AI Chatbot with Sentiment and In Kind Responses

✅ 1. Azure Resources Checklist

Install These

brew install dotnet-sdk #This will be version 9 which will not work but gives you the rest of the stuff needed.
brew install azure-cli

Assuming you already have created a Resource Group named don-test-rg, an OpenAI deployment named don-openai-useast, and deployed a gpt model named gpt-4.1 inside openai. Those all require the portal.

Install this from the webpage and follow directions. dotnet 8.x

Create a folder in your workspace and save the files at the end of this document.

Open a terminal, set your folder to the one where the files are saved, then run these commands:

These commands are only needed for Windows users:

dotnet nuget add source https://api.nuget.org/v3/index.json -n nuget.org
npm install -g azure-functions-core-tools@4 --unsafe-perm true
dotnet nuget locals all --clear
dotnet restore

The rest of the instructions are for every OS:

dotnet add package Microsoft.Azure.Functions.Worker.Extensions.OpenApi --version 1.4.0
dotnet --list-sdks
dotnet add package Azure.AI.TextAnalytics --version 5.3.0
dotnet add package Azure.AI.OpenAI --version 2.0.0

Create Storage Account

az storage account create \
  --name dehamerfuncstorage \
  --location eastus \
  --resource-group don-test-rg \
  --sku Standard_LRS

Create Function Plan for Windows

az functionapp plan create \
  --resource-group don-test-rg \
  --name dehamerfuncplan \
  --location eastus \
  --sku B1 \
  --is-linux false

Create Function App

az functionapp create \
  --resource-group don-test-rg \
  --name dehamerfuncapp \
  --plan dehamerfuncplan \
  --storage-account dehamerfunctstorage \
  --runtime dotnet-isolated \
  --functions-version 4 \
  --os-type Windows

Create TexAnalytics

az cognitiveservices account create \
  --name dehamersentimentai \
  --resource-group don-test-rg \
  --location eastus \
  --kind TextAnalytics \
  --sku S \
  --custom-domain dehamersentimentai

az cognitiveservices account update \
  --name dehamersentimentai \
  --resource-group don-test-rg \
  --set properties.publicNetworkAccess=Enabled

Get keys and endpoint for export and app settings

export TEXT_ANALYTICS_ENDPOINT=`az cognitiveservices account show \ 
  --name dehamersentimentai \
  --resource-group don-test-rg \
  --query "properties.endpoint"|sed -e s:\"::g`

export TEXT_ANALYTICS_KEY=`az cognitiveservices account keys list \
  --name dehamersentamentai \
  --resource-group don-test-rg \
  --query "key1" -o tsv`

az functionapp config appsettings set \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --settings \
  TEXT_ANALYTICS_ENDPOINT=$TEXT_ANALYTICS_ENDPOINT \
  TEXT_ANALYTICS_KEY=$TEXT_ANALYTICS_KEY
  
export OPENAI_KEY=`az cognitiveservices account keys list \
  --name don-openai-useast \
  --resource-group don-test-rg \
  --query "key1" -o tsv`

export OPENAI_ENDPOINT=https://don-openai-useast.openai.azure.com/

export OPENAI_DEPLOYMENT=gpt-4.1

az functionapp config appsettings set \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --settings \
  OPENAI_KEY=$OPENAI_KEY \
  OPENAI_ENDPOINT=$OPENAI_ENDPOINT
  OPENAI_DEPLOYMENT=$OPENAI_DEPLOYMENT
  
#Confirm they were set
az functionapp config appsettings list \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --query "[?starts_with(name, 'TEXT_') || starts_with(name, 'OPENAI_')]" -o table

Create a directory to store the functionapp files

mkdir ~/FeedbackFunctionDotNet
cd ~/FeedbackFunctionDotNet

Save the FeedbackFunction.cs, FeedbackFunction.csproj, global.json, host.json, local.settings.json, and program.cs into the FeedBackFunctionDotNet directory.

