← All Docs

Quality & Evaluation

Evaluate & Improve

Evaluate & Improve is Webase's automated quality improvement loop. It evaluates your app, fixes the issues it finds, and re-evaluates — repeating until your app passes or the maximum number of iterations is reached.

Free to use. Like standalone evaluations, the Evaluate & Improve loop does not consume any of your monthly messages. The AI fixes and re-evaluations are all included at no extra cost.

How It Works

The Evaluate & Improve loop follows a simple cycle: evaluate your app, identify issues, fix them, and then evaluate again to verify the fixes worked. This cycle repeats automatically.

1

Evaluate

The AI reviewer scores your app and identifies specific requirements that failed. Each issue includes a description of what went wrong and what needs to change.

2

Fix

The AI takes the evaluation feedback and updates your source files to address the identified issues. It modifies components, fixes imports, corrects logic errors, and improves code structure.

3

Re-Evaluate

After applying fixes, the AI runs a fresh evaluation to check whether the issues have been resolved and whether the score has improved.

4

Repeat or Stop

If the app now passes (score of 70 or above), the loop stops. If issues remain, the cycle repeats. The loop runs up to 3 iterations maximum to prevent endless cycles.

Running Evaluate & Improve

To start the automated improvement loop:

  1. Open your app in the Application Editor.
  2. Click the Evaluate & Improve button in the toolbar.
  3. Watch the progress as the AI evaluates, fixes, and re-evaluates your app.
  4. Review the final results when the loop completes.

The process typically takes 1 to 3 minutes depending on how many issues are found and how many iterations are needed. There is a short 2-second delay between iterations to respect API rate limits.

What Happens During Each Iteration

In each iteration, the AI focuses on the most impactful issues first:

  • Runtime errors — Broken imports, missing components, and crash-causing bugs are fixed first since they prevent the app from working at all.
  • Failed requirements — Features that are missing or incomplete are added or corrected.
  • Quality improvements — Code organization, user experience, and best practice compliance are improved last.

Maximum 3 iterations. The loop runs at most 3 times. If your app still has issues after 3 iterations, you can review the remaining findings and address them manually through the chat, or run Evaluate & Improve again for another round.

When to Use Evaluate & Improve

This feature works best in these situations:

  • After initial generation — Run it right after generating a new app to catch and fix any issues from the first pass.
  • After major changes — If you made significant updates through chat, run Evaluate & Improve to ensure nothing broke.
  • Before sharing or deploying — Use it as a final quality check before publishing your app.

Comparing Evaluate vs. Evaluate & Improve

  • Evaluate only checks your app and reports issues. It does not make changes. Use it when you want to understand the current state of your app.
  • Evaluate & Improve checks your app and fixes the issues automatically. Use it when you want the AI to handle the fixes for you.

Both are free and can be run as many times as you like. See App Evaluation for more details on how scoring and requirements work.