Platform Support

Understand which providers are available on each platform and how to handle platform differences.

Overview

Conduit supports macOS, iOS, visionOS, and Linux. Not all providers are available on every platform — local inference requires specific hardware, while cloud providers work everywhere with network access.

Platform Matrix

Platform	Status	Available Providers
macOS 14+	Full	MLX, Anthropic, OpenAI, OpenRouter, Ollama, HuggingFace, Kimi, MiniMax, CoreML, Llama, Foundation Models*
iOS 17+	Full	MLX, Anthropic, OpenAI, OpenRouter, HuggingFace, Kimi, MiniMax, CoreML, Foundation Models*
visionOS 1+	Full	MLX, Anthropic, OpenAI, OpenRouter, HuggingFace, Kimi, MiniMax, CoreML, Foundation Models*
Linux	Cloud-only	Anthropic, OpenAI, OpenRouter, Ollama, HuggingFace, Kimi, MiniMax

*Foundation Models requires iOS 26+, macOS 26+, or visionOS 26+.

Provider Platform Requirements

Local Providers

Provider	Requirements
MLXProvider	Apple Silicon (M1+), macOS 14+ / iOS 17+ / visionOS 1+
FoundationModelsProvider	iOS 26+ / macOS 26+ / visionOS 26+
CoreMLProvider	macOS 15+ / iOS 18+ / tvOS 18+ / watchOS 11+ / visionOS 2+
LlamaProvider	Any platform with llama.swift support
Ollama (via OpenAIProvider)	Any platform running Ollama server

Cloud Providers

All cloud providers work on every platform with network access:

• AnthropicProvider
• OpenAIProvider (OpenAI, OpenRouter, Azure)
• HuggingFaceProvider
• KimiProvider
• MiniMaxProvider

Linux

Conduit supports Linux for server-side Swift deployments. Build and test with Swift 6.2+:

swift build
swift test

No traits are enabled by default, so MLX and Foundation Models dependencies are excluded automatically.

Linux Limitations

• MLX Provider: Requires Apple Silicon with Metal GPU (not available on Linux)
• Foundation Models: Requires Apple platforms (not available on Linux)
• CoreML Provider: Requires Apple frameworks (not available on Linux)
• Image Generation: GeneratedImage.image returns nil on Linux (use data or save(to:))
• Keychain: Token storage falls back to environment variables

Local Inference on Linux

Use Ollama via OpenAIProvider for local inference on Linux:

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model
ollama pull llama3.2

import Conduit

let provider = OpenAIProvider(endpoint: .ollama(), authentication: .none)
let response = try await provider.generate(
    "Hello from Linux!",
    model: .ollama("llama3.2"),
    config: .default
)

Conditional Compilation

Use trait flags to guard platform-specific code:

#if CONDUIT_TRAIT_MLX
// MLX is available
let mlx = MLXProvider()
#endif

#if CONDUIT_TRAIT_ANTHROPIC
// Anthropic is available
let anthropic = AnthropicProvider(apiKey: "...")
#endif

#if canImport(FoundationModels)
// Foundation Models is available (iOS 26+, macOS 26+)
if #available(iOS 26.0, macOS 26.0, visionOS 26.0, *) {
    let fm = FoundationModelsProvider()
}
#endif

Writing Cross-Platform Code

Use TextGenerator as a generic constraint for fully portable code:

func generate<P: TextGenerator>(
    with provider: P,
    model: P.ModelID,
    prompt: String
) async throws -> String {
    try await provider.generate(prompt, model: model, config: .default)
}

This function works with any provider on any platform.

Minimum Deployment Targets

Platform	Minimum Version
iOS	17.0
macOS	14.0
visionOS	1.0
Linux	Swift 6.2+