In the myths, Dedalus was the master craftsman that gave men powerful wings. Ariadne was the clever problem-solver that threaded the way out of his labyrinth. Together, they made the impossible navigable.

Ariadne brings composable, structured workflows to Dedalus Labs' agent platform, built off the backs of Effect and Effect-AI.

It aims to provide

• Full type-safety and observability from error accumulation.
• Composability for ease of evaluation and unit testing.

Documentation

Read the full documentation for comprehensive guides:

Guide	Description
Getting Started	Installation and quick start
Language Model	Text generation, structured outputs, streaming
Tools & Toolkits	Define and implement tool calling
Chat	Stateful multi-turn conversations
MCP Servers	Model Context Protocol integration
Execution Planning	Retries, fallbacks, and resilience
Embeddings	Vector embeddings
Configuration	Client and model options
Error Handling	Error types and patterns

Installation

bun install ariadne

Quick Start

import { LanguageModel, Tool, Toolkit } from "@src/ariadne"
import { DedalusClient, DedalusLanguageModel } from "@src/dedalus-labs"
import * as FetchHttpClient from "@effect/platform/FetchHttpClient"
import { Effect, Layer, Schema } from "effect"
import * as Redacted from "effect/Redacted"

// 1. Create the client layer
const Dedalus = DedalusClient.layer({
  apiKey: Redacted.make(process.env.DEDALUS_API_KEY!),
}).pipe(Layer.provide(FetchHttpClient.layer))

// 2. Create a model (supports multi-provider routing)
const Gpt4o = DedalusLanguageModel.model("openai/gpt-4o")

// 3. Define a tool (optional)
const Calculator = Tool.make("Calculator", {
  description: "Perform arithmetic",
  parameters: {
    operation: Schema.Literal("add", "subtract", "multiply", "divide"),
    a: Schema.Number,
    b: Schema.Number,
  },
  success: Schema.Number,
})

const toolkit = Toolkit.make(Calculator)

const toolkitLive = toolkit.toLayer({
  Calculator: ({ operation, a, b }) =>
    Effect.succeed(
      operation === "add" ? a + b :
      operation === "subtract" ? a - b :
      operation === "multiply" ? a * b :
      a / b,
    ),
})

// 4. Generate text with tools
const program = LanguageModel.generateText({
  prompt: "What is 15 multiplied by 7?",
  toolkit,
})

// 5. Run
const result = await program.pipe(
  Effect.provide(toolkitLive),
  Effect.provide(Gpt4o),
  Effect.provide(Dedalus),
  Effect.runPromise,
)

console.log(result.text)
console.log(result.toolResults) // [{ name: "Calculator", result: 105 }]

Development

# Install dependencies
bun install

# Build
bun run build

# Develop with watch mode
bun run dev

# Run tests
bun run test

# Lint and format
bun run lint
bun run format

Effect AI Packages

This repo includes a fork of the Effect AI packages. Upstream changes are monitored and selectively merged as needed.

Contributing

Please see CONTRIBUTING.md for contribution guidelines.

License

MIT