fix(shell): render text-only A2A fallbacks

[stablegenius49]

Summary

• convert text-only A2A task responses into a simple A2UI Text surface in the Lit shell client
• preserve normal A2UI data-part responses unchanged
• add regression tests for data-part passthrough and single/multi-text fallbacks

Fixes #585.

Why

The Lit shell client currently drops text parts and only forwards data parts into the A2UI processor. When an agent returns a text-only fallback message, the UI ends up blank even though the backend sent a useful error/fallback string.

Testing

• npm install in renderers/web_core
• npm run build:tsc in renderers/web_core
• npm install in renderers/markdown/markdown-it
• npm run build in renderers/markdown/markdown-it
• npm install in samples/client/lit/shell
• npm run build:tsc in samples/client/lit/shell
• node --test --enable-source-maps dist/tests/client.test.js

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[gemini-code-assist]

Code Review

This pull request introduces a helpful fallback mechanism to render text-only A2A responses, preventing a blank UI. The new convertA2APartsToMessages function correctly transforms text parts into A2UI components, and the logic is well-supported by new regression tests. I've provided one suggestion to refactor the new function to improve its maintainability by reducing some code duplication.

[gemini-code-assist]

The logic for handling single and multiple text messages is a bit repetitive. Both branches of the if statement construct and return a similar array of beginRendering and surfaceUpdate messages. You can reduce this duplication by first determining the rootId and components array based on the number of text messages, and then constructing the final message array once.

  let rootId: string;
  let components: v0_8.Types.ComponentInstance[];

  if (textMessages.length === 1) {
    rootId = textFallbackComponentId(0);
    components = [
      {
        id: rootId,
        component: {
          Text: {
            usageHint: "body" as const,
            text: { literalString: textMessages[0] },
          },
        },
      },
    ];
  } else {
    rootId = TEXT_FALLBACK_ROOT_ID;
    const childIds = textMessages.map((_, index) =>
      textFallbackComponentId(index)
    );
    components = [
      {
        id: TEXT_FALLBACK_ROOT_ID,
        component: {
          Column: {
            children: {
              explicitList: childIds,
            },
          },
        },
      },
      ...textMessages.map((text, index) => ({
        id: childIds[index],
        component: {
          Text: {
            usageHint: "body" as const,
            text: { literalString: text },
          },
        },
      })),
    ];
  }

  return [
    {
      beginRendering: {
        root: rootId,
        surfaceId: TEXT_FALLBACK_SURFACE_ID,
      },
    },
    {
      surfaceUpdate: {
        surfaceId: TEXT_FALLBACK_SURFACE_ID,
        components,
      },
    },
  ];