Skip to content
Courses/Aulafy foundations/Local-first environment map

Local-first environment map

Before installing models, agents, or RAG, you need to understand the pieces that make an AI setup reproducible and private.

Why start here

Aulafy is designed for building with AI without depending on one account, one cloud, or one tool. That requires a small foundation: terminal, Python, Git, isolated environments, Docker, and judgment about what runs locally and what can call an external API.

You do not need to be a systems administrator. You do need to repeat a project tomorrow, explain how to run it, and review what changed when AI edited files.

The stack pieces

  • Terminal: where you start models, servers, scripts, and tools such as Codex, Claude Code, Ollama, or Docker.
  • Modern Python: the glue for scripts, RAG, evals, automations, and model tests.
  • Git: the history of code, prompts, configuration, and technical decisions.
  • Docker: the practical way to run services such as Qdrant, n8n, Open WebUI, or databases without polluting your machine.
  • Local models: useful for privacy, predictable cost, repetitive tasks, and reproducible learning.

Local vs cloud rule of thumb

Use local when data is sensitive, repetitive, cheap to process on your machine, or when you want to learn how the system works. Use cloud when you need the highest quality, huge context, stronger reasoning, or an integration you do not have locally yet.

TerminalCode
decision:
  sensitive_data: local
  fast_prototype: cloud_or_local
  repetitive_cost: local
  highest_quality: frontier
  technical_learning: local
  hybrid_production: gateway_with_routing

Deliverable

By the end of this lesson you should be able to draw your stack: which tool edits code, which model answers, where data lives, how the project is versioned, and how it starts again.