Support IBM Watsonx.ai LLM provider

[barvhaim]

Adds a support to IBM's Watsonx.ai as LLM provider and embeddings provider

.env

LLM_PROVIDER="watsonxai"
EMBEDDING_PROVIDER="watsonxai"

WATSONX_APIKEY="xxx"
WATSONX_URL="https://us-south.ml.cloud.ibm.com"
WATSONX_PROJECT_ID="xxx"
WATSONX_EMBEDDING_MODEL="intfloat/multilingual-e5-large"

FAST_LLM_MODEL="ibm/granite-13b-instruct-v2"
DEFAULT_LLM_MODEL="ibm/granite-13b-instruct-v2"
SMART_LLM_MODEL="meta-llama/llama-3-405b-instruct"

[Arieg419]

Thanks for adding this support!

@assafelovic I think having this as a configuration file in the configuration directory would be good.

Continuing previous discussions, we'd like configurations to be declarative and have everything in the config file except for API keys.

wdyt?

[assafelovic]

Hey @barvhaim can you remove this line? The idea is to not have any optional dependencies. If and when someone will use it then a manual install will be necessary. Thanks!