xintongsong commented on issue #335: URL: https://github.com/apache/flink-agents/issues/335#issuecomment-3988029356
Good questions. Here are my two cents: > Language precedence: When should the base implementation live in Java vs Python? (intentional decision, rather than which ever was implemented first) I think it depends on the sdk supports of the integrating system. If sdk is only provided for one language, or provided in both languages but one is more mature than the other, we should choose the language with better supports. Otherwise, if sdk is provided for both language with equal maturity, I'd be fine with either language. > Native implementation: When should we implement natively in both languages vs reuse one through the bridge? Cross-language resource access currently has a limitation of not supporting async execution, because of a bug in the upstream project pemja, which we are trying to fix next. Once this is resolved, I don't see a necessity in providing native implementation for both languages. Another point worth considering is the performance issue. The overhead of cross-language interactions may affects the performance. However, since currently the performance bottleneck is apparently in calling the LLMs, it's probably too early to worry about the cross-language overhead. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
