Crate openai_rust_sdk 

OpenAI Rust SDK

openai_rust_sdk is an unofficial, async, type-safe Rust client for the OpenAI API. The Responses API is the recommended starting point for new integrations; the crate also covers conversations, streaming and tools, Batch, files, vector stores, media, Realtime, fine-tuning, evals, and administration APIs.

This community-maintained crate is not affiliated with, endorsed by, or maintained by OpenAI. The API changes quickly, so consult the dated coverage matrix before depending on a particular endpoint in production.

Quick start

Set OPENAI_API_KEY and choose OPENAI_MODEL explicitly (for example, gpt-5.6-luna), then create a response:

use openai_rust_sdk::{OpenAIError, from_env, models::CreateResponseRequest};

#[tokio::main]
async fn main() -> openai_rust_sdk::Result<()> {
    let client = from_env()?;
    let model = std::env::var("OPENAI_MODEL")
        .ok()
        .map(|value| value.trim().to_owned())
        .filter(|value| !value.is_empty())
        .ok_or_else(|| {
            OpenAIError::InvalidRequest(
                "OPENAI_MODEL must be set to a non-empty model ID".to_owned(),
            )
        })?;
    let request = CreateResponseRequest::new_text(
        model,
        "Explain Rust's ownership model in one concise sentence.",
    );

    let response = client.create_response_v2(&request).await?;
    println!("{}", response.output_text());
    Ok(())
}

from_env also accepts the optional OPENAI_BASE_URL environment variable. A custom base URL must be an API origin without /v1 or a trailing slash. The API key is sent as a bearer token to that origin, so only configure endpoints you trust.

API lifecycle

Responses and Conversations are the modern integration path. Assistants, Threads, and Runs remain available for migration, but OpenAI has deprecated the Assistants API and announced its shutdown for August 26, 2026. See OpenAI’s migration guide.

Cargo features

• default: the core OpenAI API client.
• yara: local validation and CLI tooling for the optional YARA-X Batch API example.
• testing: reserved compatibility feature; currently adds no dependencies.
• full: all optional capabilities; currently enables testing and yara.

See the YARA-X Batch example for exact CLI commands and security considerations.

Reliability

The shared HTTP client does not currently apply automatic retries or expose a global request timeout. Applications should add workload-appropriate timeouts, exponential backoff with jitter, and idempotency handling. More details are in the configuration guide.

Complete, runnable programs live in the repository’s examples directory, starting with quickstart.rs and responses_api.rs.

Re-exports

pub use api::batch::BatchApi;

pub use api::streaming::StreamingApi;

pub use api::AdminApi;

pub use api::AssistantsApi;

pub use api::ConversationsApi;

pub use api::CustomToolsApi;

pub use api::EvalsApi;

pub use api::FineTuningApi;

pub use api::FunctionsApi;

pub use api::RealtimeAudioApi;

pub use api::ResponsesApi;

pub use api::RunsApi;

pub use api::SkillsApi;

pub use api::ThreadsApi;

pub use api::UploadsApi;

pub use api::VectorStoresApi;

pub use api::VideosApi;

pub use builders::FunctionBuilder;

pub use builders::ObjectSchemaBuilder;

pub use client::ChatBuilder;

pub use client::OpenAIClient;

pub use client::from_env;

pub use client::from_env_with_base_url;

pub use error::OpenAIError;

pub use error::Result;

pub use models::responses_v2::CreateResponseRequest;

pub use models::responses_v2::InputTokenCountResponse;

pub use models::responses_v2::ResponseStreamEvent;

pub use prompt_engineering::Example;

pub use prompt_engineering::PromptBuilder;

pub use prompt_engineering::PromptPatterns;

pub use prompt_engineering::PromptTemplateBuilder;

pub use prompt_engineering::XmlContentBuilder;

pub use schema::EnhancedSchemaBuilder;

pub use schema::JsonSchema;

pub use schema::SchemaBuilder;

pub use testing::batch_generator::BatchJobGenerator;

pub use models::realtime::*;

pub use models::assistants::*;

pub use models::functions::*;

pub use models::responses::*;

Modules

api

OpenAI API functionality

builders

Function and schema builders

client

Client implementations and builders

constants

Constants used throughout the SDK Constants used throughout the OpenAI API SDK

error

Error types and handling

helpers

Helper utilities to reduce code duplication Helper modules for reducing code duplication

macros

Common macros for reducing code duplication Common macros for reducing code duplication across the codebase

models

Data models and types

prompt_engineering

Prompt engineering utilities and builders

schema

JSON Schema utilities JSON Schema builder for structured outputs

testing

Utilities for the optional YARA-X Batch API example

webhooks

Offline verification and parsing of OpenAI webhook deliveries. Verify webhook deliveries before parsing their exact raw request bodies.

