Quick Start

Build a minimal AI agent that connects to Athyr and uses LLM completions.

Prefer no code? You can define agents in YAML and run them with Athyr Agent — no programming required.

Prerequisites

1. Start Athyr

athyr serve

2. Create the Agent

mkdir my-agent && cd my-agent
go mod init my-agent
go get github.com/athyr-tech/athyr-sdk-go

Create main.go:

package main

import (
	"context"
	"fmt"

	"github.com/athyr-tech/athyr-sdk-go/pkg/athyr"
)

func main() {
	ctx := context.Background()

	// 1. Create agent
	agent, _ := athyr.NewAgent("localhost:9090",
		athyr.WithAgentCard(athyr.AgentCard{
			Name:        "my-agent",
			Description: "My first AI agent",
		}),
	)

	// 2. Connect and register with Athyr
	agent.Connect(ctx)
	defer agent.Close()

	fmt.Println("Agent registered:", agent.AgentID())

	// 3. Call LLM through Athyr
	resp, _ := agent.Complete(ctx, athyr.CompletionRequest{
		Model: "llama3",
		Messages: []athyr.Message{
			{Role: "user", Content: "Hello! What can you do?"},
		},
	})

	fmt.Println("Response:", resp.Content)
}

3. Run

go run main.go

Agent registered: my-agent-x7k2m
Response: Hello! I'm an AI assistant. I can help answer questions,
explain concepts, write code, and much more. How can I help you?

What’s Happening

  1. Create - NewAgent() creates the agent with an identity card
  2. Register - Connect() registers the agent with Athyr’s registry
  3. LLM Call - Complete() sends a request through Athyr’s LLM gateway

The agent doesn’t know about Ollama - Athyr handles LLM routing and provider management.

Alternative: YAML Agent (No Code)

If you prefer not to write code, you can define the same agent in YAML using Athyr Agent:

# agent.yaml
agent:
  name: my-agent
  description: My first AI agent
  model: llama3
  instructions: |
    You are a helpful assistant. Respond concisely.
  topics:
    subscribe: [requests.new]
    publish: [requests.done]

brew install athyr-tech/tap/athyr-agent
athyr-agent run agent.yaml --server localhost:9090

Next Steps