Interest in Project 12 GSoC: Visualising diagrams + UI in Gemini CLI

[lin88225]

Hi @LyalinDotCom!

I'm Aisling, a CS student at Trinity College Dublin and a Generation Google Scholar. I’ve been following the Gemini CLI repo for a bit and Project 12 really jumped out at me. I’m a huge fan of the idea of making the terminal a bit more visual.

My background is mostly full-stack, and I recently built an NLP-based automation tool for a project with Millennium Management (mostly automating Jira workflows). Practically speaking, I've been digging into the rendering side of things, basically how to bridge Puppeteer/Sharp with Sixel or Kitty without making the CLI sluggish.

I’ll be interning at Amazon this summer, so I’m really focused on building high-performance, 'dev-first' tools. I saw the packages/cli/src/ui folder has most of the Ink logic, do you think that would be the best place to start poking around?

If there are any 'good first issues' or specific files I should look at to get a better feel for the TUI engine, I'd love to try a small PR before the application window opens.

Cheers,
Aisling

[Thankyou-Cheems]

Hi Aisling!

I'm Ruikang (Karl). I just posted my initial proposal for Idea 9 (Interactive Security Policy / Sandbox Wizard, with policy visualization) here: GSoC 2026: [Idea #9]Interactive Security Policy / Sandbox Wizard (with policy visualization)

I'm planning an ASCII-first permission dashboard (with richer rendering as stretch) as part of the sandbox-setup wizard, your work on Idea 12 (terminal diagrams, Sixel/Kitty, etc.) would be a perfect fit to make the review experience even better!

If either of us gets selected for Idea 9 or Idea 12, I’d actively reach out to collaborate on the overlapping visualization parts.

Wishing you the best with your application, I am excited to see more visual magic in the terminal! 🚀

Karl

[lin88225]

Hi Karl (@Thankyou-Cheems)!

I love the idea of collaborating. If I can get the Sixel/Kitty rendering engine solid for Project 12, it would be great to use it to turn your ASCII-first security dashboard into a visual map of safe zones and network boundaries.

Best of luck with your application, and I'm looking forward to seeing how our projects might overlap!

[aniruddhaadak80]

From my point of view, the rendering side gets much more practical if the first milestone stays narrow around one explicit visualize style command, one rendering pipeline, and one reliable fallback path. The hardest part is usually not drawing a diagram once, but keeping the output trustworthy and fast across very different terminal capabilities. Starting from the existing Ink UI code makes sense, but I would keep deterministic structure and ASCII fallback close to the center of the first version.

[lin88225]

I've been thinking about that constraint for sure. I plan to define an intermediate representation for the diagram structure first, so Sixel/Kitty are just richer renderers of the same IR rather than a separate code path. Basically, ASCII output becomes the contract, not the fallback. Capability detection (checking $TERM, $TERM_PROGRAM, Sixel probe) runs once at startup and can be cached so Puppeteer stays out of M1 entirely.

Curious whether you think the IR should live close to the Ink layer or be kept fully separate from it?