✨ feat: add custom stream handle support for LobeOpenAICompatibleFactory

[hezhijie0327]

💻 变更类型 | Change Type

• ✨ feat
• 🐛 fix
• ♻️ refactor
• 💄 style
• 👷 build
• ⚡️ perf
• 📝 docs
• 🔨 chore

🔀 变更说明 | Description of Change

1. 为 LobeOpenAICompatibleFactory 引入 handleStream 及 handleTransformResponseToStream 接口，用于应用自定义流处理
2. Qwen 切回 LobeOpenAICompatibleFactory
   a. 对语言模型默认启用内置联网搜索能力，引入 enable_search (印象里之前有个 Issue 提了这个 feat，过会儿翻下) [Request] 增加通义千问打开联网搜索的开关 #4877
   b. 对 frequency_penalty presence_penalty temperature top_p 进行范围限制
   - ref: https://help.aliyun.com/zh/dashscope/developer-reference/tongyi-qianwen-vl-plus-api
   - ref: https://help.aliyun.com/zh/dashscope/developer-reference/qwen-api

功能	效果
Qwen 互联网搜索
多模态
工具调用

3. 补齐之前 Spark 缺少的 function call 的实现
   a. 通过改造实现了 Spark 的 Function Call 功能（整体工具调用效果不是特别好，存在限制）
   b. 支持流式及非流式对话
   c. 试了下工具调用支持流式，但是除了 max-32k 模型支持工具回调外，其余均不支持（curl 调用测试也为空；max-32k 存在相同 payload，上一秒还能出内容，再次执行直接返回空的问题）

Function Name	描述
transformSparkStream	处理流式
transformSparkResponseToStream	处理非流式

Spark 的工具调用流很怪，返回的是对象不是数组，通过魔改 OpenAI 流处理对非标准返回强制转换为数组

---

PS: 给讯飞开了工单，将近三个多月无进展

📝 补充信息 | Additional Information

模块	效果
对话
工具调用

---

非流式响应流

{
    "code": 0,
    "message": "Success",
    "sid": "cha000b95d2@dx193cef49db1b894532",
    "choices": [
        {
            "message": {
                "role": "assistant",
                "content": "",
                "tool_calls": {
                    "type": "function",
                    "function": {
                        "arguments": "{\"city\":\"Shanghai\"}",
                        "name": "realtime-weather____fetchCurrentWeather"
                    }
                }
            },
            "index": 0
        }
    ],
    "usage": {
        "prompt_tokens": 8,
        "completion_tokens": 0,
        "total_tokens": 8
    }
}

流式输出响应流

{
    "code": 0,
    "message": "Success",
    "sid": "cha000b6cd0@dx193cef628319a4b532",
    "id": "cha000b6cd0@dx193cef628319a4b532",
    "created": 1734344060,
    "choices": [
        {
            "delta": {
                "role": "assistant",
                "content": "",
                "tool_calls": {
                    "type": "function",
                    "function": {
                        "arguments": "{\"city\":\"Shanghai\"}",
                        "name": "realtime-weather____fetchCurrentWeather"
                    }
                }
            },
            "index": 0
        }
    ],
    "usage": {
        "prompt_tokens": 8,
        "completion_tokens": 0,
        "total_tokens": 8
    }
}

[vercel]

@hezhijie0327 is attempting to deploy a commit to the LobeHub Team on Vercel.

A member of the Team first needs to authorize it.

[lobehubbot]

👍 @hezhijie0327

Thank you for raising your pull request and contributing to our Community
Please make sure you have followed our contributing guidelines. We will review it as soon as possible.
If you encounter any problems, please feel free to connect with us.
非常感谢您提出拉取请求并为我们的社区做出贡献，请确保您已经遵循了我们的贡献指南，我们会尽快审查它。
如果您遇到任何问题，请随时与我们联系。

[codecov]

Codecov Report

Attention: Patch coverage is 87.19212% with 26 lines in your changes missing coverage. Please review.

Project coverage is 92.14%. Comparing base (0361ced) to head (c0f38d2).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
src/libs/agent-runtime/utils/streams/spark.ts	82.83%	23 Missing ⚠️
src/libs/agent-runtime/qwen/index.ts	93.61%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5039      +/-   ##
==========================================
- Coverage   92.18%   92.14%   -0.05%     
==========================================
  Files         549      550       +1     
  Lines       41575    41643      +68     
  Branches     3734     2536    -1198     
==========================================
+ Hits        38327    38371      +44     
- Misses       3248     3272      +24     

Flag	Coverage Δ
app	92.14% <87.19%> (-0.05%)	⬇️
server	97.65% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[arvinxx]

这块是不是最好看下哪里需要给工厂函数开槽的，给开个槽？单独实现一个感觉后续的维护成本比较高

[hezhijie0327]

加了 handleStream 及 handleTtransformResponseToStream 接口，功能上没啥问题了 ，还差一个非流式转流式的自定义处理

[hezhijie0327]

@arvinxx 改造好了

1. 顺带把 Qwen 也切换回 LobeOpenAICompatibleFactory 并引入 handleStream: QwenAIStream 处理自定义流
2. 为 Qwen Chat 模型引入 enable_search 默认启用内建联网功能
3. 为 Qwen 引入 QwenLegacyModels 数组，用于处理部分模型不支持 presence_penalty

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

@arvinxx has been transformed

1. By the way, switch Qwen back to LobeOpenAICompatibleFactory and introduce handleStream: QwenAIStream to process custom streams
2. Introduce enable_search to the Qwen Chat model to enable the built-in networking function by default
3. Introduce the QwenLegacyModels array to Qwen to handle some models that do not support presence_penalty