Swift LLM Provider Integration

A reusable Codex skill and Swift provider kit for integrating multi-provider LLM support into iOS and macOS apps.

This repository packages one skill with a deployment-agnostic integration model:

• provider abstraction first
• direct or relay transport as a project choice
• reusable Swift templates for provider adapters and streaming support

What It Includes

• SKILL.md: the skill definition and workflow
• agents/openai.yaml: UI metadata for skill discovery
• assets/SwiftLLMProviderKit/: reusable Swift source templates
• references/: abstraction model, provider matrix, playbook, validation checklist
• scripts/install_provider_kit.py: copy the Swift kit into a target app
• scripts/validate_repo.py: validate frontmatter and Swift template type-checking

Scope

This skill is aimed at Swift projects that need a reusable LLM integration layer for:

• OpenAI Responses and Chat Completions
• OpenAI-compatible providers
• Aliyun DashScope / Qwen variants
• Doubao Ark chat and responses flows
• Tencent Hunyuan compatible API and official signed API
• direct provider access or backend relay access

Installation

Clone this repository, then place it in your Codex skills directory.

A typical local setup looks like this:

git clone https://github.com/voderment/LLM-Integration.git ~/.codex/skills/swift-llm-provider-integration

If you keep it elsewhere, copy or symlink the repository folder into your Codex skills directory.

Using the Skill

Example prompt:

Use $swift-llm-provider-integration to add complete OpenAI and multi-provider LLM support to this Swift app.

Installing the Swift Provider Kit Into a Project

python3 scripts/install_provider_kit.py --dest /absolute/path/to/YourApp/Core/LLM

This copies the reusable Swift files into your target project.

Core Design

The repository uses one layered model:

1. canonical provider and request abstractions
2. provider adapters for protocol families
3. transport abstraction for direct or relay deployment
4. feature logic above both layers

Read references/abstraction-model.md first if you want the rationale.

Validation

Run repository validation locally:

python3 scripts/validate_repo.py

This checks:

• SKILL.md frontmatter
• required top-level files and folders
• Swift template type-checking with swiftc

Provider Catalog Refresh

Current curated model lists live in assets/SwiftLLMProviderKit/LLMProviderPresets.swift, with provider-side notes and official source links in references/provider-matrix.md.

Usage Guide

For step-by-step installation and integration guidance, start from references/usage-guide.md.

Repository Layout

.
├── SKILL.md
├── agents/
├── assets/
│   └── SwiftLLMProviderKit/
├── references/
├── scripts/
└── .github/workflows/

License

MIT. See LICENSE.