Improvement/openai chat

[uwwint]

This changes the prompt to cahtGPT to generate a better support by

1. Structuring the response into the sections: "Cause of the error" and "How to resolve the error using Galaxy"
2. Removes the username from the response, making the the response feel inlined into the web page, also it gets rid of responses that look like a letter (starting with "Hi @Username" and ending in "Kind Regards, ..."
3. Instructing the agent that it is only providing one shot support, removing sentences like "Feel free to reach out if the error persists" For which there is no UI component.

How to test the changes?

(Select all options that apply)

• [ x ] I've included appropriate automated tests.
• This is a refactoring of components with existing test coverage.
• [ x ] Instructions for manual testing are as follows:
  1. cause a job to fail
  2. use the wizzard to analyse the error

License

• I agree to license these and all my past contributions to the core galaxy codebase under the MIT license.

[uwwint]

@nuwang

[dannon]

Thanks for the tweaks; +1 to dropping the user-addressing part; it did feel a bit too kitschy. I followed up with a more general prompt that refines this further while still addressing the point of one-shot support, suggesting the user go ahead and file that bug report if they still need help.

This also reminded me that I want to follow up and swap this json to yaml for easier customization by end users, especially considering the additional interactions we're working on that'll be coming with the next release.