Initializing Your Workspace: A Technical Onboarding Guide

Initializing Your Workspace: A Technical Onboarding Guide

Welcome to the platform. This guide assumes you understand the basics of software development and are ready to integrate AI-assisted workflows into your daily stack.

We will cover the initialization process, environment configuration, and your first agentic execution.

Phase 1: Environment Initialization (Minutes 0-2)

Autobahn is cloud-native but local-aware. When you initialize your workspace, you are spinning up a containerized environment that mirrors a standard Linux development machine.

1. 1Authentication: We recommend GitHub OAuth. This grants our agents read/write access to your repositories, enabling the "Full Context" mode where the AI understands your entire dependency tree.
2. 2Project Import: You have three entry vectors:

Phase 2: The Interface Architecture (Minutes 3-5)

The IDE is composed of three primary subsystems:

• The Editor (Monaco): This is the standard VS Code-like experience you are used to. It supports LSP (Language Server Protocol) for TypeScript, Python, and Go.
• The Terminal: A real, persistent shell. You can run npm install, docker build, or any standard Linux command.
• The Agent Interface: This is not just a chatbot. It is a command center for the Orchestrator.

[!TIP] Pro Tip: You can drag the Agent Interface to a second monitor or toggle it with Cmd+J to keep your focus on the code.

Phase 3: Executing Your First Agentic Workflow (Minutes 6-10)

Let's bypass the "Hello World" and do something useful. We will implement a robust utility function.

The Prompt: > "Create a rate-limiter.ts utility using a sliding window algorithm. It should support Redis as a backend but fallback to in-memory if Redis is not available."

The Execution Flow: 1. Routing: The Orchestrator identifies this as a "Logic Generation" task and routes it to Claude 3.5 Sonnet for its superior reasoning capabilities. 2. Generation: The model generates the class structure, interfaces, and implementation. 3. Self-Correction: If the generated code references a missing Redis client, the agent will either prompt you to install it or scaffold a mock interface.

The Result: You don't just get a code snippet; you get a file created in your file system, ready to be imported.

Optimizing for Cost

Autobahn uses a tiered model routing system to optimize your spend:

• Tier 1 (Local/Flash): For simple refactors ("rename variable", "add comments"). Cost: Near Zero.
• Tier 2 (High-Performance): For complex logic and architecture. Cost: Standard Market Rates.

The cost indicator in the UI shows you exactly which tier is active. Learn to "prompt down" for simple tasks to keep your burn rate low.