Migrate to llama-index 0.10.0+

[rwood-97]

Updating to new llama-index

Fixes #162

Notes:

• I've used legacy for the messages_to_prompt and completion_to_prompt import. Potentially we should update but I can't find the completion_to_prompt in the new llama-index. Maybe here or here is relevant.
• I'm returning Settings in setup_settings but I don't really know if I need to do this.

[rchan26]

thanks @rwood-97! I updated a few of the other libraries and I think this largely works now. There were a few issues with like setting up the tokenizers for each model and also your comment about using legacy code from the llama_index library which I've fixed.

as discussed on Slack, we'll port over that code as we can customise it for different models we might use and just in case llama-index drops this in the future. it would be good to have an easy way for us to pass in prompt templates and different ways to construct prompts, as I think that's actually why we weren't successful with using the Huggingface models previously as we just didn't configure them properly (I think)

can you have a last check over please and we can merge?

[rwood-97]

OSError: You are trying to access a gated repo.
Make sure to request access at https://huggingface.co/meta-llama/Llama-2-7b-chat-hf and pass a token having permission to this repo either by logging in with `huggingface-cli login` or by passing `token=<your_token>`.

Just logging this error incase we need to address it

Update: I needed to request acces, its now in review

[rwood-97]

Okay all good :)

[rchan26]

ah yeah, this is a good spot @rwood-97. I added a line in there to load in the tokenizer for Llama2 which needs requesting. I think we would need to maybe add a line to load in a environment variable for a HUGGINGFACE_TOKEN. we would need that also for the Huggingface model as well if the user ever wanted to use a gated model that required some sort of access first like Llama or Gemma. I'll make a quick issue