注意避雷 command code chat 端点 不按标准协议来实现

作者：#Benin 板块：#开发调优 deepseek 官网明确表示每次都需要回传 reasoning_content 内容确保模型性能，但是 command code 的chat 端点不传 reasoning_content 而是传的 reasoning_detail，直接导致harness不回传+不显示思考内容，非常诡异的协议实现，且还不支持response和messages协议。 以下是commandcode的响应 { "id": "gen_01M2KB20EAZKASB1K0JXMEZ3XN", "object": "chat.completion", "created": 1789503014, "model": "deepseek/deepseek-v4.1-flash", "choices": [{ "index": 0, "message": { "role": "assistant", "content": "391", "reasoning": "We need answer only number. 17*23=391. Must reply just number. ", "reasoning_details": [ { "type": "reasoning.text", "text": "We need answer only number. ...", "format": "unknown", "index": 0 } ] }, "logprobs": null, "finish_reason": "stop" }], "usage": { "prompt_tokens": 43, "completion_tokens": 21, "total_tokens": 64, "completion_tokens_details": { "reasoning_tokens": 19, "image_tokens": 0 } } } 刚购买我就试用了一下，首字很慢，现在虽然首字解决了但是从根本上协议就有问题，三天前我就退款邮件发了，至今没有响应。