fix(ollama): add nemotron-3-ultra to model selector

[htazq]

Description

The Ollama model selector includes nemotron-3-nano:30b and nemotron-3-super, but does not include nemotron-3-ultra.

nemotron-3-ultra can already be used when it is configured as the current/custom model, so this appears to be a preset list omission rather than a provider/runtime issue.

Expected behavior

nemotron-3-ultra should be available in the Ollama default model selector, near the existing Nemotron entries.

Actual behavior

The model selector does not show nemotron-3-ultra, so users have to enter it manually as a custom model.

Proposed change

Add nemotron-3-ultra to the Ollama model preset list.

Scope

This only updates the selector/model preset list. It does not change Ollama provider behavior, routing, authentication, or request formatting.

Tests

• uv run --extra dev ruff check hermes_cli\models.py tests\hermes_cli\test_ollama_cloud_provider.py
• uv run --extra dev python -m pytest -o addopts="" tests\hermes_cli\test_ollama_cloud_provider.py

Note: the repository pytest config uses --timeout-method=signal, which raises AttributeError: module 'signal' has no attribute 'SIGALRM' on Windows. The targeted pytest run overrides addopts for the local Windows validation run.

[tonydwb]

Code Review Summary

Verdict: Approved

Looks Good

• Adds nemotron-3-ultra to Ollama cloud model selector, positioned after its sibling anchor models (nemotron-3-nano:30b, nemotron-3-super) in the merged list.
• Includes dedicated test case test_adds_curated_nemotron_ultra_when_registry_omits_it.
• No security concerns, no debug artifacts.

---

Reviewed by Hermes Agent

[teknium1]

Thanks for the focused selector fix. I found one current-main cache-path gap that needs resolution before this static fallback can provide the requested guarantee.

Problems

• fetch_ollama_cloud_models() returns a fresh cache before the PR's merge point (hermes_cli/models.py:3732-3736) and returns stale cache on total discovery failure (hermes_cli/models.py:3775-3778). A cache created without this model will bypass the proposed insertion; the new test uses force_refresh=True and does not cover either path.
• The current Ollama Cloud contract says the catalog is dynamic with “no static list to maintain” (tests/hermes_cli/test_ollama_cloud_provider.py:105-110). The PR should establish why this source-specific static exception is necessary.

Suggested changes

• Route cached and freshly discovered results through one idempotent curation step, with cache-path coverage, if a curated exception is intended.
• Document the verified upstream-source gap or update the discovery source instead of maintaining a local exception.

This is an automated hermes-sweeper review.

[teknium1]

This insertion is skipped when fetch_ollama_cloud_models() returns its fresh disk cache or stale-cache fallback first (current main: hermes_cli/models.py:3732-3736, 3775-3778). Apply an idempotent post-processing step to every return path, and add cache coverage, if this fallback is meant to guarantee picker availability.