Imported from ThinkWithOps/ai-devops-systems-lab (AGENTS.md). Install upstream with npx skills add ThinkWithOps/ai-devops-systems-lab. Copyright stays with the author.

AGENTS.md

This file provides guidance to Codex (Codex.ai/code) when working with code in this repository.

What This Repository Is

This is a meta-repository and guidance system — not a runnable application. It contains four Codex instruction systems for generating a YouTube series of 30 portfolio-grade AI + DevOps projects. There are no build, test, or lint commands.

Four Instruction Systems

Project Generation Workflow

Step 1 — Use .Codex/ system prompt in Codex terminal to generate the main AI platform.

Step 2 — Use .Codex-demoapps/ system prompt in Codex terminal to generate a realistic demo app the platform will operate on.

Step 3 (optional) — Use .Codex-aws/ to extend with AWS services.

Step 4 — Use .Codex-nontechnical/ in Codex Chat/Web (not terminal) to generate YouTube scripts, thumbnails, LinkedIn posts, Dev.to articles.

Minimum Prompt Format (Step 1)

Use the local .Codex system for this project.

Project Name: [NAME]
Viral YouTube Title: [TITLE]
Tech Stack: [STACK]

Non-Negotiable Standards (from .Codex/project_rules.md)

Every generated project must:

• Have a visible web UI (Next.js + React + Tailwind) — no CLI-only outputs
• Have a FastAPI backend
• Include AI orchestration (LangChain / LangGraph / AutoGen)
• Include a vector database or retrieval layer (Chroma / Weaviate / Pinecone)
• Include Docker deployment (docker-compose.yml)
• Default to free/local-first LLM tooling (Ollama over OpenAI)

Standard Tech Stack

• Frontend: Next.js, React, Tailwind CSS
• Backend: FastAPI
• AI: LangChain, LangGraph, AutoGen
• Vector DB: Chroma, Weaviate, Pinecone
• LLMs: Ollama (default), OpenAI, HuggingFace
• Infra: Docker, Kubernetes (when relevant), Prometheus + Grafana (observability)

Standard Generated Project Structure

project-name/
├── frontend/          # Next.js dashboard UI
├── backend/           # FastAPI + services
├── agents/            # AI orchestration
├── vectorstore/       # Vector DB setup
├── infra/             # Docker, Kubernetes
├── docs/              # Architecture docs
├── scripts/           # Setup and utility scripts
├── docker-compose.yml
└── README.md

UI Layout Standard

Every project dashboard uses a consistent SaaS-style layout (see .Codex/ui_layout_pattern.md):

• Header: project name, status, environment, primary action button
• Sidebar: Dashboard, Agents, Logs, Architecture, Metrics, Results, Settings
• Body: 4–6 KPI cards + main visual panel + AI insights panel + live log strip

The 30-Project Sequence

Defined in README.md. Project 1 is AI DevOps Copilot (ChatGPT for DevOps Engineers). All 30 project prompts are pre-written in .Codex/example_prompt.md.

Key Reference Files

• WORKFLOW.md — step-by-step operating guide with the exact sequence for Project 01
• .Codex/system_prompt.md — core generation philosophy and role definition
• .Codex/project_rules.md — full list of non-negotiable requirements
• .Codex/generation_rules.md — what must be auto-generated from minimal input
• .Codex/portfolio_output_format.md — required section order for all generated projects
• .Codex/example_prompt.md — ready-to-run prompts for all 30 projects