Macros

file_err

Convenience macro for FileError mapping with format string

http_delete

Macro to generate HTTP DELETE method wrappers

http_delete_beta

Macro to generate HTTP DELETE method wrappers with beta headers

http_get

Macro to generate simple HTTP GET method wrappers

http_get_beta

Macro to generate HTTP GET method wrappers with beta headers

http_post

Macro to generate HTTP POST method wrappers

http_post_beta

Macro to generate HTTP POST method wrappers with beta headers

impl_api_client

Macro to implement the ApiClient trait for a struct that contains an HttpClient

impl_api_test_suite

Macro to generate comprehensive test suites for API clients and models

impl_audio_format_methods

Macro to generate common audio format methods for audio builders

impl_builder

Macro to implement the basic Builder trait

impl_builder_build

Macro to generate standard builder build() methods with field validation

impl_builder_methods

Macro for implementing common builder methods

impl_builder_setters

Macro to implement standard builder pattern setter methods

impl_bytes_to_human_readable

Macro to generate bytes to human readable formatting

impl_bytes_usage

Generic macro to implement BytesUsage

impl_comprehensive_test_suite

Macro to generate comprehensive test suites that can replace entire test modules

impl_default_object_type

Macro to generate default object type functions

impl_delete_response

Generic macro to implement DeleteResponse

impl_enum_display

Macro to generate Display implementations for enums that serialize as lowercase strings

impl_file_supported

Generic macro to implement FileSupported

impl_filterable_list_response

Generic macro to implement FilterableListResponse

impl_fine_tuning_params

Macro to generate fine-tuning job parameter types with pagination

impl_fluent_setters

Macro to generate fluent setter methods for builders

impl_format_methods

Macro to generate common format methods for builders

impl_image_size_methods

Macro to generate common size methods for image builders

impl_list_params

Macro to generate list parameter structures with pagination support

impl_list_params_with_builder

Macro to generate common list parameter types with builder pattern methods This eliminates massive duplication in vector stores, fine tuning, and similar modules

impl_list_query_params

Macro to implement ListQueryParams trait for pagination structures

impl_list_response

Macro to generate list response structures

impl_list_response_methods

Macro to generate list response filtering methods for responses with data field

impl_map_setters

Macro to generate builder methods for HashMap fields

impl_metadata_builder

Generic macro to implement MetadataBuilder for any builder type

impl_option_setters

Macro to generate builder methods for optional non-string fields

impl_pagination_params

Generic macro to implement PaginationParams

impl_params_test_suite

Macro to generate parameter testing suites with pagination

impl_query_params

Macro for implementing query parameter building with common fields

impl_response_constructors

Macro to generate success/failure response constructors

impl_run_builder_methods

Macro to generate run builder methods that delegate to RunConfigurationBuilder trait

impl_run_config_builder

Macro to generate RunConfigurationBuilder trait implementations

impl_status_enum

Generic macro to implement StatusEnum for snake_case serialized enums

impl_status_filters

Macro to generate status filtering methods for list responses

impl_status_methods

Macro to generate status checking methods for entities

impl_string_setters

Macro to generate builder methods for optional string fields

impl_test_helpers

Macro to generate common test helper functions for request/response testing

impl_transcription_format_methods

Macro to generate common transcription format methods

impl_usage_methods

Macro to generate usage bytes methods for list responses

impl_validation

Macro to implement common validation methods

impl_vec_setters

Macro to generate builder methods for vector fields

impl_with_format

Macro to implement WithFormat trait

impl_with_n

Macro to implement WithN trait

impl_with_quality

Macro to implement WithQuality trait

impl_with_size

Macro to implement WithSize trait

impl_with_speed

Macro to implement WithSpeed trait

impl_with_temperature

Macro to implement WithTemperature trait

impl_with_user

Macro to implement WithUser trait

invalid_request_err

Convenience macro for InvalidRequest mapping

map_err

Macro to simplify error mapping with format strings Usage: map_err!(FileError, "Failed to read file: {}")

network_err

Convenience macro for Unknown error mapping (network errors)

parse_err

Convenience macro for ParseError mapping

parsing_err

Convenience macro for parsing errors that go to InvalidRequest

request_err

Convenience macro for RequestError mapping

streaming_err

Convenience macro for Streaming error mapping

validation_err

Convenience macro for validation errors (parsing errors that go to InvalidRequest)

Traits

De

A data structure that can be deserialized from any data format supported by Serde.

Ser

A data structure that can be serialized into any data format supported by Serde.

Derive Macros

De

Derive macro available if serde is built with features = ["derive"].

Ser

Derive macro available if serde is built with features = ["derive"].