AI Prompt Batch Testing Tool

Read text in a folder as prompts, batch call AI APIs, and save response results, supporting multiple API formats and concurrent processing.

Bug report

Loading tool, please wait...

If the online tool fails to load or run, try the desktop tool.https://tools.yikeaigc.com/

Tool Usage

Return to old version

A desktop version is available with more powerful, stable features. Open desktop version

Instructions

Software Usage Instructions

Set the input folder: In "Path Settings", select the folder that stores the prompt text files. Each text file in the folder will be treated as a complete prompt. To process files in subfolders, check "Traverse Subdirectories"; if you want the output directory to retain the original subdirectory hierarchy, you can also check "Keep Original Path Structure".
Set the save directory: Select the output directory for the processing results. Responses returned by the API will be saved in .txt format with the same name as the input file. If duplicate names occur, numbers will be appended automatically to distinguish them.
Configure API parameters:
- API format: Supports OpenAI, NewAPI, sab2api, Gemini, Anthropic
- Model name: Enter the model name to call, such as gpt-4o-mini
- API address: For OpenAI-compatible APIs, you can enter only the domain name, and the program will automatically complete the path
- API Key: Enter the access key for the corresponding API
- File extension: Default is .txt; separate multiple extensions with English commas
Set run controls: Set the concurrency, timeout for a single request, and the number of automatic retries after a request fails according to your API quota.
Start processing: Click the "Start" button, and the tool will read the prompts one by one and call the API. Empty text files will be skipped automatically.
View progress and logs: View the real-time processing status in the "Progress and Logs" area. After processing is complete, the log will summarize the number of successful, failed, and skipped items.

Frequently Asked Questions

A: Each text file in the input folder is a complete prompt, and all text inside the file will be sent to the model as a single request. It is recommended to place only prompt text files in the input folder to avoid mixing in other unrelated files.

A: The tool supports API formats such as OpenAI, NewAPI, sab2api, Gemini, and Anthropic. For OpenAI-compatible APIs, you can enter only the domain name in the API address field, and the program will automatically complete the /v1/chat/completions path.

A: The response results will be saved to the save directory you set and output in .txt format with the same name as the input file. If a duplicate name occurs, the tool will automatically append a number to the file name to distinguish it and will not overwrite existing results.

A: Concurrency refers to the number of prompts processed at the same time. Please set it appropriately according to the API rate quota; setting it too high may trigger API rate limits. Timeout is the maximum waiting time for a single request, and retry count is the maximum number of automatic retries after a request fails. When the network is unstable, increasing the retry count appropriately can help improve the success rate.

A: "Traverse subdirectories" determines whether prompt files in subfolders are read. "Keep original path structure" is available only after "Traverse subdirectories" is checked. When checked, the output directory will recreate the subdirectory hierarchy within the main input folder, making it easy to match results with the original files one by one.

A: Blank text files will be automatically skipped and marked as SKIP in the log. Common causes of processing failure include an incorrect API address or key, an incorrect model name, insufficient API quota, or network errors. You can troubleshoot according to the error messages in the log and confirm that the API parameters are filled in correctly.

A: Yes. By default, only .txt files are processed in "File extensions". You can also enter multiple extensions separated by English commas, and the tool will only read and process text files with matching extensions.