openclaw - 💡(How to fix) Fix [Bug]: Image tool and infer image describe fail with "chat content is empty" error - MiniMax image understanding broken [3 pull requests]

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

Image tool and infer image describe fail with "chat content is empty" error - Basic image understanding broken since update The 'image' tool used for image understanding/description has stopped working after the OpenClaw update (around late April/early May 2026). This is a BASIC FUNCTIONALITY that was working before the update and is critical for daily use.

Before (working):

• The image tool worked correctly until around April 30, 2026
• Could successfully describe images from URLs using the image tool
• Could use 'openclaw infer image describe' CLI command

After (broken):

• Since the update, image tool returns error: "chat content is empty (2013)"
• The CLI 'infer image describe' command also fails but misinterprets URLs as local file paths

Error messages received:

1. Via image tool: "Image model failed (minimax-cn/MiniMax-M2.5): 400 {"type":"error","error":{"type":"invalid_request_error","message":"invalid params, chat content is empty (2013)"},"request_id":"..."}"
2. Via infer CLI: "Error: No description returned for image: C:\Users\LinGao.openclaw\workspace\https:\httpbin.org\image\png" (URL incorrectly treated as file path)

Steps to reproduce

1.Call the image tool with any image URL: image(image="https://httpbin.org/image/png", prompt="Describe this image")2. 2.OR run CLI: openclaw infer image describe --file https://httpbin.org/image/png 3.Both result in errors - no image description is returned

Expected behavior

The image tool should successfully download the image and pass it to the vision model for description At minimum, the vision model (MiniMax-M2.5) should receive the image data and return a description

Actual behavior

Error "chat content is empty" indicates the image data is NOT being passed to the model API at all This suggests a regression in the media fetching or image-to-model pipeline

OpenClaw version

2026.5.12

Operating system

Windows 10

Install method

No response

Model

MiniMax-M2.5 (configured with "input": ["text", "image"])

Provider / routing chain

openclaw infer image describe → MiniMax API (MiniMax-M2.5 vision model)

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

URGENT - This is a basic, core functionality. Image understanding is essential for tasks like:

Reading invoices/receipts from images Describing screenshots Processing user-submitted images Many daily workflows depend on this feature.

Please fix ASAP. This regression blocks critical daily operations.

Additional information

No response