Protocol-based AI adapter foundation for Elixir - unified abstractions for gemini_ex, claude_agent_sdk, codex_sdk with automatic fallback, capability detection, and telemetry

Altar.AI

Unified AI adapter foundation for Elixir - Protocol-based abstractions for multiple AI providers

Features

• Protocol-Based Architecture - Uses protocols instead of behaviours for maximum flexibility
• Runtime Capability Detection - Introspect what each adapter supports at runtime
• Composite Adapters - Automatic fallback chains across multiple providers
• Framework Agnostic - No dependencies on FlowStone, Synapse, or other frameworks
• Unified Telemetry - Standard telemetry events for monitoring and debugging
• Comprehensive Testing - Mock adapters and test utilities included

Supported Providers

• Gemini - Google Gemini AI (via gemini_ex)
• Gemini - Anthropic Claude (via claude_agent_sdk)
• Codex - OpenAI models (via codex_sdk)
• OpenAI - OpenAI chat + embeddings (via openai_ex)
• Fallback - Heuristic fallback (no external API required)
• Mock - Configurable mock for testing

All SDK dependencies are optional - Altar.AI works with whatever you have installed.

Installation

Add altar_ai to your list of dependencies in mix.exs:

def deps do
  [
    {:altar_ai, "~> 0.1.0"},
    # Optional: Add the AI SDKs you want to use
    # {:gemini_ex, "~> 0.1.0"},
    # {:claude_agent_sdk, "~> 0.1.0"},
    # {:codex_sdk, "~> 0.1.0"},
    # {:openai_ex, "~> 0.9.18"}
  ]
end

Quick Start

Basic Usage

# Create an adapter
adapter = Altar.AI.Adapters.Gemini.new(api_key: "your-api-key")

# Generate text
{:ok, response} = Altar.AI.generate(adapter, "Explain Elixir protoc