[Bug]: the sixth Compression strategy for current round makes llm response empty

[tcsp]

AgentScope-Java is an open-source project. To involve a broader community, we recommend asking your questions in English.

Describe the bug
the sixth Compression strategy for current round makes llm response empty.

To Reproduce
Steps to reproduce the behavior:

1. enbale the AutoContextMeomry for reactAgent
2. chat and make context tokens greater than maxTokens in config
3. the AutoContextMemory trigger the sixth compression strategy
4. the compressed result returned normally
5. however, when the compressed result was put back into the context and a reasoning request was made again, an empty result was returned

reasoning request response just like:

{"choices":[{"delta":{"content":"","role":"assistant"},"finish_reason":"stop","index":0}],"created":1768463958,"id":"021768463956351c97886b1884070617909158979a5aac856af9d","model":"doubao-seed-1-8-251228","service_tier":"default","object":"chat.completion.chunk","usage":null}

{"choices":[],"created":1768463958,"id":"021768463956351c97886b1884070617909158979a5aac856af9d","model":"doubao-seed-1-8-251228","object":"chat.completion.chunk","usage":{"completion_tokens":0,"prompt_tokens":26762,"total_tokens":26762,"prompt_tokens_details":{"cached_tokens":0},"completion_tokens_details":{"reasoning_tokens":0}}}

Expected behavior
First, the reasoning request response should not be empty.

Second, I would rather agentscope could provide a switch control to skip the sixth compression policy.

Environment (please complete the following information):

• AgentScope-Java Version: 1.0.7
• Java Version: 17