The Jetify AI SDK for Go offers a unified interface for Go developers to build AI applications and agents, addressing the fragmentation and vendor lock-in common with interacting across multiple AI providers. It provides an idiomatic Go experience, abstracting provider-specific APIs and enabling robust, multi-modal AI development with production-ready features.

How It Works

This SDK provides a provider-agnostic abstraction layer for language models, embeddings, and image generation, inspired by Vercel's TypeScript SDK. It prioritizes idiomatic Go design, strong typing, and production-readiness with features like automatic retries, rate limiting, and provider failover. Key advantages include first-class support for multi-modal inputs (text, images, files) and tool calling across different providers, simplifying complex AI integrations.

Quick Start & Requirements

• Install: go get go.jetify.com/ai
• Prerequisites: Go development environment.
• Links: Examples directory, API Reference documentation.

Highlighted Details

• Multi-Provider Support: Integrates with OpenAI, Anthropic, with more planned.
• Multi-Modal Inputs: Handles text, images, and files within conversations.
• Tool Calling: Supports function calling with parallel execution.
• Language Models: Offers text generation with streaming and non-streaming support.
• Provider-Specific Features: Includes OpenAI's web search, computer use, and file search tools, and Anthropic's advanced reasoning and tool use.

Maintenance & Community

Maintained by Jetify, the project offers community support via Discord (https://discord.gg/jetify), GitHub Discussions, and GitHub Issues for bug reports and feature requests.

Licensing & Compatibility

Licensed under the Apache 2.0 License, permitting commercial use and integration with closed-source applications.

Limitations & Caveats

The SDK is currently in Public Alpha, meaning breaking changes are expected and may occur with minor version updates. It is not yet stable enough for production use at scale, with General Availability (v1) planned for later.