Hello,
This is GPT Prompt Tester.
With the v0.6.6 update, we introduced the new Eval / Improve feature and improved the stability of Playground options along with internal structural enhancements.
Below are the key updates included in this release.
✨ Key Updates
🔍 Eval / Improve Feature Added (Pro Only)
A new Eval (Evaluation) feature has been added to analyze prompt execution results.
Based on the evaluation results, you can also use the Improve workflow to generate improved versions of your prompt.
This allows users to run a more structured cycle of prompt evaluation and refinement.
※ The Eval / Improve feature is available for Pro license users only.

Prompt Result Analysis Menu
In the Playground panel:
- Click the More menu (①) located to the right of the [Run] button.
- Select [Analyze Result] (②) to generate the evaluation immediately.
Below is an example of the Evaluation result screen.

Prompt Result Evaluation Screen
To generate improvement suggestions:
- Click the Improve (④) tab at the top of the panel.
- To return to the evaluation results, click the Evaluation (③) tab.

This allows you to move easily between analysis and improvement workflows.
Prompt Evaluation and Improvement Flow
When you open the Improve tab, the system generates prompt improvement suggestions based on the evaluation.
After generation, you can click the [Regenerate Improve] button in the upper-right corner to create a new suggestion at any time.

Applying the Improved Prompt
Once an improved prompt is generated, you can immediately apply it by clicking:
[Apply to Current Panel]
This allows you to quickly test the improved prompt within the same Playground panel.
🧩 Improved Playground Web Search Tool Conditions
The conditions for selecting the Web Search tool in Playground have been updated.
If the JSON output format is selected:
- The Web Search tool will be disabled
This prevents confusion that may occur when combining JSON output and web search results.
🧪 json_schema Status Update
The json_schema output option is currently under development.
In the current version, it is displayed as “Coming Soon.”
Full support will be added in a future update.
🛠️ Other Improvements
- Internal processing logic has been improved to enhance overall stability
- Edge cases that caused unstable behavior in certain situations have been fixed
GPT Prompt Tester will continue to evolve as a tool focused on:
prompt experimentation · comparison · evaluation · improvement
If you encounter any issues or have suggestions for improvements, please feel free to share them with us anytime.
Thank you 🙏
— GPT Prompt Tester Team