FeedbackFunction.cs

using System.Net.Http.Headers;
using System.Text;
using System.Text.Json;
using Azure;
using Azure.AI.TextAnalytics;
using Microsoft.Azure.Functions.Worker;
using Microsoft.Azure.Functions.Worker.Http;
using Microsoft.Extensions.Logging;

public class DeHamerFeedbackFunction
{
    private readonly ILogger _logger;

    public DeHamerFeedbackFunction(ILoggerFactory loggerFactory)
    {
        _logger = loggerFactory.CreateLogger<DeHamerFeedbackFunction>();
    }

    [Function("DeHamerFeedbackFunction")]
    public async Task<HttpResponseData> Run(
        [HttpTrigger(AuthorizationLevel.Anonymous, "post", Route = null)]
        HttpRequestData req)
    {
        _logger.LogInformation("DeHamerFeedbackFunction triggered.");

        var requestBody = await new StreamReader(req.Body).ReadToEndAsync();
        var data = JsonSerializer.Deserialize<Dictionary<string, string>>(requestBody);
        string feedback = data?["feedback"] ?? string.Empty;

        // Get Text Analytics config
        var textEndpoint = Environment.GetEnvironmentVariable("TEXT_ANALYTICS_ENDPOINT");
        var textKey = Environment.GetEnvironmentVariable("TEXT_ANALYTICS_KEY");

        _logger.LogInformation("Using Text Analytics endpoint: {endpoint}", textEndpoint);

        var credentials = new AzureKeyCredential(textKey);
        var client = new TextAnalyticsClient(new Uri(textEndpoint), credentials);
        var documentSentiment = await client.AnalyzeSentimentAsync(feedback);
        var sentiment = documentSentiment.Value.Sentiment.ToString().ToLower();

        _logger.LogInformation("Sentiment detected: {sentiment}", sentiment);

        // Compose prompt
        string prompt = sentiment switch
        {
            "positive" => $"Respond in a cheerful and thankful tone to the following positive customer feedback:\n\"{feedback}\"",
            "negative" => $"Respond in a harsh, sarcastic, and annoyed tone to the following negative customer feedback:\n\"{feedback}\"",
            _ => $"Respond neutrally and professionally to the following feedback:\n\"{feedback}\""
        };

        // OpenAI Setup
        var openaiEndpoint = Environment.GetEnvironmentVariable("OPENAI_ENDPOINT")?.TrimEnd('/');
        var openaiKey = Environment.GetEnvironmentVariable("OPENAI_KEY");
        var deployment = Environment.GetEnvironmentVariable("OPENAI_DEPLOYMENT");

        _logger.LogInformation("Calling OpenAI at: {url}", $"{openaiEndpoint}/openai/deployments/{deployment}/chat/completions");

        using var httpClient = new HttpClient();
        httpClient.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", openaiKey);

        var payload = JsonSerializer.Serialize(new
        {
            messages = new[]
            {
                new { role = "system", content = "You are a customer support agent who mirrors the customer's sentiment tone." },
                new { role = "user", content = prompt }
            }
        });

        _logger.LogInformation("Payload sent to OpenAI: {payload}", payload);

        var response = await httpClient.PostAsync(
            $"{openaiEndpoint}/openai/deployments/{deployment}/chat/completions?api-version=2024-02-15-preview",
            new StringContent(payload, Encoding.UTF8, "application/json"));

        if (!response.IsSuccessStatusCode)
        {
            var errorDetails = await response.Content.ReadAsStringAsync();
            _logger.LogError("OpenAI API call failed: {status} - {details}", response.StatusCode, errorDetails);

            var errorResponse = req.CreateResponse(System.Net.HttpStatusCode.InternalServerError);
            await errorResponse.WriteAsJsonAsync(new { error = "Failed to get OpenAI response", status = response.StatusCode, details = errorDetails });
            return errorResponse;
        }

        var json = JsonDocument.Parse(await response.Content.ReadAsStringAsync());
        _logger.LogInformation("Raw OpenAI response: {json}", json);

        var message = json.RootElement.GetProperty("choices")[0].GetProperty("message").GetProperty("content").GetString();

        var result = new
        {
            sentiment = sentiment,
            message = message
        };

        var responseData = req.CreateResponse(System.Net.HttpStatusCode.OK);
        await responseData.WriteAsJsonAsync(result);
        return responseData;
    }
}

