5 / 5
LLM evaluation

Set a temperature for repeatable tests

Easy30 pts~12 min
  • Determinism
  • Temperature
  • Seeds
Practice app · Acme Support Assistant

A deterministic LLM-style support assistant with retrieval (RAG), JSON mode, safety policies and tool calls, exposed via UI and API.

BASE_URL
/api/practice
Console app
/lab/ai-testing-set-a-temperature-for-repeatable-tests

Your starter code already declares BASE_URL — call the API relative to it.

Objective

Pin temperature and seed so the same request yields byte-identical responses.

Your task

  1. 1Ask "When is customer support available?" with temperature 0 and seed 42.
  2. 2Repeat the identical request.
  3. 3Assert output_text and id are identical across both calls and contain “09:00–18:00 UTC”.

Acceptance criteria

  • POST /ai/chat returns 200
  • temperature and seed are set
  • At least 2 assertions pass

LLM evaluation · AI Testing · Foundations of AI testing