I’m building a SaaS application using an open-source project licensed under EPL 2.0, and I want to include my own AI functionality. For now, I plan to only offer the product as a service (not distribute the software), but in the future, I might want to create a separate desktop AI app. Given these plans, should I keep my AI code separate from the EPL 2.0 codebase from the start, or can I combine them for now and separate them later if needed? What are the pros and cons of each approach, especially regarding licensing, maintainability, and future flexibility? I am a solo developer.