> Initializes a new instance of <see cref="EmbeddingClient"/>. </summary>
/// <param name="model"> The name of the model to use in requests sent to the service. To learn more about the available models, see <see href="https://platform.openai.com/docs/models"/>. </param>
/// <param name="apiKey"> The API key to authenticate with the service. </param>
/// <exception cref="ArgumentNullException"> <paramref name="model"/> or <paramref name="apiKey"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="model"/> is an empty string, and was expected to be non-empty. </exception>
public EmbeddingClient(string model, string apiKey) : this(model, new ApiKeyCredential(apiKey), new OpenAIClientOptions())
{
}
// CUSTOM:
// - Added `model` parameter.
// - Used a custom pipeline.
// - Demoted the endpoint parameter to be a property in the options class.
/// <summary> Initializes a new instance of <see cref="EmbeddingClient"/>. </summary>
/// <param name="model"> The name of the model to use in requests sent to the service. To learn more about the available models, see <see href="https://platform.openai.com/docs/models"/>. </param>
/// <param name="credential"> The API key to authenticate with the service. </param>
/// <exception cref="ArgumentNullException"> <paramref name="model"/> or <paramref name="credential"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="model"/> is an empty string, and was expected to be non-empty. </exception>
public EmbeddingClient(string model, ApiKeyCredential credential) : this(model, credential, new OpenAIClientOptions())
{
}
// CUSTOM:
// - Added `model` parameter.
// - Used a custom pipeline.
// - Demoted the endpoint parameter to be a property in the options class.
/// <summary> Initializes a new instance of <see cref="EmbeddingClient"/>. </summary>
/// <param name="model"> The name of the model to use in requests sent to the service. To learn more about the available models, see <see href="https://platform.openai.com/docs/models"/>. </param>
/// <param name="credential"> The API key to authenticate with the service. </param>
/// <param name="options"> The options to configure the client. </param>
/// <exception cref="ArgumentNullException"> <paramref name="model"/> or <paramref name="credential"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="model"/> is an empty string, and was expected to be non-empty. </exception>
public EmbeddingClient(string model, ApiKeyCredential credential, OpenAIClientOptions options)
{
Argument.AssertNotNullOrEmpty(model, nameof(model));
Argument.AssertNotNull(credential, nameof(credential));
options ??= new OpenAIClientOptions();
_model = model;
Pipeline = OpenAIClient.CreatePipeline(credential, options);
_endpoint = OpenAIClient.GetEndpoint(options);
}
// CUSTOM:
// - Added `model` parameter.
// - Used a custom pipeline.
// - Demoted the endpoint parameter to be a property in the options class.
// - Made protected.
/// <summary> Initializes a new instance of <see cref="EmbeddingClient"/>. </summary>
/// <param name="pipeline"> The HTTP pipeline to send and receive REST requests and responses. </param>
/// <param name="model"> The name of the model to use in requests sent to the service. To learn more about the available models, see <see href="https://platform.openai.com/docs/models"/>. </param>
/// <param name="options"> The options to configure the client. </param>
/// <exception cref="ArgumentNullException"> <paramref name="pipeline"/> or <paramref name="model"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="model"/> is an empty string, and was expected to be non-empty. </exception>
protected internal EmbeddingClient(ClientPipeline pipeline, string model, OpenAIClientOptions options)
{
Argument.AssertNotNull(pipeline, nameof(pipeline));
Argument.AssertNotNullOrEmpty(model, nameof(model));
options ??= new OpenAIClientOptions();
_model = model;
Pipeline = pipeline;
_endpoint = OpenAIClient.GetEndpoint(options);
}
// CUSTOM: Added to simplify generating a single embedding from a string input.
