docs(llm): stop claiming NVIDIA_API_KEY is auto-read for nvidia_nim (NVBug 6622594)

[kheiss-uwzoo]

Summary

• Follow-up to docs(extraction): pass NVIDIA_API_KEY explicitly in Live RAG examples (NVBug 6622594) #2575 / NVBug 6622594 QA: Live RAG examples on main already pass api_key="os.environ/NVIDIA_API_KEY", but LiteLLMClient still claimed NVIDIA_API_KEY was read automatically.
• The class docstring now states that omitted api_key uses LiteLLM provider-native lookup, that nvidia_nim reads NVIDIA_NIM_API_KEY (not NVIDIA_API_KEY), and that os.environ/NVIDIA_API_KEY forwards the hosted-inference key.

Fixes the remaining API-docstring gap on NVBug 6622594.

26.08.1 released artifacts

• Tag 26.08.1 does not contain docs(extraction): pass NVIDIA_API_KEY explicitly in Live RAG examples (NVBug 6622594) #2575. There is no 26.08.1 maintenance branch to cherry-pick onto.
• docs.nvidia.com 26.8.1 publishes from main, so this docstring (and the already merged examples) land on the next docs publish.
• PyPI / packaged README at tag 26.08.1 stay on the older Live RAG examples until a later package cut. That is a release-train item, not this PR.

Test plan

• Confirm LiteLLMClient.__doc__ no longer lists NVIDIA_API_KEY as an auto-read variable
• Confirm the docstring names NVIDIA_NIM_API_KEY for nvidia_nim and os.environ/NVIDIA_API_KEY as the explicit forward
• Confirm the PR diff is only nemo_retriever/src/nemo_retriever/models/llm/clients/litellm.py

[greptile-apps]

Greptile Summary

Updates the LiteLLMClient class documentation to distinguish LiteLLM’s provider-native credential lookup from explicit environment-variable forwarding.

• Documents NVIDIA_NIM_API_KEY as the native variable for the default nvidia_nim provider.
• Explains how to forward NVIDIA_API_KEY explicitly using api_key="os.environ/NVIDIA_API_KEY".

Confidence Score: 5/5

The PR appears safe to merge because it only corrects credential documentation and introduces no runtime behavior changes.

The documented explicit environment reference is resolved and forwarded at call time, while an omitted key remains delegated to LiteLLM’s provider-native lookup.

Important Files Changed

Filename	Overview
nemo_retriever/src/nemo_retriever/models/llm/clients/litellm.py	The revised credential guidance is consistent with the client’s omitted-key and explicit environment-reference handling; no actionable issue was identified.

_{Reviews (1): Last reviewed commit: "docs(llm): stop claiming NVIDIA_API_KEY ..." | Re-trigger Greptile}