FeeedbackFunction.csproj

<Project Sdk="Microsoft.NET.Sdk.Worker">
  <PropertyGroup>
    <TargetFramework>net8.0</TargetFramework>
    <AzureFunctionsVersion>v4</AzureFunctionsVersion>
    <OutputType>Exe</OutputType>
    <Nullable>enable</Nullable>
    <ImplicitUsings>enable</ImplicitUsings>
  </PropertyGroup>

  <ItemGroup>
    <PackageReference Include="Azure.AI.TextAnalytics" Version="5.3.0" />
    <PackageReference Include="Microsoft.Azure.Functions.Worker" Version="1.17.0" />
    <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.OpenApi" Version="1.5.0" />
    <PackageReference Include="Microsoft.Azure.Functions.Worker.Sdk" Version="1.17.0" OutputItemType="Analyzer" />
    <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.Http" Version="3.1.0" />
    <PackageReference Include="Microsoft.Extensions.Logging.Console" Version="8.0.0" />
    <PackageReference Include="System.Text.Json" Version="8.0.0" />
  </ItemGroup>

  <ItemGroup>
    <None Update="host.json">
      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
    </None>
    <None Update="local.settings.json">
      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
      <CopyToPublishDirectory>Never</CopyToPublishDirectory>
    </None>
  </ItemGroup>

</Project>

Make sure this matches the version of dotnet 8.x you install.

global.json

{
  "sdk": {
    "version": "8.0.411"
  }
}

host.json

{
  "version": "2.0"
}

local.settings.json

{
  "IsEncrypted": false,
  "Values": {
    "AzureWebJobsStorage": "UseDevelopmentStorage=true",
    "FUNCTIONS_WORKER_RUNTIME": "dotnet-isolated"
  }
}

program.cs

using Microsoft.Extensions.Hosting;
using Microsoft.Extensions.DependencyInjection;

var host = new HostBuilder()
    .ConfigureFunctionsWorkerDefaults()
    .Build();

host.Run();

From the FeedbackFunctionDonnet directory:

dotnet clean
dotnet publish -c Release -o publish
cd publish
zip -r ../publish.zip .
cd ..
az functionapp deployment source config-zip --resource-group don-test-rg --name dehamerfuncapp --src publish.zip

After running you will see additional files in the FeedbackFunctionDotnet directory. This is normal.

ls
bin                     FeedbackFunction.csproj host.json               obj                     publish
FeedbackFunction.cs     global.json             local.settings.json     Program.cs              publish.zip

To test if it is working

export FUNCTION_KEY=`az functionapp function keys list \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --function-name dehamerfeedbackfunction --query "default" | sed -e s:\"::`

curl -X POST https://dehamerfuncapp.azurewebsites.net/api/dehamerfeedbackfunction \
  -H "x-functions-key: $FUNCTION_KEY" \
  -H "Content-Type: application/json" \
  -d '{"feedback": "This service was trash. I want my time back."}'

You should see something like:

{"sentiment":"negative","message":"Oh, absolutely\u2014because we totally specialize in time travel, right? Sorry our \u0022trash\u0022 service didn\u2019t meet your sky-high expectations. Your valuable time is obviously worth so much, so thanks for investing it with us just to let us know how you feel."}%

Confirmation tests

az functionapp function list --name dehamerfuncapp --resource-group don-test-rg --query "[].{name:name, status:invokeUrlTemplate}" -o table
Name                                    Status
--------------------------------------  --------------------------------------------------------------------
dehamerfuncapp/DeHamerFeedbackFunction  https://dehamerfuncapp.azurewebsites.net/api/dehamerfeedbackfunction

az functionapp show --name dehamerfuncapp --resource-group don-test-rg --query "enabledHostNames"
[
  "dehamerfuncapp.azurewebsites.net",
  "dehamerfuncapp.scm.azurewebsites.net"
]

az functionapp function list \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --query "[].invokeUrlTemplate" \
  --output tsv

https://dehamerfuncapp.azurewebsites.net/api/dehamerfeedbackfunction

az functionapp function show \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --function-name dehamerfeedbackfunction \
  --query "isDisabled"

false

If true

az functionapp function update \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --function-name dehamerfeedbackfunction \
  --set isDisabled=false

Tailing logs if enabled. Notice it says webapp, not functionapp. This is because after deployment the functionapp is in a similar deployment mode to webapps and this command tails logs for both types. You would run this in one terminal logged in with az login and then run the curl from another terminal, also logged in unless you already have the key in your curl statement.

az webapp log tail \
  --name dehamerfuncapp \
  --resource-group don-test-rg

If you want it to autostart after issues.

