fm.cr

Crystal bindings for Apple's FoundationModels framework. Run on-device AI powered by Apple Intelligence directly from Crystal.

Requires macOS 26+ (Tahoe) with Apple Intelligence enabled.

Overview

fm.cr provides a Crystal interface to Apple's on-device language model through a Swift FFI bridge. It supports text generation, streaming, structured output with JSON schemas, tool calling, and context management -- all running locally on Apple Silicon.

Quick Links

• Getting Started -- Installation, prerequisites, and your first program
• Basic Usage -- Models, sessions, and generating responses
• API Reference -- Complete API documentation

Features

• Text Generation -- Synchronous and streaming responses
• Structured Output -- Type-safe responses using Generable or raw JSON schemas
• Tool Calling -- Extend model capabilities with custom tools
• Context Management -- Transcript export, token estimation, and automatic compaction
• Session Restore -- Save and restore conversation state via transcripts

Installation

Add fm.cr to your shard.yml:

dependencies:
  fm:
    github: hahwul/fm.cr

Then run:

shards install

The native Swift FFI library is built automatically via the postinstall script.