Nitro raycast extension. Using LLM right from Raycast.
This is a shameless wrapper around Nitro, to allow easy use in Raycast.
Currently, this is just a placeholder until I can make it work. Behold, the wave of Llamas is coming to you soon!
Reference for logical flow: raycast/extensions | Add speedtest #302
- Using
@janhq/nitro-nodewhen it is published on NPM instead of hacking around the installation hooks. See: janhq/jan#1635 - Auto check and update latest release of
nitroor let user choose the version they want to use. - Shamelessly copy the user interface of OpenAI chat and Hugging Face chat. Sorry I'm not a designer so don't sue me for copying the UI from the other chat models.
- Let users choose models they want to use and download it automatically.
- Let users pick the
prompt_templateon the known list to try on the new model. If they accept theprompt_templateit will be used. - Let LLM (tiny-llama) helps guessing the best prompt template to use and highlight it to user.
- Choice of installation directory and model directory.
- Option to use hugging-chat-api instead of running models locally? Am I going to abuse their servers?!?
- Be really lazy and summon Llamas to work for me by utilizing Raycast.
- Don't abandon this project like other pet projects in the past (most even never leave local machine)
- Summon as many Lllamas as possible.
- Make my robot waifu become real!