You can view the version history of specialized agents in Optimizely Opal, letting you view any changes and debug issues that arise.
Access your agents
- Go to home.optimizely.com
- Select your organization.
-
Select Opal.
-
Click Agents.
-
Select the Your Agents tab.
View specialized agent history
You can view a specialized agent's history, which includes previous versions, their modification dates and times, and the individuals responsible for those updates.
To view a specialized agent's history, complete the following:
- Follow the steps in the Access your agents section.
- Click the specialized agent name or click More (...) > Edit Agent.
-
Click View version history.
View previous version of a specialized agent
You can view a previous version of a specialized agent using the Version History. To do so, complete the steps in the View version history section and then click on a previous version of the specialized agent.
Or click More (...) > View this version on the previous version you want view.
Restore a specialized agent to a previous version
Version control for specialized agents acts as a safety, letting you recover from unforeseen problems, and supports iterative development and deployment with reduced-risk development and deployment. Restoring a version of a specialized agent replaces your current agent configuration and creates a new version.
You might need to restore an agent to a previous version for the following reasons:
- Introduction of bugs or errors – New versions, despite rigorous testing, can sometimes introduce unexpected bugs or break existing functionalities. Restoring to a stable previous version lets you quickly mitigate these issues and minimize impact on operations while a fix is developed.
- Performance degradation - A new agent version might inadvertently lead to slower response times, increased resource consumption, or other performance bottlenecks. Rolling back can restore optimal performance.
- Undesired behavior or output – The agent might start generating responses or taking actions that do not align with its intended purpose or brand guidelines. Reverting to a known good version ensures consistent and appropriate behavior.
- Security vulnerabilities – A new update could unintentionally introduce security flaws or expose sensitive data. Restoring to a previous, secure version protects your systems and information.
- Accidental configuration changes or corruption – Human error can sometimes lead to incorrect configurations or even corruption of the agent's current state. Restoring from a backup or previous version can recover the agent.
- Failed experimentation or A/B testing – If you experiment with new agent capabilities or parameters and the results are negative or inconclusive, restoring to the baseline version lets you discard the failed experiment without permanent impact.
- Feature removal or regression – A new version might remove a feature essential for certain workflows, or a previously working feature might stop functioning correctly (regression).
- Auditing and reproducibility – For auditing purposes or to reproduce a specific historical behavior, deploy an exact past version of the agent.
- Negative user feedback – If users report a significant drop in satisfaction or usability after an update, reverting to a previous version can address their concerns while you re-evaluate the changes.
If you want to revert your specialized agent to a previous version, you can do so in the following ways:
- Follow the steps in the View version history section.
- Click on the version you want to restore.
-
Click Restore.
- Click Restore Version on the confirmation page.
Or
- Follow the steps in the View version history section.
- Click More (...) on the version you want to revert to.
-
Click Restore this version.
- Click Restore Version on the confirmation page.
Create a specialized agent based off a previous version
Creating an agent from a previous version lets you leverage the stability of a proven design while gaining the flexibility to innovate, test, and adapt without compromising your existing operations. You might want to create a specialized agent based on a previous version for the following key reasons:
- Isolated feature development – You can develop and test new features or significant changes in a separate, new specialized agent without impacting the live, stable version. This minimizes risk and lets you thoroughly validate the new agent before potential deployment.
- A/B Testing and experimentation – You can launch a new agent version with specific changes (for instance, a different prompt, new tool integrations, or updated logic) alongside the existing production agent. This enables A/B testing to measure the impact of your changes on key metrics without disrupting the primary user experience.
- Branching for customization – You can use a previous stable version as a foundation to create specialized agents tailored for different use cases, departments, or customer segments. This lets you target enhancements while maintaining a consistent core functionality.
- Parallel development streams – Multiple teams or developers can work concurrently on different improvements or features. Each team can start from a shared, previous version, fostering independent development and reducing merge conflicts.
- Safe iteration and learning – If a recent update introduced issues that are not critical enough for an immediate rollback, you can use a previous stable version as a starting point to re-implement the desired changes more carefully, incorporating lessons learned from the problematic version.
To create a specialized agent based off a previous version, complete the following:
- Follow the steps in the View version history section.
- Click More (...) on the version you create a new agent based off of.
-
Click Duplicate as a new agent.
- Update the agent, including the Id, to make it unique.
- Click Save.
Article is closed for comments.