Getting Started Tutorial

This tutorial will guide you through your first atloop task.

Prerequisites

  • Python 3.8 or higher

  • atloop installed (see Installation)

  • A sandbox backend (or use local_test mode)

  • LLM API credentials configured

Step 1: Configure atloop

Create configuration file at ~/.atloop/config/atloop.yaml:

ai:
  completion:
    model: "deepseek-chat"
    api_base: "https://api.deepseek.com"
    api_key: "${DEEPSEEK_API_KEY}"

sandbox:
  base_url: "http://127.0.0.1:8080"
  local_test: true  # Use local test mode for development

Step 2: Create a Simple Task

Create a Python script:

from atloop.api import TaskRunner
from atloop.config.models import TaskSpec, Budget

task = TaskSpec(
    task_id="my-first-task",
    goal="Add a hello_world() function to main.py",
    workspace_root="./test-project",
    task_type="feature",
    budget=Budget(max_llm_calls=10, max_tool_calls=50),
)

runner = TaskRunner()
result = runner.execute(task)
print(result)

Step 3: Review Results

Check the result:

if result["status"] == "success":
    print(f"✓ Task completed in {result['step']} steps")
    # Review the diff
    print(result.get('diff', ''))
else:
    print(f"❌ Task failed: {result.get('reason')}")

Next Steps