aiface

aiface is a versatile and unified multi-provider client for large language models (LLMs) in Rust. It streamlines interactions with over 30 LLM providers, including renowned names like Claude, OpenAI, Gemini, and GitHub Copilot, through a single, type-safe API. This capability allows developers to easily switch between providers, enhancing flexibility and functionality in AI applications.

Key Features

• Unified Connector: Utilize one Connector to access various providers seamlessly. Change the provider simply by modifying the name passed to Connector::new.
• Robust Streaming with Error Handling: Supports automatic retries upon encountering network errors (429/5xx) with built-in backoff strategies. A reset marker helps discard any partial responses, ensuring smooth chat experiences.
• Context Awareness: A comprehensive model catalog facilitates context-window discovery and classifies errors intelligently. This feature aids in maintaining efficient interactions, even when unexpected issues arise.
• Integrated OAuth Support: Offers built-in support for various OAuth flows, including Authorization Code and PKCE. Tokens are securely stored using the operating system's keyring, providing enhanced security for sensitive credentials.
• Multimodal Capabilities: Allows for inline transmission of images in base64 format through chat, enabling rich content generation. The generate_image function produces images via participating providers. Chat responses can also include embedded images for a more vivid interaction experience.
• Type-Safe Decision Making: Introduces System One decision models that allow for probability-based typed decisions under one Connector::decide contract, simplifying complex decision-making processes.

Example Usage

The following Rust snippet demonstrates how to set up a chat connection with the OpenAI provider:

use aiface::connector::Connector;

let reply = Connector::new("openai")?
    .with_model("gpt-4o")
    .with_temperature(0.7)
    .with_api_key("sk-...")
    .chat("Explain Rust ownership")
    .await?;

Documentation

Comprehensive guides for all the available modules can be found in the docs/ directory. This includes sections on the connector overview, client builder API, decision models, and more. Additionally, a practical walkthrough is available in the example located at examples/client.rs.

To leverage the powerful capabilities of aiface, Rust edition 2024 or higher (≥ 1.85) is required. This project offers a streamlined approach for developers looking to integrate multiple LLMs into their applications, enhancing productivity and capabilities in AI-driven projects.