added gemini stt and tts feature

[SanTech0927]

No description provided.

[dbrockmann]

Hey, thanks for the contribution! 🚀

Looked through it, and the code looks solid, but a few things:

1. The API for STT like this does not support streaming, so you correctly batch it. But do you also use it with joinly conversational, and if so, how? I just see that the latency with this is quite high, right?
2. I noticed that the google-generativeai has been marked deprecated since August. I think it would still be ok to merge for now, but it would probably be nice to migrate the implementation to use the maintained google-genai package (migration docs).
3. Few linting/typing errors. It's best to run uv run ruff format (does style/formatting), uv run ruff check --fix (checks linting rules/applies fixes), and uv run pyright (checks typing), and then fix the rest of the errors manually.

Happy to see what you think about it! If you don't have time to continue working on this, we can also find a way to merge it in. Just let me know 👍

[SanTech0927]

I’ve made the suggested changes in the code and updated it to use the newer versions accordingly. I’m currently working on integrating the live API of Gemini to improve the latency and overall performance.

Apologies for the delay — had my exams recently, so I got to it a bit late. Thanks for your patience! 🚀

[dbrockmann]

Awesome, thanks for your work 🥳

I also tested it, and it works, but yeah, latency is really hard. It's just that the Gemini API like this is not really made for low latency, right? Still happy to merge it in now, feel free to then open another PR when you work on live implementations 🚀

I just did a small tweak with the dependencies, as with this updated implementation it uses google-genai which was only pulled in by joinly-client (through pydantic-ai).