Is there a high-level chat option?

[AnOpenSauceDev]

Is your feature request related to a problem? Please describe.
I'm currently working on a project that involves llama-cpp-python's high-level API, and i wanted to know if there was a high-level chat feature, as chat completion tends to produce an unintelligible mess when i try talking to it.

Describe the solution you'd like
It would be nice if i could get some documentation on how to properly chat with the ai, or some alternate way to do this, without making my codebase too messy.

Describe alternatives you've considered

• Running the model low-level (i couldn't find anything related to -ngl, which made me give up pretty quickly)
• running the chat completion model with a prompt, but that lead to it repeatedly affirming me that Moscow was the largest European city. Either that, or hallucinating quite a lot. (it also couldn't recall what i said either.)

Additional context
Model: vicuna 7B
GPU: RTX 3060ti, compiled with cuBLAS
I'm running this native on my Pop_OS machine.

[gjmulder]

Check out oobabooga/text-generation-webui which can use llama-cpp-python. I've had a lot of success with llama-cpp-telegram_bot using Vicuna and WizardLM GGML models @TheBloke has converted.

The key is to understand how the models you are using were trained and to ensure your prompt template conforms to the training template.

[AnOpenSauceDev]

I tried a few of the tricks they used, and got some good results. Thanks!