Inkline
Menu

Eval pipeline

Templates
Eval pipeline 8 items and 7 connections. no yes Prompt or model change Load eval dataset Run the system on everycase Score with judges (LLM +rules) Compare with the baseline Any regression? Ship the change Block and report
Eval pipeline — preview
Flowchart

How a prompt or model change is gated: run the eval set, score with LLM and rule judges, compare with the baseline and block regressions.

What is inside

  • Dataset → run → judge → score
  • Regression gate
Use this template

Opens in the playground, no account needed. Sign up to save it.

Show the Inkline text
diagram flow "Eval pipeline"
  start    change  "Prompt or model change"
  io       dataset "Load eval dataset"
  step     run     "Run the system on every case"
  step     judge   "Score with judges (LLM + rules)"
  step     report  "Compare with the baseline"
  decision gate    "Any regression?"
  end      ship    "Ship the change"
  end      block   "Block and report"

  change -> dataset -> run -> judge -> report -> gate
  gate   -> ship  : no
  gate   -> block : yes