Description

Two of the four sub-issues NVB 6187477 (GitHub Issue #3727) reported against quickstart.html have not been addressed in the v0.0.53 .mdx rewrite. The other two (Issue #1 wizard prompt-order narrative and Issue #4 :11435 ollama-auth-proxy explanation) were fixed and NVB 6187477 is now closed-verified. This bug tracks the two remaining sub-issues so the original can stay closed.

Sub-issue A — NVIDIA Endpoints curated model list still stale

docs/get-started/quickstart.mdx line 111 still lists exactly the five models the original NVB 6187477 flagged:

• Nemotron 3 Super 120B
• GLM-5
• MiniMax M2.7
• GPT-OSS 120B
• DeepSeek V4 Pro

The wizard actually offers seven curated entries plus Other. The two missing from the doc:

• Nemotron 3 Nano Omni 30B (nvidia/nemotron-3-nano-omni-30b-a3b-reasoning)
• Kimi K2.6 (moonshotai/kimi-k2.6)

The Gemini list at line 193 IS refreshed with 3.x preview entries — that side is in good shape. NVIDIA Endpoints side has not been refreshed since the v0.0.44 report.

Sub-issue B — Option 7 example block missing dynamic running/suggested suffix

quickstart.mdx lines 78-86 show the literal:

7) Local Ollama (localhost:11434)

When Ollama is actually running on the host, the wizard prints:

7) Local Ollama (localhost:11434) — running (suggested)

The literal example block does not mirror what the user sees during onboard. The narrative at line 92 (Local Ollama appears when NemoClaw detects a usable local Ollama path or can offer an install or start action for your platform.) describes the dynamic detection, but the literal block hides the suffix.

Environment

Device:        MacBook Pro (Apple M4)
OS:            macOS 26.x (Darwin 25.1.0, arm64)
Architecture:  arm64
Node.js:       v23.10.0
npm:           11.3.0
Docker:        29.2.1 (Colima)
OpenShell CLI: 0.0.44
NemoClaw:      v0.0.53
OpenClaw:      2026.5.22
Doc file:      docs/get-started/quickstart.mdx (file extension migrated .html/.md to .mdx since the original report)

Steps to Reproduce

1. View docs/get-started/quickstart.mdx (or the published HTML).
2. Compare the NVIDIA Endpoints model list at line 111 against the actual wizard catalog when running nemoclaw onboard.
3. Compare the Option 7 example block (lines 78-86) against the wizard output when ollama is running locally.

Expected Result

Sub-issue A — Refresh the NVIDIA Endpoints curated model list to mirror the wizard's actual offerings, OR drop the literal list in favor of a "consult build.nvidia.com" link so the doc cannot drift again on the next catalog update.

Sub-issue B — Update the Option 7 literal example to show the dynamic — running (suggested) suffix, OR add an annotation immediately after the literal block clarifying that the suffix appears when ollama is detected locally.

Actual Result

Both sub-issues persist verbatim from the v0.0.44 report on NVB 6187477. See the 2026-05-28 retest comment on NVB 6187477 for full doc grep evidence.

Logs

Doc grep evidence on macOS v0.0.53 (2026-05-28):

$ sed -n '78,92p' /Users/lynnh/.nemoclaw/source/docs/get-started/quickstart.mdx
... (literal Option 7 line still static, no — running (suggested) suffix) ...

$ grep -n "Nemotron\|GLM\|MiniMax\|GPT-OSS\|DeepSeek\|Kimi K2" /Users/lynnh/.nemoclaw/source/docs/get-started/quickstart.mdx
111:3. At the `Choose model [1]:` prompt, pick a curated model from the list
     (for example, `Nemotron 3 Super 120B`, `GLM-5`, `MiniMax M2.7`,
     `GPT-OSS 120B`, or `DeepSeek V4 Pro`), ...
(Kimi K2.6 and Nemotron 3 Nano Omni 30B absent)

Related

Follow-up of NVB 6187477 ([macOS][Docs][GitHub Issue #3727] quickstart.html onboard wizard description out of sync with v0.0.44).

Original bug had 4 sub-issues; Issue #1 (prompt-order narrative) and Issue #4 (:11435 ollama-auth-proxy explanation) are fixed in v0.0.53; Issue #2 (sub-issue A here) and Issue #3 (sub-issue B here) remain.

NVB#6235152