Skip to content

ai-ops

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

ai-ops

README.md

AI-OPS

What AI-OPS Is

AI-OPS is the standardized operating model for AI-assisted engineering across the DevFlow monorepo. It defines how planning, implementation, review, and acceptance interact so that work remains consistent, auditable, and safe for production SaaS products.

Why It Exists

  • Consistency: Multiple products (Next.js apps, RPA, APIs, packages) share one way of working with AI tools.
  • Risk control: Clear separation between ideation (ChatGPT) and execution (Cursor) reduces blind refactors and architectural drift.
  • Quality bar: Linked documents (workflow, acceptance criteria, definition of done) make “done” objective, not subjective.
  • Onboarding: New engineers and AI sessions start from the same playbook.

How It Integrates with DevFlow

Layer Role
Repository Source of truth: Turborepo, PNPM, shared packages, app-specific boundaries.
AI-OPS Process layer: when to plan, how to prompt, how to validate, when a task is complete.
Product teams Apply AI-OPS per feature; extend only where product rules differ (document deltas in PRs).

AI-OPS does not replace product specs, security review, or release management. It structures how AI tools participate in delivery.

High-Level Flow

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  ChatGPT    │     │   Cursor    │     │  ChatGPT    │
│  Planning   │────▶│  Execution  │────▶│ Validation  │
└─────────────┘     └─────────────┘     └─────────────┘
       │                    │                    │
       │                    │                    │
       ▼                    ▼                    ▼
  Intent, scope,      Code & tests        Review vs
  constraints,        in-repo             acceptance
  prompts ready                             criteria
  1. ChatGPT (planning): Clarify requirements, surface risks, produce structured prompts and acceptance notes.
  2. Cursor (execution): Implement against the repo; follow CURSOR_RULES.md and product conventions.
  3. ChatGPT (validation): Compare implementation intent to ACCEPTANCE_CRITERIA.md and DEFINITION_OF_DONE.md; flag gaps before merge.

Iteration loops between Cursor and validation until the task meets the definition of done.

Document Map

File Purpose
AI_WORKFLOW.md Official step-by-step workflow and anti-patterns.
CURSOR_RULES.md Execution rules for Cursor and code quality expectations.
PROMPT_TEMPLATE.md Reusable prompts for feature, bugfix, refactor, infra.
ACCEPTANCE_CRITERIA.md Standard acceptance model by domain.
DEFINITION_OF_DONE.md Objective completion checklist.

AI-OPS is maintained as part of DevFlow engineering standards. Propose changes via PR with rationale.