/// <summary> Generates an embedding representing the text input. </summary>
/// <param name="input"> The text input to generate an embedding for. </param>
/// <param name="options"> The options to configure the embedding generation. </param>
/// <param name="cancellationToken"> A token that can be used to cancel this method call. </param>
/// <exception cref="ArgumentNullException"> <paramref name="input"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="input"/> is an empty string, and was expected to be non-empty. </exception>
public virtual async Task<ClientResult<OpenAIEmbedding>> GenerateEmbeddingAsync(string input, EmbeddingGenerationOptions options = null, CancellationToken cancellationToken = default)
{
Argument.AssertNotNullOrEmpty(input, nameof(input));
options ??= new();
CreateEmbeddingGenerationOptions(input, ref options);
using BinaryContent content = options;
ClientResult result = await GenerateEmbeddingsAsync(content, cancellationToken.ToRequestOptions()).ConfigureAwait(false);
return ClientResult.FromValue(((OpenAIEmbeddingCollection)result).FirstOrDefault(), result.GetRawResponse());
}
// CUSTOM: Added to simplify generating a single embedding from a string input.
/// <summary> Generates an embedding representing the text input. </summary>
/// <param name="input"> The text input to generate an embedding for. </param>
/// <param name="options"> The options to configure the embedding generation. </param>
/// <param name="cancellationToken"> A token that can be used to cancel this method call. </param>
/// <exception cref="ArgumentNullException"> <paramref name="input"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="input"/> is an empty string, and was expected to be non-empty. </exception>
public virtual ClientResult<OpenAIEmbedding> GenerateEmbedding(string input, EmbeddingGenerationOptions options = null, CancellationToken cancellationToken = default)
{
Argument.AssertNotNullOrEmpty(input, nameof(input));
options ??= new();
CreateEmbeddingGenerationOptions(input, ref options);
using BinaryContent content = options;
ClientResult result = GenerateEmbeddings(content, cancellationToken.ToRequestOptions());
return ClientResult.FromValue(((OpenAIEmbeddingCollection)result).FirstOrDefault(), result.GetRawResponse());
}
// CUSTOM: Added to simplify passing the input as a collection of strings instead of BinaryData.
/// <summary> Generates embeddings representing the text inputs. </summary>
/// <param name="inputs"> The text inputs to generate embeddings for. </param>
/// <param name="options"> The options to configure the embedding generation. </param>
/// <param name="cancellationToken"> A token that can be used to cancel this method call. </param>
/// <exception cref="ArgumentNullException"> <paramref name="inputs"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="inputs"/> is an empty collection, and was expected to be non-empty. </exception>
public virtual async Task<ClientResult<OpenAIEmbeddingCollection>> GenerateEmbeddingsAsync(IEnumerable<string> inputs, EmbeddingGenerationOptions options = null, CancellationToken cancellationToken = default)
{
Argument.AssertNotNullOrEmpty(inputs, nameof(inputs));
options ??= new();
CreateEmbeddingGenerationOptions(inputs, ref options);
using BinaryContent content = options;
ClientResult result = await GenerateEmbeddingsAsync(content, cancellationToken.ToRequestOptions()).ConfigureAwait(false);
return ClientResult.FromValue((OpenAIEmbeddingCollection)result, result.GetRawResponse());
}
// CUSTOM: Added to simplify passing the input as a collection of strings instead of BinaryData.
/// <summary> Generates embeddings representing the text inputs. </summary>
/// <param name="inputs"> The text inputs to generate embeddings for. </param>
/// <param name="options"> The options to configure the embedding generation. </param>
/// <param name="cancellationToken"> A token that can be used to cancel this method call. </param>
/// <exception cref="ArgumentNullException"> <paramref name="inputs"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="inputs"/> is an empty collection, and was expected to be non-empty. </exception>
public virtual ClientResult<OpenAIEmbeddingCollection> GenerateEmbeddings(IEnumerable<string> inputs, EmbeddingGenerationOptions options = null, CancellationToken cancellationToken = default)
{
Argument.AssertNotNullOrEmpty(inputs, nameof(inputs));
options ??= new();
CreateEmbeddingGenerationOptions(inputs, ref options);
using BinaryContent content = options;
ClientResult result = GenerateEmbeddings(content, cancellationToken.ToRequestOptions());
return ClientResult.FromValue((OpenAIEmbeddingCollection)result, result.GetRawResponse());
}
// CUSTOM: Added to simplify passing the input as a collection of ReadOnlyMemory tokens instead of BinaryData.
