pandaapo commented on PR #4817: URL: https://github.com/apache/eventmesh/pull/4817#issuecomment-2041656698
@Pil0tXia You have correctly drawn the architecture of this PR (with the specified prompts idea first proposed by the author and your prompts template idea), so I believe you understand the core functionality implemented in this PR now. As for the design of the prompt templates and how to send specified prompt, I suggest leaving them to the developer without limiting the developer's thought, as the author creatively thought of multiple data formats can being specified before. Of course the developer can adopt your good design. The developer can even temporarily not develop prompt templates and submit the feature of prompt templates after multiple PRs submitted for this Connector. > this checking process may not save much performance overhead compared to deserialization. Maybe. Maybe the overhead of checking format is less than the overhead of building CloudEvent item by item. Suitable code can even do without explicitly checking format. > By not relying on the JSON format generated by ChatGPT, more stable results can be obtained. I said earlier to make it configurable, with the option to choose how it currently runs like what this PR do, and the option to ask for a specified format to be returned according to a specified prompt. The _stable results_ here only make this Connector feature-poor in practice. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
