Description

GenAI interactions are hard to read in Aspire's default telemetry experience:

• GenAI information is spread across span attributes, span events and log entries.
• Input and output content is embedded in complex JSON payloads. It is possible but awkward to read simple text messages from JSON payloads. Large large text messages, or messages which are JSON/XML encoded inside JSON, are very difficult to understand.
• The standard has changed significately 3 times so far.

This PR adds a specialized GenAI dialog for viewing GenAI operations. When GenAI telemetry is detected then a button is added to GenAI spans and logs. The button launches a GenAI viewer dialog and displays content in an easy-to-use UI. This approach is similar to what Aspire does for log entries with exception messages, which have a dialog dedicated to viewing log entry exception details.

CircuitBrain is the button icon and seems like the most universial icon for GenAI at the moment. Feedback welcome.

The dialog displays highlevel information about the GenAI operation (name, model, start time, duration, token usage). It also displays all the input and output information. What were the system instructions, input messages and tool call results (including the message history of assistant responses) and output messages.

Clicking on a message displays its content in more detail. The "Raw" tab displays the raw text without formatting. The "Preview" tab displays content in a user friendly preview to the user:

• JSON and XML content is indented and highlighted.
• Other content is assumed to be markdown, and rendered as markdown.

There is a button to copy message content to the clipboard.

Other PR features/changes:

• There have been 3 different ways of representing GenAI input and output content. All 3 are supported. Each is attempted, newest to oldest, to find where content is stored.
• The dialog has "Next" and "Previous" buttons so you can navigate between GenAI operations in the current span without closing and reopening the dialog.
• Change chat AI integration to use telemetry reported by MEAI instead of OpenAI or Azure.AI.Inference packages. MEAI is activitely updated. Telemetry in other packages is out of date and doesn't receive updates.

To do:

• Add a configuration option to Aspire chat AI integration about capturing sensitive data (i.e. message content)
• Move text to resource files

Major follow up work, that's not part of this PR:

• Add ability to quickly filter traces and spans to known types. e.g. select "GenAI" in a dropdown to only see traces and spans that have GenAI operations in them.

Fixes # (issue)

Checklist

• Is this feature complete?
  • Yes. Ready to ship.
  • No. Follow-up changes expected.
• Are you including unit tests for the changes and scenario tests if relevant?
  • Yes
  • No
• Did you add public API?
  • Yes
    • If yes, did you have an API Review for it?
      • Yes
      • No
    • Did you add <remarks /> and <code /> elements on your triple slash comments?
      • Yes
      • No
  • No
• Does the change make any security assumptions or guarantees?
  • Yes
    • If yes, have you done a threat model and had a security review?
      • Yes
      • No
  • No
• Does the change require an update in our Aspire docs?
  • Yes
    • Link to aspire-docs issue (consider using one of the following templates):
      • New (or update) doc-idea template
      • New breaking-change template
      • New diagnostic template
  • No