openai-ergonomic 0.26.0

Ergonomic Rust wrapper for OpenAI API
docs.rs failed to build openai-ergonomic-0.26.0
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Visit the last successful build: openai-ergonomic-0.26.1

openai-ergonomic

Crates.io docs.rs MSRV Dependencies CI License

Ergonomic Rust client for the OpenAI API and the services that speak it (Azure OpenAI, OpenRouter, Groq, Ollama, gateways): fluent builders for chat completions, the Responses API, embeddings, images, audio, files, assistants, batches and more, on top of the generated openai-client-base. It sends through a tower-based HTTP client (timvw-rust-runtime), natively and on WebAssembly.

Quick start

cargo add openai-ergonomic
cargo add tokio --features full
use openai_ergonomic::Client;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Reads OPENAI_API_KEY (and the other OPENAI_* / AZURE_OPENAI_* variables).
    let client = Client::from_env()?.build();
    let response = client
        .send_chat(client.chat().system("Be brief.").user("Hello, world!"))
        .await?;
    println!("{}", response.content().unwrap_or_default());
    Ok(())
}

Getting started goes on from here: the Responses API, streaming, errors and the examples.

Key features

  • Builders for chat completions, the Responses API (/responses, with typed streaming events and stored-response operations), embeddings, images, audio, files, uploads, vector stores, assistants and threads, batches, fine-tuning, moderations and usage (getting started, API coverage).
  • Providers: OpenAI, Azure OpenAI, OpenRouter, Groq, Ollama or any OpenAI-compatible API, with refreshable credentials (Entra ID tokens, gateways) through an AuthProvider (configuration).
  • A tower HTTP stack with retries (honouring Retry-After), timeouts, stream idle timeouts, rate-limit headers, a retry classification of errors and layers of your own (configuration).
  • Streaming of chat completions and of Responses API events, with completion reports (streaming).
  • Tools for chat completions and the Responses API: typed function tools, tools that return images or files, custom (grammar) tools, namespaces and tools only known at runtime (MCP), in a registry that answers every call (tools).
  • An agent loop on both APIs: ToolLoop runs the model's tool calls until it answers, with history repair, calls started while the response streams, ordered parallel dispatch, retries of whole requests, cancellation, deferred tools with a tool search, stateless requests and per-call rate limits (agent loop).
  • Interceptors around every request and stream, with a Langfuse / OpenTelemetry integration (interceptors).
  • Bring your own types: send and decode types of your own through the same client (chat and responses).
  • Files and media: parallel, retried large uploads; images, speech and file contents streamed to disk (files and media).
  • Webhooks: verify OpenAI webhook signatures and decode their events (webhooks).
  • WebAssembly: builds for wasm32-unknown-unknown (browsers, Node.js, Cloudflare Workers) (WebAssembly).

Documentation

Getting started installing, a client, chat, the Responses API, errors, the examples
Configuration providers (Azure, OpenRouter, Groq, Ollama, custom), authentication, retries, timeouts, custom HTTP clients, per-call options
Chat and responses, Responses workflows the builders, structured output, vision, bring your own types
Streaming, completion contract streamed chat completions and responses, completion reports
Tool framework, tool orchestration typed tools, the registry, answering calls on both APIs, multi-turn tool calling
Agent loop ToolLoop: the tool loop on both APIs, history repair, parallel calls, retries, cancellation, tool search
Interceptors, Langfuse integration hooks around requests and streams, typed state, Langfuse / OpenTelemetry
Files and media, vector stores uploads (Upload), large files, images, speech, downloads
Webhooks signature verification and events (feature webhook)
WebAssembly building for wasm32-unknown-unknown, what differs, the Node.js tests
Migration upgrading release by release
Examples index, architecture the examples/ directory; how the crate is put together
Development, contributing building and testing, local CI, the WebAssembly suite
API reference docs.rs

License

Licensed under either of Apache License, Version 2.0 or MIT license at your option.

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.