#Optional
az webapp config set \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --always-on true

Get a list of your functionapps

az functionapp function list \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --output table

Href                                                                              InvokeUrlTemplate                                                     IsDisabled    Language         Location    Name                                    ResourceGroup    ScriptHref                                                                            TestData    TestDataHref
--------------------------------------------------------------------------------  --------------------------------------------------------------------  ------------  ---------------  ----------  --------------------------------------  ---------------  ------------------------------------------------------------------------------------  ----------  --------------------------------------------------------------------------------------------------------
https://dehamerfuncapp.azurewebsites.net/admin/functions/DeHamerFeedbackFunction  https://dehamerfuncapp.azurewebsites.net/api/dehamerfeedbackfunction  False         dotnet-isolated  East US     dehamerfuncapp/DeHamerFeedbackFunction  don-test-rg      https://dehamerfuncapp.azurewebsites.net/admin/vfs/site/wwwroot/FeedbackFunction.dll              https://dehamerfuncapp.azurewebsites.net/admin/vfs/data/Functions/sampledata/DeHamerFeedbackFunction.dat

Only do this for testing. It's not secure in the real world unless you don't want to have domain restrictions on access to the functionapp. I had to do it so I could test from my home webserver.

az functionapp cors add \
  --name dehamerfuncapp \
  --resource-group don-test-rg \
  --allowed-origins "*"

If you have a webserver that you trust, you can create this webpage and test it:

index.html

<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8" />
  <title>Dennis Feedback Page</title>
  <style>
    body {
      background-color: white;
      color: red;
      font-family: Arial, sans-serif;
      margin: 40px;
    }

    header {
      display: flex;
      align-items: center;
      margin-bottom: 40px;
    }

    header img {
      height: 50px;
      margin-right: 20px;
    }

    h1 {
      font-size: 2em;
    }

    textarea {
      width: 100%;
      height: 100px;
      font-size: 1em;
      padding: 10px;
    }

    button {
      margin-top: 10px;
      padding: 10px 20px;
      font-size: 1em;
      background-color: red;
      color: white;
      border: none;
      cursor: pointer;
    }

    #response {
      margin-top: 20px;
      font-weight: bold;
      white-space: pre-line;
    }
  </style>
</head>
<body>
  <header>
    <img src="https://webobjects2.cdw.com/is/image/CDW/cdw23-logo-no-tag-88x48?$transparent$" alt="CDW Logo" />
    <h1>Dennis Feedback Page</h1>
  </header>

  <form id="feedbackForm">
    <label for="feedback">Enter your feedback:</label><br>
    <textarea id="feedback" name="feedback" required></textarea><br>
    <button type="submit">Submit</button>
  </form>

  <div id="response"></div>

  <script>
    document.getElementById('feedbackForm').addEventListener('submit', async function (e) {
      e.preventDefault();

      const feedback = document.getElementById('feedback').value;
      const responseDiv = document.getElementById('response');

      responseDiv.textContent = "Processing...";

      try {
        const res = await fetch('https://dehamerfuncapp.azurewebsites.net/api/dehamerfeedbackfunction', {
          method: 'POST',
          headers: {
            'Content-Type': 'application/json'
          },
          body: JSON.stringify({ feedback })
        });

        if (!res.ok) {
          throw new Error("API error: " + res.status);
        }

        const data = await res.json();
        responseDiv.innerHTML = `<strong>Sentiment:</strong> ${data.sentiment}<br><strong>Response:</strong> ${data.message}`;
      } catch (err) {
        responseDiv.textContent = "Error: " + err.message;
      }
    });
  </script>
</body>
</html>

After resolving early build issues and switching from Linux to a Windows-hosted function plan due to SDK compatibility, the deployment succeeded with the following workflow:

- Rewrote the function class to eliminate naming conflicts
- Set AuthorizationLevel to Anonymous to permit open feedback submission
- Used curl to verify responses from the deployed function
- Created an HTML frontend to POST feedback to the function endpoint

AI Knowledge