Production Prompt System Architecture

# Production Prompt System Architecture

You are a software architect specializing in AI systems. Design a production prompt management system for [APPLICATION TYPE].

## Why Prompts Need Engineering

Explain the risks of treating prompts as static strings: silent regressions when models update, no visibility into what changed, inability to roll back a bad change, no way to know which prompt version is running in production.

## System Components

Define each component of the production prompt system:
- Prompt registry: versioned storage with unique IDs and deployment history
- Testing pipeline: unit tests for prompt behavior, regression suite on every change
- Staging environment: test against production-equivalent model before deploying
- Rollback mechanism: instant revert to the previous version with zero downtime
- A/B testing layer: route a percentage of traffic to the new prompt version
- Monitoring and alerting: track output quality metrics and alert on degradation

## Prompt as Code

Define the format for storing a prompt as code:
- Version identifier and changelog
- Variables and their types
- The prompt template itself
- Test cases (input + expected output)
- Evaluation metrics

## Architecture Diagram

Describe the data flow from prompt change request through testing, staging, deployment, and monitoring for [APPLICATION TYPE].