This demo adds a Chat Playground page to a basic .NET Aspire solution. It provides an experience similar to that of the Chat Playground feature in Azure AI Foundry.

The source code demonstrates using the streaming variant of the completions API to display the chat response fragments as they are received from the API call.

1. Create a resource group. Name it something like AI-RG.

   Don't have an Azure account? Create a free account from https://azure.microsoft.com. The $200 credit allows for plenty of experimentation for a month.

2. In the resource group, create an Azure OpenAI service resource. Preferrably choose EastUS2, if there is enough quota.

3. Click on the created in step 2.

4. Click on the Explore Azure AI Foundry portal.

5. In Azure AI Foundry, click on Deployments in the left nav menu.

6. Click on + Deploy Model. Click on Deploy Base Model from the dropdown.

7. Click on the model name. Copy the following pieces of information: a. Endpoint subdomain: It is usually openai.azure.com but could be cognitiveservices.azure.com. b. Service name (the name of the resource created in step 2) c. Model name (e.g. gpt-4.1) d. Deployment name (e.g. gpt-4.1) e. The API key

1. Open the solution in Visual Studio (the free Community Edition can be used).
2. From the Build menu, choose Build Solution.
3. Run the solution (press F5 key).

1. From the Aspire dashboard's Resources tab, click on the first link under the Endpoints column for the webfrontend row.
2. Click on Chat Playground from the left nav.
3. Plug the settings gathered in the Pre-requisites section above into the page's fields and click on the Get Answer button.
4. Modify the system prompt and test with different user prompts changing the sampling temperature between 0 and 2. Higher values enable more creative answers, but beware that as you get close to 2, the results can be unpredictable and non-deterministic (LSD-ish).