Crate ai_sdk_openai

Crate ai_sdk_openai 

Source
Expand description

§OpenAI Provider Implementation

Production-ready OpenAI provider implementing the AI SDK provider specification. This crate delivers complete access to OpenAI’s model portfolio including GPT language models, DALL-E image generation, Whisper transcription, embeddings, and text-to-speech capabilities.

§Supported Models

  • GPT Language Models - GPT-4, GPT-4 Turbo, GPT-3.5 Turbo, o1 (reasoning models)
  • Text Embeddings - text-embedding-3-small, text-embedding-3-large, ada-002
  • Image Generation - DALL-E 2, DALL-E 3 with quality and style controls
  • Speech Synthesis - TTS-1 (standard), TTS-1-HD (high definition)
  • Speech Transcription - Whisper-1 with timestamps and translations

§Features

  • Full Specification Compliance: Implements all v3 provider traits
  • Streaming Support: Server-sent events for real-time token streaming
  • Tool Calling: Native function calling with parallel tool execution
  • Vision Support: Multimodal inputs with image URLs and base64 data
  • Structured Output: JSON mode and response format constraints
  • Error Handling: Comprehensive error types with retry guidance
  • Request Inspection: Access to raw request/response bodies for debugging

§Quick Start

§Basic Text Generation

ⓘ
use ai_sdk_openai::OpenAIChatModel;
use ai_sdk_provider::LanguageModel;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let api_key = std::env::var("OPENAI_API_KEY")?;
    let model = OpenAIChatModel::new("gpt-4", api_key);

    let response = model
        .generate("Explain photosynthesis in simple terms")
        .temperature(0.7)
        .max_tokens(200)
        .await?;

    println!("{}", response.text());
    Ok(())
}

§Using the Provider Factory

ⓘ
use ai_sdk_openai::OpenAIProvider;
use ai_sdk_provider::ProviderV3;

let provider = OpenAIProvider::new("your-api-key");
let chat_model = provider.language_model("gpt-4");
let embedding_model = provider.embedding_model("text-embedding-3-small");

§Configuration

§API Key

Obtain your OpenAI API key from: https://platform.openai.com/api-keys

Set via environment variable:

export OPENAI_API_KEY=sk-...

Or pass directly to model constructors:

ⓘ
let model = OpenAIChatModel::new("gpt-4", "sk-...");

§Custom Base URL

Use OpenAI-compatible endpoints:

ⓘ
use ai_sdk_openai::{OpenAIChatModel, OpenAIConfig};

let config = OpenAIConfig::new("your-api-key")
    .base_url("https://your-proxy.com/v1");
let model = OpenAIChatModel::from_config("gpt-4", config);

Modules§

model_detection
Model detection utilities for OpenAI models.
responses
OpenAI Responses API implementation.

Structs§

OpenAIChatModel
OpenAI chat completion implementation of the language model interface.
OpenAIConfig
Configuration for OpenAI API endpoints and authentication.
OpenAIEmbeddingModel
OpenAI implementation of embedding model.
OpenAIImageModel
OpenAI implementation of image model.
OpenAIProvider
OpenAI provider for creating and managing all OpenAI API model instances.
OpenAISpeechModel
OpenAI implementation of speech model.
OpenAITranscriptionModel
OpenAI implementation of transcription model.
OpenAIUrlOptions
Parameters for dynamic URL construction in OpenAI API requests.

Enums§

MultimodalError
Error types for multimodal conversion
OpenAIContentPart
OpenAI content part (can be text, image, or audio)
OpenAIError
Errors that can occur when using the OpenAI provider.

Functions§

convert_audio_part
Converts an audio file to OpenAI format
convert_image_part
Converts an image file to OpenAI format