Disclaimer: This is an experiment for SDK v0.20. LLMs are subject to hallucination and can make mistakes. Please save any existing work to source control and inspect the LLM’s changes carefully. If you have issues, please revert to the upgrade documentation.
Pre-requisite: You have successfully updated your computer to Kepler SDK v0.20 per the update instructions.
Using an AI Agent to accelerate your Kepler v0.20 Upgrade
For an easier upgrade of your project to the v0.20 SDK, we recommend using an AI agent that integrates directly into your IDE through an extension. These extensions typically provide an interactive prompt window, allowing you to clearly see and review proposed changes within your familiar development environment. To assist you, we have provided a sample prompt that has the specific instructions for the v0.20 upgrade: ksdk-upgrade-prompt.txt (10.3 KB)
Agent and Model Compatibility
Our testing has shown that IDE-integrated autonomous coding agents such as Cline using the most capable models such as Claude 4 Sonnet provide the most seamless experience. However, you can use any AI agent and model that effectively handles complete prompts and code modifications. The key requirement is that the agent can process your codebase (with your permission of course) and communicate changes in a clear, interactive manner.
Working with Plan/Act Systems
Many AI agents operate using a Plan/Act paradigm. When working with these systems, we recommend first engaging the agent in “Plan” mode to review the proposed changes. Once you’ve validated the upgrade strategy, you can switch to “Act” mode to implement the changes. This two-step approach provides better control and visibility over the upgrade process.
Remember that while these are our recommended approaches based on testing, the most important factor is choosing a solution that integrates well with your existing development workflow and provides clear visibility into the upgrade process.
Sample workflow (using Cline)
-
Put coding agent in “Plan” mode and prompt with the sample prompt (i.e you can copy/paste the text into the prompt).
-
Review the generated upgrade plan, and note the major code changes. Based on the package, this would typically include:
package.jsonreact-native.config.jsconanfile.py- For project with native bits:
CMakeLists.txt
-
When you are ready to make code changes, put Cline in “Act” and follow the prompts to complete the changes
-
Now that you’ve run the upgrade tool, it’s time to ensure everything works as expected. Start by building your package using your standard workflow command (such as
kepler build). This first build will help identify any immediate issues that need your attention. Test your newly built artifact on either your simulator or physical device. This hands-on validation will help you catch any behavioral changes that might have been introduced during the upgrade.
We strongly recommend reviewing the changes before deploying to production. Take a moment to look through the upgraded code, paying particular attention to any sections that were automatically modified. Remember, a successful upgrade is about more than just running a tool - it’s about ensuring your application continues to work as expected on a device. If you run into any questions along the way, consult our upgrade documentation or reach out for any questions on the community forum!
Did you like this? “Like” this post or let us know on the community forums!