Add classifyCategories function for llm scorecard

[nlarew]

Jira: (EAI-1107) Add Question Categorizations to Case Data

Changes

• Automatic classifier for future categories
  • based on https://docs.google.com/spreadsheets/d/1g2XdsNa1T2_vhJWlxDInvG0UlWmKUvsuqTA_BFMGzK0/edit?gid=1488777287#gid=1488777287

[Copilot]

Pull Request Overview

This PR introduces a new LLM-based classifier for question categorizations, adds a sample categories CSV, and updates package dependencies.

• Add a comprehensive sample of categorized questions in CSV format for the classifier.
• Implement classifyCategories.ts with an OpenAI-powered classifier and inline category definitions.
• Reorder and dedupe dependencies in package.json for the llm-scorecard package.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
packages/scripts/src/llm-scorecard/examples/categories.csv	Add example dataset of questions with dataset, label, and question
packages/scripts/src/llm-scorecard/classifyCategories.ts	Create classifyCategories function using makeClassifier
packages/scripts/package.json	Reorder and remove duplicate dependency entries

Comments suppressed due to low confidence (2)

packages/scripts/src/llm-scorecard/classifyCategories.ts:145

• [nitpick] There are no unit tests for classifyCategories. Adding tests with known prompts and expected categories will help ensure the classifier behaves as intended.

export const classifyCategories = makeClassifier({

packages/scripts/src/llm-scorecard/classifyCategories.ts:145

• Add a JSDoc comment above this export to explain the purpose, input format, and output structure of classifyCategories for future maintainers.

export const classifyCategories = makeClassifier({

[Copilot]

The header row ends with an extra comma, which may lead to an unintended empty column when parsing. Please remove the trailing comma so it reads Dataset,Label,Question.

Suggested change

	Dataset,Label,Question,
	Dataset,Label,Question

[Copilot]

[nitpick] This inline definition of many category examples can become hard to maintain as you add new categories. Consider loading this data from an external JSON or CSV file and transforming it into classificationTypes to keep the code concise.