instructure/canvas-lms · error · CedarAi::Errors::GraderError

friendly_cedar_error_for(e)

Error message

friendly_cedar_error_for(e)

What it means

GradeService#call wraps Cedar (AI grading) API calls. CedarClient validation-style errors (ValidationError, CedarLimitReachedError, UnsupportedLanguageError, ContentTooLongError) have static user-safe messages, so they are re-raised as CedarAi::Errors::GraderError with a message produced by friendly_cedar_error_for(e). The bracketed message is just the raise expression; the actual message depends on the Cedar error.

Solutions

1. Read the raised GraderError message to identify which Cedar condition fired (limit vs validation vs language vs length).
2. Shorten the submission/rubric content (trim via normalize_rubric_for_prompt, truncate essay text) before sending to Cedar.
3. Wait or request a limit increase if CedarLimitReachedError; back off and retry the grading job later.
4. Ensure the submission language is supported or skip Cedar grading for unsupported languages.

Example fix

// before
service.call(...) # content too long -> GraderError
// after
prompt = GradeService.normalize_rubric_for_prompt(rubric_data)
service.call(...) if prompt.to_s.length <= MAX_CEDAR_CONTENT_LENGTH

begin
  GradeService.call(...)
rescue CedarAi::Errors::GraderError => e
  render json: { error: e.message }, status: :unprocessable_entity
end

When it happens

Trigger: Cedar API rejects the grading request: prompt/rubric content too long, request exceeds a Cedar usage limit, unsupported language detected, or input fails Cedar client-side validation during grading submission.

Common situations: Very large submissions or rubrics hitting Cedar content limits; account-level Cedar rate/usage limits exhausted; grading essays in a language Cedar doesn't support; malformed rubric data normalized incorrectly before send.

Related errors

• An unexpected error occurred while grading.
• A new_id, '# ', referenced an existing # and the # with #…
• A new_integration_id, '#
• A student referenced a non-existent user #
• A user did not pass validation

AI-assisted analysis of instructure/canvas-lms@1c9f0bb801 (2026-09-15). Data as JSON: /api/errors/ed17a967f66ac949. Report an issue: GitHub.

Appendix: source

Thrown at app/services/grade_service.rb:58

    begin
      grading_results = CedarClient.grade_essay(
        description: @assignment,
        essay: @essay,
        rubric: cedar_rubric,
        feature_slug: "grading-assistance",
        root_account_uuid: @root_account_uuid,
        current_user: @current_user
      )

      map_grade_essay_results_to_canvas(grading_results, @rubric)
    # These subclasses have static, user-safe messages and can be shown directly.
    # If new CedarClientError subclasses are added with user-safe messages, add them here.
    rescue InstructureMiscPlugin::Extensions::CedarClient::ValidationError,
           InstructureMiscPlugin::Extensions::CedarClient::CedarLimitReachedError,
           InstructureMiscPlugin::Extensions::CedarClient::UnsupportedLanguageError,
           InstructureMiscPlugin::Extensions::CedarClient::ContentTooLongError => e
      raise CedarAi::Errors::GraderError, friendly_cedar_error_for(e)
    rescue InstructureMiscPlugin::Extensions::CedarClient::CedarClientError => e
      Rails.logger.error("[GradeService] Cedar API error: #{e.message}")
      raise CedarAi::Errors::GraderError, friendly_cedar_error_for(e)
    rescue => e
      Rails.logger.error("[GradeService] Unexpected error: #{e.class}: #{e.message}")
      raise CedarAi::Errors::GraderError, I18n.t("An unexpected error occurred while grading.")
    end
  end

  def self.normalize_rubric_for_prompt(rubric_data)
    rubric_data.each_with_object({}) do |criterion, acc|
      key = criterion[:description]
      acc[key] = {
        "Criteria" => (criterion[:ratings] || []).map do |rating|
          {
            "Description" => rating_description_for(criterion, rating),
            "Points" => rating[:points]
          }

View on GitHub (pinned to 1c9f0bb801)