Fixed :

Capture ChatGPT conversations using the newer logged-in and logged-out layouts, while keeping support for older layouts.
Preserve indentation and blank lines inside fenced code blocks during capture and Markdown export.
Capture ChatGPT's newer code-block containers and language labels.
Attach handoff files to ChatGPT's unrestricted file input instead of its camera input.
Preserve ordinary text before code blocks instead of mistaking it for a language label.
Preserve spaces between nested text elements, table rows, line breaks, and ordered lists.
Fall back to existing code sources when ChatGPT code wrappers differ, instead of silently exporting empty blocks.
Deduplicate nested message containers and retain mixed-layout and image-only messages.
Include images adjacent to their message inside the same ChatGPT turn/list item.
Render PDF code backgrounds before the code text, preserve inline-code identifiers, and support longer Markdown fences.
Keep short PDF code examples together and paginate long code/table content without hiding text.
Restrict ChatGPT capture to the visible active conversation; ignore hidden cached chats, sidebars, dialogs and unsent drafts.
Reject page/DOM changes during capture and mismatched source URLs instead of saving a mixed conversation.
Pin capture to the chosen tab's top frame, bound injection retries, and prevent duplicate saves/receiver listeners.
Record the capture's source URL in new saved entries and Markdown/PDF exports.

What's new in v1.2.0:

• Added full support for DeepSeek (chat.deepseek.com) — capture threads and resume chats seamlessly.
• Added direct "Export to JSON" option in conversation details.
• Improved KaTeX math formula extraction and rendering in Markdown and PDF exports.
• Polished 2x2 export layout and updated platform logos.