test(ci): drop unsupported NIM reasoningEffort knob

[cursor]

Closed as proven duplicate of canonical #881

Exact evidence before closure:

• test(ci): drop unsupported NIM reasoningEffort knob #885 exact head acf1f37e79dbc1f5df259583011962ce2edb82cf is a direct fast-forward child of test(ci): lock local OpenCode small_model to NVIDIA NIM #881 predecessor 8f3f1631a69c73b27100a35ca4296fc7ab3e933c (ahead_by=1, behind_by=0).
• The single successor commit removes the unsupported NIM reasoningEffort option, locks the local {env:NVIDIA_API_KEY} vs organization-secret boundary, and adds the associated doctoring/regressions.
• Canonical test(ci): lock local OpenCode small_model to NVIDIA NIM #881 branch cursor/bc-accd3e03-65c6-4684-b661-34e3744ec4ac-ac55 was fast-forwarded without force to this exact acf1f37e… head and refetched successfully.
• Therefore test(ci): drop unsupported NIM reasoningEffort knob #885 has no remaining unique semantic/file delta outside test(ci): lock local OpenCode small_model to NVIDIA NIM #881.

No checks/reviews transfer from this duplicate. #881 must regenerate exact-current-head CI/security/SBOM/supply-chain/review evidence, and #879 remains a separate divergent predecessor until its unique delta is adjudicated rather than guessed away.

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[cursor]

Review (head acf1f37e)

This is the NIM-contract landing vehicle. Prefer this over #879 (6ca06d3f) and #881 (8f3f1631). Do not merge more than one of those three.

Held on this head:

• enabled_providers is ["nvidia-nim"]
• {env:NVIDIA_API_KEY} at https://integrate.api.nvidia.com/v1
• small_model is nvidia-nim/meta/llama-3.3-70b-instruct
• Nemotron Super has no reasoningEffort
• leftover deny-list includes github-models, STRIX_GITHUB_MODELS_TOKEN, COPILOT_GITHUB_TOKEN, openai/gpt-5, openai/o3, openai/o4-mini, and models.github.ai
• config text also rejects {env:NVIDIA_NIM_API_KEY} so the local binding is not renamed to the org secret
• review-agent credential names and central scheduler routing stay unchanged
• doctoring is in docs/doctoring/opencode-nvidia-nim-contract.md with APA 7th NVIDIA citations

No Critical findings. Do not self-approve or merge until exact-head CI and a qualifying independent non-author last-push approval exist.

Do not mix Stem Lab (#884) or lock-in residual (#880) onto this PR.

  

_{Sent by Cursor Automation: Fix Issues}

[cursor]

Review (acf1f37e vs develop@acdbea63)

This PR is closed unmerged. The same head is the current landing on open #881. Do not merge #885.

Scope: 7 files, +162/−50. Local OpenCode stays NVIDIA NIM only. This head drops OpenAI-style reasoningEffort on nvidia/llama-3.3-nemotron-super-49b-v1.5, keeps {env:NVIDIA_API_KEY} → https://integrate.api.nvidia.com/v1, rejects {env:NVIDIA_NIM_API_KEY} inside opencode.jsonc, and leaves review-agent credentials in ContextualWisdomLab/.github.

CodeRabbit CLI agent auth timed out in this environment. Findings are from the CodeRabbit reviewer agent plus an exact-head config parse.

CodeRabbit raised 2 issues, both Info. Critical: 0. Warning: 0.

Held on this SHA:

• enabled_providers is ["nvidia-nim"] only
• default and small_model stay on nvidia-nim
• no options.reasoningEffort on Nemotron Super 49B v1.5
• leftover github-models / STRIX_GITHUB_MODELS_TOKEN / COPILOT_GITHUB_TOKEN / openai/o3 / openai/o4-mini strings are absent
• scheduler credential test no longer reads local opencode.jsonc

Non-blocking follow-ups belong on #881, not a third NIM PR: dedupe the two contract tests, and lock reasoningEffort against the whole opencode.jsonc text the same way leftover tokens are locked.

Do not restore COPILOT_GITHUB_TOKEN. Do not rename the local binding to {env:NVIDIA_NIM_API_KEY}. Do not mix Stem Lab onto this slice.

  

_{Sent by Cursor Automation: Fix Issues}