Vision Support Update (0.0.4.5)

[wz-ml]

Feature update with image support and bugfixes! New additions:

• Support for .jpeg / .png image uploads on both prompt-on-the-fly and chatGPT popups (including multi-image support).
• A screenshot + snipping feature to make sending snips of your current webpage to chatGPT that much easier.
• Added keyboard shortcuts for image uploading & screenshots.

Major Bugfixes:

• Fixed a major bug introduced in the last update where opening a new chat popup while an existing one receives a stream causes the text stream to flow to the new one instead.
• Fixed a bug regarding text streams not stopping after creating a popup after a refresh.
• Promptless selection text now appears for both ChatGPT and prompt-on-the-fly interfaces.

Notes:

• The snipping feature fails when the current tab's a PDF! Chrome renders PDFs in an iframe, which Chrome's CaptureVisibleTab is unable to capture. Seems difficult to work around. This may be intentional behavior on Chrome's part as a security feature to prevent extensions from reading iframe data.
• I've done some manual bug-hunting, but I'm far from comprehensive. @giosilvi Let me know if any bugs appear! In particular, I haven't played around with the hotkeys, so testing on that would be nice.

[giosilvi]

Looks great so far!

[giosilvi]

I pushed a fix to the title of the popup, to make it less clutter let's just use the last user message.