/// <summary> Generates embeddings representing the tokenized text inputs. </summary>
/// <param name="inputs"> The tokenized text inputs to generate embeddings for. </param>
/// <param name="options"> The options to configure the embedding generation. </param>
/// <param name="cancellationToken"> A token that can be used to cancel this method call. </param>
/// <exception cref="ArgumentNullException"> <paramref name="inputs"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="inputs"/> is an empty collection, and was expected to be non-empty. </exception>
public virtual async Task<ClientResult<OpenAIEmbeddingCollection>> GenerateEmbeddingsAsync(IEnumerable<ReadOnlyMemory<int>> inputs, EmbeddingGenerationOptions options = null, CancellationToken cancellationToken = default)
{
Argument.AssertNotNullOrEmpty(inputs, nameof(inputs));
options ??= new();
CreateEmbeddingGenerationOptions(inputs, ref options);
using BinaryContent content = options;
ClientResult result = await GenerateEmbeddingsAsync(content, cancellationToken.ToRequestOptions()).ConfigureAwait(false);
return ClientResult.FromValue((OpenAIEmbeddingCollection)result, result.GetRawResponse());
}
// CUSTOM: Added to simplify passing the input as a collection of ReadOnlyMemory of tokens instead of BinaryData.
/// <summary> Generates embeddings representing the tokenized text inputs. </summary>
/// <param name="inputs"> The tokenized text inputs to generate embeddings for. </param>
/// <param name="options"> The options to configure the embedding generation. </param>
/// <param name="cancellationToken"> A token that can be used to cancel this method call. </param>
/// <exception cref="ArgumentNullException"> <paramref name="inputs"/> is null. </exception>
/// <exception cref="ArgumentException"> <paramref name="inputs"/> is an empty collection, and was expected to be non-empty. </exception>
public virtual ClientResult<OpenAIEmbeddingCollection> GenerateEmbeddings(IEnumerable<ReadOnlyMemory<int>> inputs, EmbeddingGenerationOptions options = null, CancellationToken cancellationToken = default)
{
Argument.AssertNotNullOrEmpty(inputs, nameof(inputs));
options ??= new();
CreateEmbeddingGenerationOptions(inputs, ref options);
using BinaryContent content = options;
ClientResult result = GenerateEmbeddings(content, cancellationToken.ToRequestOptions());
return ClientResult.FromValue((OpenAIEmbeddingCollection)result, result.GetRawResponse());
}
private void CreateEmbeddingGenerationOptions(string input, ref EmbeddingGenerationOptions options)
{
using MemoryStream stream = new();
using Utf8JsonWriter writer = new(stream);
writer.WriteStringValue(input);
writer.Flush();
options.Input = BinaryData.FromBytes(stream.ToArray());
options.Model = _model;
options.EncodingFormat = InternalCreateEmbeddingRequestEncodingFormat.Base64;
}
private void CreateEmbeddingGenerationOptions(IEnumerable<string> inputs, ref EmbeddingGenerationOptions options)
{
using MemoryStream stream = new();
using Utf8JsonWriter writer = new(stream);
writer.WriteStartArray();
foreach (string input in inputs)
{
writer.WriteStringValue(input);
}
writer.WriteEndArray();
writer.Flush();
options.Input = BinaryData.FromBytes(stream.ToArray());
options.Model = _model;
options.EncodingFormat = InternalCreateEmbeddingRequestEncodingFormat.Base64;
}
private void CreateEmbeddingGenerationOptions(IEnumerable<ReadOnlyMemory<int>> inputs, ref EmbeddingGenerationOptions options)
{
using MemoryStream stream = new();
using Utf8JsonWriter writer = new(stream);
writer.WriteStartArray();
foreach (ReadOnlyMemory<int> input in inputs)
{
writer.WriteStartArray();
foreach (int tokenId in input.ToArray())
{
writer.WriteNumberValue(tokenId);
}
writer.WriteEndArray();
}
writer.WriteEndArray();
writer.Flush();
options.Input = BinaryData.FromBytes(stream.ToArray());
options.Model = _model;
options.EncodingFormat = InternalCreateEmbeddingRequestEncodingFormat.Base64;
}
}openai/openai-dotnet
Publicmirrored from https://github.com/openai/openai-dotnetAvailable
src/Custom/Embeddings/EmbeddingClient.cs
266lines · modepreview