Recreated settings changes - Adds serveral options for llamacpp and ollama

[icsy7867]

Original PR here:
#1677

llama-cpp https://llama-cpp-python.readthedocs.io/en/latest/api-reference/
https://docs.llamaindex.ai/en/stable/examples/llm/llama_2_llama_cpp.html#

ollama - https://github.com/run-llama/llama_index/blob/eeb2a60387b8ae1994005ad0eebb672ee02074ff/llama-index-integrations/llms/llama-index-llms-ollama/llama_index/llms/ollama/base.py

No configurable changes. -
openailike - https://docs.llamaindex.ai/en/stable/examples/llm/localai.html#localai

Not sure about the model_kwargs. The value is references for openai, but I could not find documentation on what values were allowed.
openai - https://github.com/run-llama/llama_index/blob/eeb2a60387b8ae1994005ad0eebb672ee02074ff/llama-index-integrations/llms/llama-index-llms-openai/llama_index/llms/openai/base.py
https://docs.llamaindex.ai/en/stable/examples/llm/openai.html

For the text/description I used the values found here:
https://github.com/ollama/ollama/blob/main/docs/modelfile.md#valid-parameters-and-values

LlamaCPP, where it used the same K/V, had the same values. However my setup is currently using ollama, need some testing done for LlamaCPP.

I also added the temperature under the main llm.settings. This should allow the models that supports this value to be edited/changed.

[icsy7867]

Hmm small bug... num_predict: 128 doesnt do what I think it does. It tells llamaindex the maximum size of the response. So this should probably be set to -1 or -2 by default.

It is odd though, that the default says "128", but if you dont set that kwarg, you get a larger response.

[icsy7867]

Looking at the ollama code:
https://github.com/ollama/ollama/blob/f878e91070af750709f1b3195eeb9fbdcaad2bef/openai/openai.go#L174

	if r.MaxTokens != nil {
		options["num_predict"] = *r.MaxTokens
	}

It looks like the default is 128, unless you have max tokens set. Then it just makes the value the same as the max tokens. Alternatively. setting this to "Max new tokens" might make more sense.