rollback: can't solve performance issues, rollback to lfc

fix: move to gpu when running

fix: load models on startup

fix: run faiss on gpu

feat: add disclaimer

fix: correct prompt / context combination

fix: change to placeholder

feat: add prefill

feat: accelerate retrieve fn

feat: use Zephyr

feat: improve description

feat: add bestter description

feat: change context slider default value to ensure best results

feat: test original chunksize of 200

feat: test original chunksize of 512

feat: widen kontext window

fix: switch back to llama due to lack of compatibility

fix: update to transformers >4.8 compatibility

fix: use synchronous generation

feat: test microsoft phi

fix: correct to the right import

feat: switch back to google flan

feat: add seperate faiss search field

feat/todo (#2)

fix: revert the UI changes

feat: add headings

feat: add more cleare UI

feat: add more cleare UI

fix: correct use of input accordion

fix: remove invalid argument

feat: rework ui

fix: use markdown boxes

fix: rename components

feat: further improve UI

feat: further improve UI

feat: further improve UI

feat: further improve UI

feat: change ui to markdown view

feat: add seperate context box

feat: add pdf links

fix: revert to last functional version

fix: revert to last functional version

fix: only use gpu when running

feat: improve metadata handling

feat: further optimize for gpu speed

feat: use nvidia llama with optional reasoning

feat: imporve sysprompt

feat: use microsoft phi-2

feat: decrease chunksize

feat: use tinyLlama