A Go LLM SDK for Streaming and Tool-Calling AI Backends (with a React Frontend Library)

Grafana has released ai-sdk, which supports LLM calls, streaming, tool-calling, and structured outputs in the Go language environment. This SDK shares its design philosophy with Vercel's AI SDK, and its distinguishing feature is maintaining wire compatibility with TypeScript frontends.

The key features are as follows:

• React compatibility: SSE (Server-Sent Events) generated from a Go backend can be used directly in React hooks such as useChat, useCompletion, and useObject without any separate protocol conversion.
• Support for various models: Supports Anthropic, Amazon Bedrock, OpenAI, and OpenAI-compatible APIs.
• Tools and structured outputs: Regular Go functions can be invoked as model tools, and schema-validated data can be generated in object and array formats.
• Production environment controls: Provides features suitable for production environments, supporting timeouts, fallback, logging, Prometheus metrics, and agent observability settings.