Fix custom and off topic to use llm_base

[steven10a]

Fixes reported [BUG #29 ]

• topical-alignment check hard codes temperature=0 which is not valid for gpt-5 series models and causes an error

The fix:

• All LLM based checks should be using llm-base which properly handles the temperature param
• Additionally noticed user-defined-llms was not properly using llm-base, fixed that
• Updated tests for both guardrails

[Copilot]

Pull Request Overview

This PR refactors the user-defined LLM and topical alignment guardrails to use a common createLLMCheckFn factory function, reducing code duplication and improving maintainability.

• Replaces manual LLM check implementations with calls to createLLMCheckFn from llm-base.ts
• Renames exported check functions from userDefinedLLMCheck/topicalAlignmentCheck to userDefinedLLM/topicalAlignment
• Rewrites tests to verify the refactored implementation with multiple GPT model variants

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
src/checks/user-defined-llm.ts	Refactored to use createLLMCheckFn factory, reducing ~150 lines of duplicate LLM handling code and renaming the exported check function
src/checks/topical-alignment.ts	Refactored to use createLLMCheckFn factory, reducing ~80 lines of duplicate LLM handling code and renaming the exported check function
src/tests/unit/checks/user-defined-llm.test.ts	Updated tests to verify the refactored implementation with model-specific temperature tests for gpt-4, gpt-5, gpt-4o, and gpt-3.5-turbo
src/tests/unit/checks/topical-alignment.test.ts	Updated tests to verify the refactored implementation with model-specific temperature tests for gpt-4, gpt-5, gpt-4o, and gpt-3.5-turbo

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[gabor-openai]

TY