OpenAI API VBA function returns #Value! but MsgBox displays response

题意:"OpenAI API VBA 函数返回 #Value!,但 MsgBox 显示响应"

问题背景:

I am trying to integrate the OpenAI API into Excel. The http request to OpenAI chat completion works correctly and the response is OK. When I display it with a MsgBox, it looks fine.

"我正在尝试将 OpenAI API 集成到 Excel 中。对 OpenAI 聊天完成的 HTTP 请求正常工作,响应也正常。当我使用 MsgBox 显示它时,它看起来很好。"

But when the function is called in a sheet, the returned value is #VALUE!

"但当在工作表中调用该函数时,返回的值是 #VALUE!"

python 复制代码
Public Function GPT(InputPrompt) As String
  'Define variables
  Dim request As Object
  Dim text, response, API, api_key, DisplayText, GPTModel As String
  Dim GPTTemp As Double
  Dim startPos As Long
  Dim rng As Range
  Dim httpRequest As Object
  Dim countArray As Variant

 
  'API Info
  API = "https://api.openai.com/v1/chat/completions"
  api_key = Trim(Range("API_Key").value)
  'Note: for future upgrades, please replace with GPT-4 etc
  GPTModel = Range("Model_Number").value
  
  If api_key = "" Then 'API key is missing!
    MsgBox "Error: API cannot be blank! Please go to 'Configuration' tab and enter a valid OpenAI API key", vbExclamation, "GPT for Excel"
    frmStatus.Hide
    Exit Function
  End If
    
  'Clean input text and make JSON safe
  text = CleanInput(InputPrompt)
  
  'Create request object
  Set httpRequest = CreateObject("MSXML2.XMLHTTP")
  'Set httpRequest = New MSXML2.XMLHTTP60
  
  'Get temp from Config panel
  GPTTemp = Range("GPT_temperature").value
  
  'Assemble request body
  Dim requestBody As String
  requestBody = "{""model"": ""gpt-4"", ""messages"": [{""role"": ""user"", ""content"": """ & text & """}], ""temperature"": 0.7}"

  With httpRequest
     .Open "POST", API, False
     .setRequestHeader "Content-Type", "application/json"
     .setRequestHeader "Authorization", "Bearer " & api_key
     .send (requestBody)
  End With

  If httpRequest.Status = 200 Then 'Successfully called OpenAI API
   response = httpRequest.responseText
   'Get usage info from response object
   countArr = ExtractUsageInfo(response)
    
   startPos = InStr(response, """content"":") + Len("""content"":") + 2
   endPos = InStr(startPos, response, "},")
   DisplayText = Trim(Mid(response, startPos, endPos - startPos))
   DisplayText = Mid(DisplayText, 1, Len(DisplayText) - 2)
   DisplayText = CleanOutput(DisplayText)
    
   Set request = Nothing
   If FillActive = False Then frmStatus.Hide
   MsgBox (GPT)
   GPT = DisplayText
   MsgBox (GPT)
   If Range("Log_Data").value = "Yes" Then
    Call UpdateLogFile(countArr)
   End If
  Exit Function

End Function

问题解决:

You are limited as to what you can do in a custom worksheet function. While I can't find an explicit reference to making web or API calls in the MS Docs my experience of worksheet functions leads me to not be surprised that this doesn't work.

"在自定义工作表函数中,你的操作是有限的。虽然我在微软文档中找不到明确提到进行 Web 或 API 调用的内容,但根据我对工作表函数的经验,这种方法不起作用并不令人意外。"

An alternative would be to add a Button or Shape to your worksheet then associated that button with a Sub such as

"另一种方法是在你的工作表中添加一个按钮或形状,然后将该按钮与一个 `Sub` 关联,例如:"

python 复制代码
Sub ButtonClicked()
    Dim Response As String, InputPrompt As String
    InputPrompt = CStr(ThisWorkbook.Sheets("Sheet1").Range("A1").Value)
    Response = GPT(InputPrompt)
    ThisWorkbook.Sheets("Sheet1").Range("A2").Value = Response
End Sub

Which uses your existing GPT function to get a response for the text in cell A1 and loads it into cell A2 (both in 'Sheet1' in this case ... obviously you can change the cells and worksheet as needed).

"这个方法使用你现有的 `GPT` 函数来获取单元格 A1 中的文本的响应,并将其加载到单元格 A2 中(在本例中,这两个单元格都在 'Sheet1' 中......显然,你可以根据需要更改单元格和工作表)。"

相关推荐
三声三视2 小时前
75 条文章索引被一条 add 清成 1 条,退出码还是 0:tri-article 的 index.py 我读了 205 行
人工智能·ai·skill·tri-skills·tri-article
小葱炖豆腐3 小时前
python绘制excel折线图
python·excel·numpy·pandas·matplotlib
牛油果子哥q3 小时前
生产级AI项目上线全流程:Docker容器化、服务编排、监控告警、日志收集、容灾降级、线上运维闭环
人工智能·ai
星辰徐哥3 小时前
本地视频预览别只自己看:把Remotion动效项目发给客户远程验收
docker·ai·node.js·html·音视频·react·remotion
HYDtomako6 小时前
Agent checkpoint设计
ai·agent·checkpoint·claude code
七灵微7 小时前
【AI】代码实战- 基于时间序列的轴承检测1
人工智能·机器学习·ai·pandas·matplotlib
RSABLOCKCHAIN7 小时前
读懂AI-Agent项目Career-ops项目的底层原理
ai
十正8 小时前
把表变成模型够得着的句柄
人工智能·ai·agent
长谷深风1118 小时前
支付审批的智能风险控制设计
大数据·人工智能·ai·大模型·支付·aiagent·hitl
feasibility.8 小时前
当“给机器人指路“成为一门数据工程学——SimpleNav 拆解:导航大模型的标准化之争
ai·机器人·无人机·导航·具身智能·vla·vln