Skip to content
This repository was archived by the owner on Jul 22, 2025. It is now read-only.

Conversation

@SamSaffron
Copy link
Member

  • If JSON is broken for structured output then lean on a more forgiving parser
  • Gemini 2.5 flash does not support temp, support opting out
  • Evals for assistant were broken, fix interface
  • Add some missing LLMs
  • Translator was not mapped correctly to the feature - fix that
  • Don't mix XML in prompt for translator

- If JSON is broken for structured output then lean on a more forgiving parser
- Gemini 2.5 flash does not support temp, support opting out
- Evals for assistant were broken, fix interface
- Add some missing LLMs
- Translator was not mapped correctly to the feature - fix that
- Don't mix XML in prompt for translator
@SamSaffron
Copy link
Member Author

SamSaffron commented Jul 3, 2025 via email

if bad_json
helper_response = helper_response.join if helper_response.is_a?(Array)
helper_response =
DiscourseAi::Utils::BestEffortJsonParser.extract_key(
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is great! I suggest we move it inside #read_buffered_property.

@partial_json_tracker << raw
end

def broken?
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No need for this if we move the best effort parser inside this object.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

actually we still need it cause on finish when streaming we need to double check.

@SamSaffron SamSaffron merged commit ab5edae into main Jul 4, 2025
6 checks passed
@SamSaffron SamSaffron deleted the evals2 branch July 4, 2025 04:47
keegangeorge added a commit that referenced this pull request Jul 7, 2025
This update fixes a regression from #1484, which caused AI helper title suggestions to begin suggesting numerous non-unique titles because it was looping through structured responses incorrectly.
keegangeorge added a commit that referenced this pull request Jul 8, 2025
This update fixes a regression from #1484, which caused AI helper title suggestions to begin suggesting numerous non-unique titles because it was looping through structured responses incorrectly.
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Development

Successfully merging this pull request may close these issues.

4 participants