Agent Can Optimize Your Deployment Environment
addedYou can now ask the Hyphen Agent to optimize a project environment based on real workload behavior.
When prompted, the Agent creates an optimizer task, analyzes CPU usage, memory usage, and traffic patterns, and generates ranked configuration recommendations to better align infrastructure with demand.
What’s new
- The notification badge now displays the Agent tasks that need approval
- Prompt the Agent to optimize a project environment
- Automatic investigation of CPU, memory, and requests per second
- Ranked recommendations for environment configuration improvements
- Suggested updates to min and max instance counts
- Suggested updates to traffic region placement
- Support for provider- and region-specific scaling adjustments
- Deployment runs triggered after approved changes are applied
How it works
The Agent creates an optimizer task and evaluates the selected project environment.
It then produces a sequence of recommended configuration updates.
Each recommendation:
- includes the proposed change
- identifies region and provider scope
- is labeled with a risk level
- may require approval before execution
Depending on the severity and confidence of a recommendation, the Agent may apply changes automatically or request approval before proceeding.
Why this matters
Right-sizing infrastructure is usually manual and reactive. The Optimizer skill evaluates real usage signals and proposes targeted improvements so your environment better matches workload demand.
- Reduce cloud spend from unused capacity
- Improve responsiveness during traffic spikes
- Strengthen availability configuration
- Adapt infrastructure to real regional demand patterns
Things to know
- Optimization runs as an Agent task with investigation and recommendation phases
- The Agent ranks the severity of each recommendation to determine whether approval is required
- High risk changes require approval before execution
- Recommendations are scoped to the selected project environment
- Future releases will allow approval behavior to be controlled by policies