Code block start sequence (```) is recognized even if it is immediately followed by text, but not recognized as end sequence in the same scenario

What's happening?

Code block start sequence (```) is recognized if it is immediately followed by text, but not recognized as end sequence in the same scenario.

Here is an example output I get from Opus 4.6 to a question about image format on OpenAI API:

**Content block structure:** Images are embedded as content blocks within a message's `content` array (which must be a list, not a plain string, when images are included). The content block type is `ChatCompletionContentPartImage`, with fields `image_url` and `type`. The structure is:

```json
{
  "type": "image_url",
  "image_url": {
    "url": "data:image/png;base64,<BASE64_DATA>",
    "detail": "high"
  }
}
```The base64 image data is passed via a data URL format: `"data:image/jpeg;base64,<your_image_data>"`.

"" in "json" is recognized as code block start, but is not recognized in "```The base64" as code block end.

It would be preferable if either both cases are recognized as code block start/end, or neither is recognized as code block start/end, unless there is a compelling argument as to why one is recognized and the other is not.

Right now answers like this are not easy to read in Big-AGI UI due to non-code sections ending up being presented as code.

Where does this happen?

Big-AGI Pro (big-agi.com)

Impact on your workflow

Low - Minor inconvenience

Environment (if applicable)

No response

Additional context

No response