AWS AppConfig adds versioned AI prompts and model settings

AWS AppConfig added two configuration types on November 19 for managing AI model settings and prompts independently of application-code deployments.
The AWS announcement says model-invocation configurations store identifiers, inference parameters and messages, including system prompts and example conversations. Provider-specific settings use custom parameters so the schema can work across AI providers.
Prompt configurations are separately versioned artifacts. Teams can change a prompt without changing its model configuration and reuse it across multiple applications.
Both types support gradual deployment and automatic rollback. Developers can release a change to progressively larger shares of traffic while monitoring operational metrics, customer experience and token usage. If a configured observability alarm fires during rollout, AppConfig can restore the previous configuration.
AWS says the new configuration types are available in every Region where AppConfig operates, at no additional cost. This is not a statement that the underlying application or model inference is free.
The addition brings prompts and model-call settings into an existing deployment-control system, giving teams a way to test changes without treating every adjustment as a full application release.
Sources and citations
Revision history
Published .