Feature Request: Allow developers to set the iframe 'allow' attribute for HtmlService

Description

This is a feature request to give developers more options to configure the allow attribute for iframes generated by HtmlService.

The immediate use case is to enable on-device AI capabilities like Gemini Nano, which requires the host iframe to include the allow="language-model" attribute to grant access to the Prompt API, as outlined in recent Chrome AI announcements. Without this, it's currently impossible to build Workspace Add-ons that leverage these new browser features.

Currently, add-on iframes are served with an extensive list of permissions (e.g., camera, geolocation, microphone), which demonstrates that a permissions framework is already in place. To provide developers with control, we suggest a new method for the HtmlOutput class, such as .setIframeAllow(permissionString). This new method could operate on a list of Google-approved permission strings, ensuring that only safe and vetted capabilities can be enabled, including the new language-model policy.

Code

Here is a hypothetical example of how the proposed feature would be used in Code.gs, along with the corresponding client-side HTML.

`Code.gs` (Hypothetical Usage)

function onOpen() {
  DocumentApp.getUi()
    .createMenu('Nano AI')
    .addItem('Show Demo', 'showSidebar')
    .addToUi();
}

function showSidebar() {
  const html = HtmlService.createHtmlOutputFromFile('sidebar')
    .setTitle('Gemini Nano Demo');
  
  // --- PROPOSED NEW METHOD ---
  // This would instruct Google to add the attribute to the iframe.
  html.setIframeAllow('language-model'); 
  
  DocumentApp.getUi().showSidebar(html);
}

`sidebar.html` (Client-side code)

<!DOCTYPE html>
<html>
  <head>
    <base target="_top">
  </head>
  <body>
    <h3>Gemini Nano Prompt Demo</h3>
    <p>Model Availability: <code id="availability">checking...</code></p>
    <button onclick="run()">Run Prompt</button>
    <pre id="output"></pre>

    <script>
      const availabilityElement = document.getElementById('availability');
      
      // Check for availability on load
      (async () => {
        if (window.ai) {
          const availability = await window.ai.canCreateTextSession();
          availabilityElement.textContent = availability;
        } else {
           availabilityElement.textContent = 'Not supported';
        }
      })();

      async function run() {
        // ... function to run the prompt ...
        // This would fail or succeed based on the iframe permission.
      }
    </script>
  </body>
</html>

Output

Current Output (The Problem)

When the sidebar is rendered, the generated <iframe> has an extensive allow attribute but lacks the necessary language-model value.

<iframe src="https://n-....googleusercontent.com/userCodeAppPanel"
  allow="accelerometer *; camera *; geolocation *; microphone *; ... ; web-share *">
</iframe>

This results in a check for the model's availability returning unavailable.

Expected Output (The Solution)

With the proposed .setIframeAllow('language-model') method, the language-model value would be appended to the existing allow attribute list.

<iframe src="https://n-....googleusercontent.com/userCodeAppPanel"
  allow="accelerometer *; camera *; ... ; web-share *; language-model *">
</iframe>

This would grant the necessary permission, and a check for the model's availability would return readily.

Steps

These are the steps to reproduce the current limitation:

1. Create a new Apps Script project bound to a Google Doc, Sheet or Slides.
2. Add the Code.gs and sidebar.html files with the content from the code sections above.
3. Save the project and reload the host document to see the "Nano AI" menu.
4. Open the sidebar via Nano AI > Show Demo.
5. Observe that "Model Availability" in the sidebar shows unavailable. This confirms the permission issue.

mhawksey/ticket.md