Adding support for NVIDIA NIM

[praateekmahajan]

NVIDIA NIM provides an API to access several OSS models through one API.

The issue is intended to track adding a provider for NIM, so that we can expose it also in goose.

Examples

Here is an example of an endpoint that supports tools : https://docs.api.nvidia.com/nim/reference/meta-llama-3_1-8b-infer
Another example of langchain agents using LLAMA 3.1 https://github.com/langchain-ai/langchain-nvidia/blob/main/cookbook/nvidia_nim_agents_llama3.1.ipynb

Possible Issue

I haven't looked too closely but since NIM supports various models, the payload for each complete will also be different. Someone who has worked on similar APIs (e.g. Databricks Endpoints) or who has looked at how NVIDIAChat works in Langchain can perhaps offer ideas here.