openai-d ~main

OpenAI API for D

This library provides unofficial D clients for OpenAI API.

Overview

Key capabilities are summarized below.

Installation

Once dmd and dub are available, add this library to your project:

dub add openai-d

Quick Start

import openai;

auto client = new OpenAIClient();
auto res = client.chatCompletion(
    chatCompletionRequest(openai.GPT4O, [userChatMessage("Hello!")]));
writeln(res.choices[0].message.content);

See the examples directory for full sample programs.

Features

Endpoint

• [x] OpenAI
• [x] Azure OpenAI Service

API

OpenAI

• [x] Responses API
• [x] Chat
• [x] tools (function_call)
• [x] structured output
• [x] input vision
• [ ] stream (design)
• [ ] Realtime (Beta) (TODO)
• [x] Audio
• [x] speech
• [x] transcription
• [x] translations
• [x] Images
• [x] Embeddings
• [ ] Evals (TODO)
• [ ] Fine-tunings (TODO)
• [ ] Graders (TODO)
• [ ] Batch (TODO)
• [x] Files — upload, list, retrieve, delete, download
• [ ] Uploads (TODO)
• [x] Models
• [x] Moderations
• [ ] Vector stores (TODO)
• [ ] Containers (TODO)
• [ ] Assistants (Beta) (TODO)
• [ ] Administration (WIP)
• [x] Admin API Keys
• [x] Invites
• [x] Users
• [x] Projects
• [x] Project users
• [x] Project service accounts
• [x] Project API keys
• [x] Project rate limits
• [x] Audit logs
• [x] Usage
• [ ] Certificates (Beta WIP)

Legacy

• [x] Completions (Legacy)

Deprecated

• Edits
• Use chat API instead. See: https://platform.openai.com/docs/deprecations/edit-models-endpoint

Optimization

• [ ] Switch HTTP client to requests (TODO)
  • Not adopted because it is less convenient due to Windows' handling of system certificates. Version flag is required for support.
  • Adopting 'requests' is expected to lead to more efficient use of Fiber in vibe.d.

Completion (Legacy)

import std;
import openai;

// Load API key from environment variable
auto client = new OpenAIClient();

// POST /completions
auto message = completionRequest("gpt-3.5-turbo-instruct", "Hello, D Programming Language!\n", 10, 0);
message.stop = "\n";

auto response = client.completion(message);

writeln(response.choices[0].text.chomp());

Embedding

import std;
import openai;

// Load API key from environment variable
auto client = new OpenAIClient();

// POST /embeddings
const request = embeddingRequest("text-embedding-3-small", "Hello, D Programming Language!");
auto response = client.embedding(request);

float[] embedding = response.data[0].embedding;
writeln(embedding.length); // text-embedding-3-small -> 1536

Files

import std.stdio;
import std.file : write;
import openai;

auto client = new OpenAIClient();
auto up = uploadFileRequest("input.jsonl", FilePurpose.FineTune);
auto uploaded = client.uploadFile(up);

auto retrieved = client.retrieveFile(uploaded.id);
writeln("retrieved: ", retrieved.filename);

auto content = client.downloadFileContent(uploaded.id);
write("copy.jsonl", content);

client.deleteFile(uploaded.id);

See examples/files for the full example.

Moderation

import std;
import openai;

// Load API key from environment variable
auto client = new OpenAIClient();

// POST /moderations
const request = moderationRequest("D is a general-purpose programming language with static typing, systems-level access, and C-like syntax. With the D Programming Language, write fast, read fast, and run fast.");
auto response = client.moderation(request);

if (response.results[0].flagged)
    writeln("Warning!");
else
    writeln("Probably safe.");

Transcription

import std;
import openai;

// Load API key from environment variable
auto client = new OpenAIClient();

// POST /audio/transcriptions
auto request = transcriptionRequest("audio.mp3", "whisper-1");
auto response = client.transcription(request);

writeln(response.text);

See examples/audio_transcription for the full example.

Translation

import std;
import openai;

// Load API key from environment variable
auto client = new OpenAIClient();

// POST /audio/translations
auto request = translationRequest("audio.mp3", "whisper-1");
auto response = client.translation(request);

writeln(response.text);

See examples/audio_translation for the full example.

Images

import std;
import openai;

// Load API key from environment variable
auto client = new OpenAIClient();

// POST /images/generations
auto request = imageGenerationRequest("A cute baby sea otter");
request.responseFormat = ImageResponseFormatB64Json;
auto response = client.imageGeneration(request);
write("image.png", Base64.decode(response.data[0].b64Json));

See examples/images for the full example.

Administration

import std;
import openai;

auto client = new OpenAIAdminClient();
auto list = client.listProjects(listProjectsRequest(20));
writeln(list.data.length);

Requires an admin API key. See examples/administration* for complete examples.

OpenAIClientConfig

Environment variables

import std.process;
import openai;

environment["OPENAI_API_KEY"] = "<Your API Key>";
environment["OPENAI_ORGANIZATION"] = "<Your Organization>";
environment["OPENAI_API_BASE"] = "https://example.api.cognitive.microsoft.com"; // optional
environment["OPENAI_DEPLOYMENT_ID"] = "<Your deployment>"; // Azure only
environment["OPENAI_API_VERSION"] = "2024-10-21"; // Azure only

auto config = OpenAIClientConfig.fromEnvironment();

assert(config.apiKey == "<Your API Key>");
assert(config.organization == "<Your Organization>");
assert(config.apiBase == "https://example.api.cognitive.microsoft.com");

Development

Requirements

dmd and dub must be installed. Use the official D installer.

License

This project is released under the MIT license.