Uploaded image for project: 'Red Hat Enterprise Linux AI'
  1. Red Hat Enterprise Linux AI
  2. RHELAI-4005

[UX] Word Count Improvements - surface warnings, not errors

XMLWordPrintable

      Goal/Problem Statement 
      Provide guidance (warnings) to users when they are selecting context and generating QnA's that if they exceed our word count guidance, the quality of the data generated can go down. Instead of making it so they can't go over, provide a reasoning to why they should not, but still allow them to proceed.

      Background
      The reason we have limits is to make sure the prompts to the teacher model don't get too big, if they get too big then the quality of the data generated goes down. As we swap out teacher models, we will be able to have bigger contexts and question and answer sizes too.

      We currently have errors implemented for both context and qna word count. We need to remove any hard limits on word count.

      Feedback 
      tasmith@redhat.com: We are using the QA UI this week and we need the context selection part of the process to not give hard limits to word count or q&a - limits need to be warnings only

      Acceptance Criteria

      • Determine the best method to alert a user of recommendations related to word count and remove barrier of requiring a user to meet these thresholds to submit
      • Pass off to Dev (Mis to work with Jeff)
      • Share with FnF team (have Jehlum, Taylor and William ack)

      Deliverables
      Update word count overage to a warning rather than error #743
      Add alert on Seed data section #753

              misjohns@redhat.com Missy Johnson
              beaumorley Beau Morley
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated:
                Resolved: