Tried it this morning - it works with openrouter, feels a little less verbose than antigravity CLI which should help with cost.
I would be happier if it wasn't javascript-based, but it's open source so it should be doable to use the models with things like https://github.com/charmbracelet/crush and
optimize a little how these alternatives work with deepseek models.
For prototyping the web interface is OK, but I don't want to rely on javascript for anything that isn't sandboxed or an isolated lab environment. As someone who lives in the terminal the web interface doesn't feel as nice to use as things like aichat, crush, or antigravity-cli.
I'll have to see if it handles remote ssh operations as well as antigravity. Some models seem to really struggle to remember where things are (locally vs. remotely) even though the context is recent. The one thing preventing me from self-hosting is the VRAM requirement for larger context sizes.
I see that people comment elsewhere that they're unsure what gemini 3.7 flash is for - the large context size makes it a good fit for agentic workflows where the agent must remember the whole conversation not to make a mistake. Anything below 128K seems very limited.
Seeing more diversity on the large context size models is good. The rising API prices are what makes me want to look at efficient self-hosting as soon as possible.
Tried it this morning - it works with openrouter, feels a little less verbose than antigravity CLI which should help with cost. I would be happier if it wasn't javascript-based, but it's open source so it should be doable to use the models with things like https://github.com/charmbracelet/crush and optimize a little how these alternatives work with deepseek models.
For prototyping the web interface is OK, but I don't want to rely on javascript for anything that isn't sandboxed or an isolated lab environment. As someone who lives in the terminal the web interface doesn't feel as nice to use as things like aichat, crush, or antigravity-cli.
I'll have to see if it handles remote ssh operations as well as antigravity. Some models seem to really struggle to remember where things are (locally vs. remotely) even though the context is recent. The one thing preventing me from self-hosting is the VRAM requirement for larger context sizes.
I see that people comment elsewhere that they're unsure what gemini 3.7 flash is for - the large context size makes it a good fit for agentic workflows where the agent must remember the whole conversation not to make a mistake. Anything below 128K seems very limited.
Seeing more diversity on the large context size models is good. The rising API prices are what makes me want to look at efficient self-hosting as soon as possible.