I've tried various URL, here two examples:

• https://i.ytimg.com/an_webp/gusXBTyg1O4/mqdefault_6s.webp?du=3000&sqp=CNjCn5YG&rs=AOn4CLBNM0alsGkVAM7rKJ5CydtKcVY2kQ
  • suggestions.key: "Too Simple"
  • Status Code: 400
• https://tineye.com/result_json
  • suggestions.key: "Download Error"
  • Status Code: 422

Looking at tineye PIP package, it seems the result is always a JSON: https://github.com/TinEye/pytineye/blob/d51c637a4b3e9b50b47739b540ab908243bf4f91/pytineye/api.py#L277 (Note: the entry point is different, we can't be sure)

So, the following code deals with more situation:

    if resp.is_error:
        # handle error codes from Tineye
        if resp.status_code in (400, 422):
            error = json_data.get('error')
            s_key = json_data.get('suggestions', {}).get('key', '')
            if s_key.startswith("Invalid image URL"):
                message = format_not_supported
            elif s_key:
                message = "TinEye:" + s_key
            else:
                message =  "TinEye:" + error
            results.append({'answer': message})
            return results
        resp.raise_for_status()

I feel we divert the usage of the answers or there are some missing features: looking at the code, we need to specify the engine name because the answers don't provide the engine.
The result is more an error, except we don't provide a way to return that.

For now:

• I would just return an empty result
• open an issue to allow engine to return usage error.
• once implemented, we can update tineye.py

Out of topic: when I've search for the query https://tineye.com/result_json python, tineye got the query and returned an error. We can avoid this if we

• move tineye to the images category.
• match only full URL without space.
• provide a category for URL search: !url https://example.com

For now:

• I would just return an empty result
• open an issue to allow engine to return usage error.

first lets merge #1412

• once implemented, we can update tineye.py

OK

• provide a category for URL search: !url https://example.com

No need for, it is a online_url_search engine:

when I've search for the query https://tineye.com/result_json python, tineye got the query and returned an error.

The https://tineye.com/result_json URL is responsible for the error / its not an image URL.