Anthropic: Fix message content flattening for grouped consecutive messages

[sbruel]

Pull Request Template

Summary

When consecutive messages from the same role were grouped together, multi-element content arrays (e.g., vision messages with image + text) would become nested arrays that failed Anthropic's API validation. The original flattening logic only handled single-element arrays, leaving multi-element arrays improperly nested.

The messages property would be something like

{
  "content": [
    [
      {
        "source": {
          "data": "...",
          "media_type": "image/png",
          "type": "base64"
        },
        "type": "image"
      },
      {
        "text": "describe this image",
        "type": "text"
      }
    ],
    "prompt text"
  ],
  "role": "user"
}

instead of

{
  "content": [
      {
        "source": {
          "data": "...",
          "media_type": "image/png",
          "type": "base64"
        },
        "type": "image"
      },
      {
        "text": "describe this image",
        "type": "text"
      },
      {
        "text": "prompt text",
        "type": "text"
      }
  ],
  "role": "user"
}

Screen.Recording.2025-06-26.at.14.07.24.mov

Error logs

• Fix "Input should be a valid dictionary or object" error when sending grouped messages with multi-element content arrays to Anthropic API.
• Improve message content flattening logic to handle both single and multi-element arrays correctly.
• Ensure raw strings are converted to proper {type: "text", text: "..."} objects.
• Maintain trailing whitespace trimming for final assistant messages in all content formats.

Change Type

• Bug fix (non-breaking change which fixes an issue)

Testing

The easiest way I could find to reproduce the bug is shown in the video above.

• Create a message using and Anthropic model.
• After the response is returned, fork the original prompt into a new conversation.
• Write another message so there are 2 consecutive user messages.

Test Configuration:

Checklist

Please delete any irrelevant options.

• My code adheres to this project's style guidelines
• I have performed a self-review of my own code
• I have commented in any complex areas of my code
• I have made pertinent documentation changes
• My changes do not introduce new warnings
• I have written tests demonstrating that my changes are effective or that my feature works
• Local unit tests pass with my changes
• Any changes dependent on mine have been merged and published in downstream modules.
• A pull request for updating the documentation has been submitted.

[danny-avila]

api/app/clients/AnthropicClient.js is no longer used after #8013