Introduction

PromptOps is a Prompt Runtime platform for AI applications. It decouples prompts, agents and workflows from code, so AI applications can hot-reload, version, publish, gray-release and observe their prompts at runtime.

Core idea:

Make prompts a properly engineered resource — Build → Version → Deploy → Runtime → Observe.

Why it exists

Problem	Description
Prompts hard-coded	Changing a prompt requires a redeploy
No version control	Hard to roll back when something breaks
No hot reload	Updating a prompt means restarting the AI service
Messy team collaboration	No ownership or permissions
No gray release	New prompts go fully live at once — risky
No observability	No idea how often prompts run, or how well

Capabilities

Module	Description
Prompt Runtime	CRUD, search, categories, multi-environment, version publish & rollback
Hot Reload	Real-time prompt-change push over WebSocket
SDK	Node / Python / Java — Python & Java zero-dependency, Node depends only on ws
Playground	Fill variables, call models, view results; version diff
Agent	A config-driven prompt + provider + model
Workflow	A render → model → transform step-orchestration engine
Observability	Audit logs, run logs, token & latency stats, live connected clients
RBAC	admin / editor / viewer roles
Gray Release	Weighted A/B version traffic split
Workspace	Resources grouped by team workspace

Architecture

• Backend: Go + Gin + GORM + SQLite, with WebSocket hot-reload.
• Frontend: Vue 3 + Vite + Naive UI + Pinia.
• SDKs: Node (built-in fetch, ws for WebSocket), Python (standard library), Java (java.net.http).
• Deployment: single binary / Docker.

Next: Quickstart · Deployment.