Skip to content

Compress Context ​

Introduction ​

In the "Java Web Project (New Version)" and "General Scenario" modes, during the source code generation step, as context accumulates, the model may "forget" early information, affecting the accuracy of subsequent code and increasing Token consumption. Therefore, we have added the "Compress Current Session" feature. You can flexibly choose between "Auto" or "Manual" modes: with auto compression enabled, the system will automatically process it when preset conditions are met; with manual mode, you can execute it on demand. You can check the context usage rate to decide whether to compress. This feature not only effectively improves the accuracy of model responses but also significantly reduces Token consumption.

Prerequisites: Context compression is only effective during the source code generation step in "Java Web Project (New Version)" and "General Scenario" modes, and can only be performed after the task is completed or cancelled.

Auto Compression ​

  1. After the source code generation task stops in "Java Web Project (New Version)" and "General Scenario" modes, hover over the "Compress Icon" and check "Auto Compress". When the context content meets the compression conditions, it will be automatically compressed.

Prerequisites: Context compression is only effective during the source code generation step in "Java Web Project (New Version)" and "General Scenario" modes, and can only be performed when the task is stopped.

Manual Compression ​

  1. After the source code generation task stops in "Java Web Project (New Version)" and "General Scenario" modes, hover over the "Compress Icon" to check the context usage rate. If the usage rate is too high, you can click the "Compress Current Session" button to compress.

Context Usage Rate Note: 1M represents 1 million tokens, approximately equal to 3 million characters. Usage rate percentage calculation: Current context characters / (500,000 * 3)

  1. When the page displays "Compression Complete", the compression is successful, the context usage rate will decrease, and you can then continue generating source code.

邮箱:chendw@feisuanyz.com 邮编:518000 地址:深圳市前海深港合作区前湾一路1号A栋201室