fix(models): prioritize Ollama Cloud models

[farena-ai]

Summary

• Prioritizes Ollama Cloud models in model listing/selection behavior.
• Updates focused tests for Ollama Cloud provider ordering.

Files changed

• hermes_cli/models.py
• tests/hermes_cli/test_ollama_cloud_provider.py

Test commands/results

python3 -m py_compile hermes_cli/models.py tests/hermes_cli/test_ollama_cloud_provider.py
venv/bin/python -m pytest tests/hermes_cli/test_ollama_cloud_provider.py -q

Result:

46 passed

Live impact

• Affects Hermes CLI/runtime model listing and selection paths when running updated code.
• No gateway platform behavior changed.
• No running service was restarted or modified.

Restart/reload needed

• CLI use picks this up when run from the updated checkout/install.
• Existing long-running processes need restart/relaunch only if they rely on this model-listing code path.
• No gateway restart is specifically required unless the gateway process itself must refresh this code.

Known caveats

• Targeted Ollama Cloud tests passed.
• Optional broader sweep before merge:

venv/bin/python -m pytest tests/hermes_cli/test_models.py tests/hermes_cli/test_ollama_cloud_provider.py -q

[teknium1]

Thanks for improving the Ollama Cloud picker. The requested behavior is still absent on current main: fetch_ollama_cloud_models() preserves live/models.dev order at hermes_cli/models.py:3746-3761.

Problems

• The new priority call in this PR only handles a newly merged result. Current fresh-cache returns at hermes_cli/models.py:3720-3724 and stale-cache fallback at hermes_cli/models.py:3763-3766 return stored order verbatim, so existing caches and offline fallback do not receive the requested prioritization.

Suggested changes

• Apply the helper on both cache-return paths, or centralize it before every return from fetch_ollama_cloud_models().
• Add tests for fresh-cache and stale-cache ordering alongside the new merge test.

Automated hermes-sweeper review.

[teknium1]

This reorders only newly merged results. Fresh-cache and stale-cache paths return their stored lists verbatim on current main (hermes_cli/models.py:3720-3724, 3763-3766), so please apply the helper to those return paths too and cover them with tests.