claude-code - 💡(How to fix) Fix Image dimension error blocks ALL responses instead of gracefully degrading [1 participants]

When a conversation contains an image exceeding 2000px dimension limit (for multi-image requests), every subsequent response becomes the same error message, making the session completely unusable: The user cannot interact with Claude at all — every input, regardless of content, returns the same error. The session is effectively bricked until the user manually runs /compact or /clear. 3. At some point, every response becomes the dimension limit error 4. User input is completely ignored — even simple text like "hello" returns the same error

• Or: at minimum, the error should only affect the specific API call, not permanently brick the session — subsequent messages without image references should work normally The error blocks ALL API calls for the entire session. The user is trapped in a loop where no input produces any useful response. /compact works as a workaround but the user has to know to try it — and if /compact itself triggers the same API path, it may also fail.

TL;DR

Pre-process images to detect and either auto-downscale or remove those exceeding the 2000px dimension limit before sending the conversation to the API.

Guidance

• Implement image pre-processing to check for dimensions exceeding the limit and apply one of the suggested fixes (auto-downscale or remove) to prevent the error from blocking the entire session.
• Consider adding a notice or warning when an image is removed or downscaled, to inform the user of the action taken.
• Review the API call logic to ensure that errors are handled per-call rather than per-session, to prevent the session from becoming unusable after a single error.
• Test the pre-processing fix with various image sizes and types to ensure it works as expected and does not introduce new issues.

Example

# Pseudo-code example of image pre-processing
def preprocess_images(conversation):
    for image in conversation.images:
        if image.width > 2000 or image.height > 2000:
            # Either downscale the image
            image.downscale(2000, 2000)
            # Or remove it with a notice
            conversation.remove_image(image)
            conversation.add_notice("Image exceeded dimension limit and was removed.")
    return conversation

Notes

The suggested fix assumes that the image pre-processing can be implemented without significant performance impact. Additionally, the fix may need to be adapted based on the specific requirements and constraints of the Claude Code API and the conversation handling logic.

Recommendation

Apply the workaround of pre-processing images to detect and handle those exceeding the dimension limit, as this approach directly addresses the root cause of the issue and can be implemented without waiting for a potential fix in a future version.