Switch: AI assistant

Overview

Switch is an in-app AI assistant designed to streamline the use of the Split product. It offers multilingual support, rapid responses, and knowledge-based assistance by utilizing our public documentation and blogs. Switch makes it easy for all developers to get the help they need, without ever leaving the Split interface. This documentation provides technical details and guidelines for developers to work with Switch effectively.

Getting Started

If Switch is available to your team you will find this icon in the bottom left.

When you click on “Ask Switch” the chat will open on the right side of the screen. You can click on one of the suggested questions to get started or type in your own question.

Example questions

• How do I set up a javascript SDK in my codebase?
• How do I create and manage feature flags in Split?
• How can I integrate Split with my application using the SDK?
• What are the best practices for naming feature flags and events?
• How do I set up an A/B test in Split?
• Can you explain how Split's targeting rules work?
• How can I use segments to target specific groups of users?
• What data types can I track using Split events for measurement?
• How do I interpret the metrics on my experiments' results?
• How can I manage user permissions and access control in Split?
• How do I export data from Split for further analysis?
• What are the implications of feature flag rollout strategies, such as canary releases?
• How do I troubleshoot issues with the Split SDK?
• Can I use Split across multiple environments, and how?
• How does Split handle data privacy and security?
• What are the best practices for cleaning up old flags and technical debt?

Accuracy

Please be aware that Switch is an automated system and is currently in Beta. It may not always provide 100% accurate information. While we strive to provide accurate information, we cannot guarantee the accuracy, completeness, or timeliness of the information provided by the chatbot. You can always reach out to support (link) for additional help

We have experienced that Switch can have hallucinations when asked to return a numerical value. When dealing with numbers, the limitations of LLMs can often be more pronounced due to the precision and context required in numerical information.

Data Security

Switch currently only uses publicly available information. No customer data is involved.

In a future iteration, there will be features that benefit from ingestion of customer data, and users will have to explicitly opt-in if they’d like to use these features.

Upcoming Enhancements

The introduction of Switch is just the beginning, as the forward-looking product roadmap of Split reflects a robust integration of AI into its services by introducing automated features that simplify and enhance the development process, including:

• Automatic Data Interpretation: Assist users in understanding complex metrics and graphs.
• Immediately Actionable Insights: Enabling Switch to not only interpret queries but also perform background tasks like API calls. This would allow a user to be alerted on their phone when a release has an issue, and ask Switch to kill the release immediately, without needing to return to a computer.
• Easy Tech Debt Collection: Identifying and resolving obsolete code with a quick and simple request to Switch. 

Feedback and Support

For support inquiries or to report issues with Switch, please submit a request via our Help Center. Provide detailed descriptions and reproduction steps for issues to facilitate expedited troubleshooting. Your feedback is instrumental for the iterative enhancement of Switch.