curl --location 'https://tokenhub-intl.tencentcloudmaas.com/plan/v3/chat/completions' \\
--header 'Authorization: Bearer $your_api_key' \\
--header 'Content-Type: application/json' \\
--data '{
  "model": "$model_id",
  "messages": [
    {
      "role": "user",
      "content": "写一首关于春天的诗"
    }
  ],
  "stream": false
}'

{
    "id": "********-****-****-****-************",
    "object": "chat.completion",
    "model": "auto",
    "created": *******,
    "choices": [
        {
            "index": 0,
            "message": {
                "role": "assistant",
                "content": "**春日漫步**\\n\\n春风拂面柳絮轻，  \\n桃红李白映湖光。  \\n碧草如烟铺小径，  \\n燕舞蜂歌映日晴。  \\n翠鸟鸣枝声细细，  \\n清泉石上映星光。  \\n万物苏醒歌盛世，  \\n人间四月尽芳菲。",
                "reasoning_content": "The user is speaking Chinese: \\"写一首关于春天的诗\\" which translates to \\"Write a poem about spring\\". So we need to comply, providing a poem about spring in Chinese.\\n\\nWe should generate a Chinese poem. There's no policy conflict. The user wants a poem about spring. That is fine.\\n\\nWe should produce a poem in Chinese, maybe a classical style or free verse. The user is Chinese, likely wants a Chinese poem.\\n\\nThe policy: There's no disallowed content. So just comply.\\n\\nWe could produce a 5- or 7-character per line poem, e.g., a \\"七绝\\" or a \\"五律\\". Or a free verse. Let's produce a short Chinese poem with vivid imagery. The poem might be 4-8 lines. Should be about spring. Use imagery of blossoming, birds, gentle breezes, green grass, etc.\\n\\nThus we can output something like:\\n\\n春风拂面柳絮轻，  \\n桃红李白映湖光。  \\n碧草如烟铺小径，  \\n燕舞蜂歌映日晴。  \\n\\nThat's a four-line poem with 7 characters each maybe? Let's count: \\"春风拂面柳絮轻\\" has 7 characters: 春风 (2) 拂 (1) 面 (1) 柳 (1) 絮 (1) 轻 (1) - it's 7? Actually Chinese characters: 春(1) 风2 拂3 面4 柳5 絮6 轻7. Yes 7. Then second line \\"桃红李白映湖光\\" also 7. Third line \\"碧草如烟铺小径\\" also 7. Fourth line \\"燕舞蜂歌映日晴\\" also 7. Good.\\n\\nAlternatively, maybe do a 5- or 7-line poem in a more modern style. But the user just said \\"写一首关于春天的诗\\", so either is fine.\\n\\nWe can also add a title: \\"春之韵\\".\\n\\nLet's provide a short poem: \\"春风拂面柳絮轻，桃红李白映湖光，碧草如烟铺小径，燕舞蜂歌映日晴\\". Possibly add more lines to fill up to 8 lines: \\"翠鸟鸣枝声细细，清泉石上映星光，万物苏醒歌盛世，人间四月尽芳菲\\". That would be 8 lines of 7 characters each (if we maintain 7 characters each). Let's see: \\"翠鸟鸣枝声细细\\" - characters: 翠1 鸟2 鸣3 枝4 声5 细6 细7: yes 7. \\"清泉石上映星光\\" - 清1 泉2 石3 上4 映5 星6 光7. \\"万物苏醒歌盛世\\" - 万1 物2 苏3 醒4 歌5 盛6 世7. \\"人间四月尽芳菲\\" - 人1 间2 三3 月4 尽5 芳6 菲7. Good. Then that is 8 lines of 7 characters each, forming a \\"七绝\\" variant. Actually it's 8 lines, maybe a \\"七律\\" (seven-character regulated verse) but 8 lines is okay.\\n\\nWe can present the poem with a title \\"春日漫步\\". Then provide the poem.\\n\\nWe should be mindful of possible Chinese punctuation. Use Chinese punctuation: \\"，\\" and \\"。\\" or just line breaks.\\n\\nWe can give a note: It's a classical style poem, describing spring scenery.\\n\\nTherefore answer: a poem.\\n\\nWill respond in Chinese, as the user wrote in Chinese. Provide the poem. That's it.\\n\\nNow output."
            },
            "finish_reason": "stop"
        }
    ],
    "usage": {
        "prompt_tokens": 46,
        "completion_tokens": 813,
        "total_tokens": 859,
        "prompt_tokens_details": {
            "cached_tokens": 0
        },
        "completion_tokens_details": {
            "reasoning_tokens": 748
        }
    }
}