Vouch request: fede-kamel

[fede-kamel]

What do you want to work on?

I’d like to add OCI Generative AI support to OpenShell. I opened issue #3879 and prepared PR #3885 with a provider configuration, tests, and documentation for OCI’s OpenAI-compatible endpoints. This first change uses bearer authentication; OCI request signing and principal-based credentials are separate follow-ups for discussion with maintainers.

Why this change?

I work on OCI Generative AI integrations at Oracle and want to make it usable through OpenShell. This would let OCI users call supported models while keeping real credentials outside the sandbox. I tested the integration against live OCI endpoints, including streaming, tool calling, vision, embeddings, and the Responses API.

Checklist

• I wrote this request myself (not AI-generated)
• I have read CONTRIBUTING.md

[fede-kamel]

Prior contributions to NVIDIA repositories, for reference:

Repository / PR	State	Title
NVIDIA/NeMo-Relay#554	merged 2026-08-12	feat(codec): add OCI Generative AI provider support
NVIDIA/NeMo-Relay#1012	merged 2026-09-09	fix: keep Relay headers out of LangChain models that reject extra_headers
NVIDIA/NeMo-Agent-Toolkit#1804	merged 2026-04-30	Add OCI LangChain support for hosted Nemotron workflows
NVIDIA/NeMo-Agent-Toolkit#1856	merged 2026-04-09	fix(memmachein): preserve conversation message order in add_items
NVIDIA/nvidia-oci-samples#12	merged 2026-09-15	Add Nemotron 3.5 Lightning vLLM endpoint sample on OKE
NVIDIA/nvidia-oci-samples#17	open	Add agentic sample: OCI Generative AI agent observed by NeMo Relay
NVIDIA/dgx-spark-playbooks#104	open	Add playbook: Keeping vLLM Up
NVIDIA/dgx-spark-playbooks#105	open	fix: correct invalid shell, YAML, and TypeScript configs

OpenShell issues filed while preparing and testing this change: #3879 (design), #3880, #3881, #3882, #3883, #3884.

[johntmyers]

/vouch

[github-actions[bot]]

@fede-kamel has been vouched by @johntmyers.

You can now submit pull requests to OpenShell. Welcome aboard.

Please read CONTRIBUTING.md before submitting.

[fede-kamel]

Thanks again for the vouch, @johntmyers. I've opened four PRs since, each fixing an issue I filed and each probed live before and after on the released 0.1.2 build:

PR	Fixes	Type
#3923	#3884: approving a second policy proposal right after the first fails with a raw FAILED_PRECONDITION	gateway + CLI
#3924	#3880: no sandbox starts on WSL 2 + Docker Desktop with default settings	Docker driver
#3910	#3881: the Runtimes page TOML examples fail gateway preflight	docs
#3907	#3883: the OpenRouter provider smoke test hangs	docs

#3923 has been through gator review; the last finding is fixed and waiting on its re-check.

When someone has a moment, could a maintainer:

1. Comment /ok to test on the four PRs? Branch Checks and Helm Lint are still waiting on it.
2. Triage bug(docker): sandboxes never become Ready on WSL 2 + Docker Desktop (supervisor dials 127.0.0.1) #3880 and ux(cli): sequential rule approve fails with raw FAILED_PRECONDITION after proposals refresh #3884? fix(policy): refresh pending proposals when the sandbox policy changes #3923 and fix(docker): reach a WSL 2 gateway through Docker Desktop's host alias #3924 change behavior, so I expect they need those issues accepted.

The before/after probe results for the two behavior fixes are in the #3923 and #3924 threads.

Happy to adjust scope or split anything up.