Mojentic

Copy Markdown View Source

Hex.pm License: MIT Elixir

An LLM integration framework for Elixir with full feature parity across Python, Rust, and TypeScript implementations.

Mojentic provides a clean abstraction over multiple LLM providers (OpenAI, Ollama) with tool support, structured output generation, streaming, and a complete event-driven agent system.

🚀 Features

  • 🔌 Multiple Providers: OpenAI and Ollama gateways
  • 🛠️ Tool Support: Extensible tool system with automatic recursive execution
  • 📊 Structured Output: Type-safe response parsing with JSON schemas
  • 🌊 Streaming: Real-time streaming with full tool calling support
  • 🔍 Tracer System: Complete observability for debugging and monitoring
  • 🤖 Agent System: Event-driven multi-agent coordination with ReAct pattern
  • 🏗️ OTP Design: GenServer-based components ready for supervision trees
  • 📦 24 Examples: Comprehensive examples demonstrating all features

📦 Installation

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

def deps do
  [
    {:mojentic, "~> 1.0.0"}
  ]
end

Quick Start

Simple Text Generation

alias Mojentic.LLM.{Broker, Message}
alias Mojentic.LLM.Gateways.Ollama

broker = Broker.new("qwen3:32b", Ollama)
messages = [Message.user("What is Elixir?")]
{:ok, response} = Broker.generate(broker, messages)
IO.puts(response)

Structured Output

schema = %{
  type: "object",
  properties: %{
    sentiment: %{type: "string"},
    confidence: %{type: "number"}
  },
  required: ["sentiment", "confidence"]
}

messages = [Message.user("I love this product!")]
{:ok, result} = Broker.generate_object(broker, messages, schema)
IO.inspect(result)
# => %{"sentiment" => "positive", "confidence" => 0.95}

Tool Usage

defmodule WeatherTool do
  @behaviour Mojentic.LLM.Tools.Tool

  @impl true
  def run(args) do
    location = Map.get(args, "location", "unknown")
    {:ok, %{location: location, temperature: 22, condition: "sunny"}}
  end

  @impl true
  def descriptor do
    %{
      type: "function",
      function: %{
        name: "get_weather",
        description: "Get current weather for a location",
        parameters: %{
          type: "object",
          properties: %{
            location: %{type: "string", description: "City name"}
          },
          required: ["location"]
        }
      }
    }
  end
end

tools = [WeatherTool]
messages = [Message.user("What's the weather in SF?")]
{:ok, response} = Broker.generate(broker, messages, tools)
IO.puts(response)

Examples

See the examples/ directory for complete runnable examples:

# Simple LLM text generation
mix run examples/simple_llm.exs

# Structured output with JSON schema
mix run examples/structured_output.exs

# Tool usage with automatic tool calling
mix run examples/tool_usage.exs

Configuration

Environment Variables

  • OLLAMA_HOST - Ollama server URL (default: http://localhost:11434)

Architecture

Mojentic is structured in three layers:

Layer 1: LLM Integration

  • Mojentic.LLM.Broker - Main interface for LLM interactions
  • Mojentic.LLM.Gateway - Behaviour for LLM provider implementations
  • Gateway implementations: Ollama, OpenAI
  • Mojentic.LLM.ChatSession - Conversational session management
  • Mojentic.LLM.TokenizerGateway - Token counting
  • Mojentic.LLM.EmbeddingsGateway - Vector embeddings
  • Comprehensive tool system with 10+ built-in tools

Layer 2: Tracer System

  • Mojentic.Tracer.System - Event recording GenServer
  • Mojentic.Tracer.EventStore - Event persistence and querying
  • Correlation ID tracking across requests
  • LLM call, response, and tool events

Layer 3: Agent System

📚 Documentation

Generate documentation locally:

mix docs
open doc/index.html

🧪 Development

# Install dependencies
mix deps.get

# Compile
mix compile

# Run tests
mix test

# Format code
mix format

# Run code quality checks
mix credo --strict

# Security audit
mix deps.audit

📄 License

MIT License - see LICENSE

Credits

Mojentic is a Mojility product by Stacey Vetzal.