fix: show LM Studio picker with base URL

[daveotero]

What does this PR do?

Makes the built-in lmstudio provider appear in the model picker when LM Studio is configured with a local base URL but no API key.

Current behavior on main: if a user sets LM_BASE_URL for local LM Studio but does not set LM_API_KEY, the built-in LM Studio provider can be omitted from the /model picker unless LM Studio is already the active provider or the auth-gated path is satisfied.

That is wrong for the common local LM Studio setup: LM Studio typically runs on localhost and does not require an API key.

Related Issue

No separate issue filed. This came from a live Hermes model-picker configuration issue on a local LM Studio setup.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✅ Tests (adding or improving test coverage)

Changes Made

• Update hermes_cli/model_switch.py so the built-in lmstudio provider is included in picker candidates when LM_BASE_URL is configured, even if LM_API_KEY is absent.
• Preserve existing behavior when:
  • LM Studio is already the active provider
  • LM Studio auth is configured
  • other provider credential gates are evaluated
• Add regression coverage in tests/hermes_cli/test_lmstudio_model_picker.py for the local no-auth LM Studio picker path.

Why this is the right scope

This is intentionally narrow: it does not add a new provider, change runtime routing, alter model resolution, or loosen auth requirements for remote providers.

It only fixes picker visibility for the existing built-in lmstudio provider when the existing local base URL configuration is enough to use it.

Typical affected configuration:

LM_BASE_URL=http://127.0.0.1:1234/v1
# no LM_API_KEY

Expected result: LM Studio appears in the model picker as an available local provider.

How to Test

Targeted regression test used before opening the PR:

/c/Users/daveotero/.hermes/hermes-agent/venv/Scripts/python.exe -m pytest tests/hermes_cli/test_lmstudio_model_picker.py -q -o 'addopts='

The test covers the bug directly: LM_BASE_URL is configured, LM_API_KEY is absent, and the picker still emits the built-in lmstudio provider row.

Platforms Tested

• Native Windows 10
• Python 3.11 Hermes environment
• Local LM Studio base URL configuration

Duplicate Search

Searched open and closed issues/PRs for LM Studio picker visibility with LM_BASE_URL / missing LM_API_KEY and did not find an existing PR or issue covering this fix.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run the targeted regression test and it passes
• I've added tests for my changes
• I've tested on my platform: native Windows 10

Documentation & Housekeeping

• Documentation update: N/A, bugfix only
• cli-config.yaml.example update: N/A, no config keys added or changed
• CONTRIBUTING.md / AGENTS.md update: N/A, no architecture/workflow change
• Cross-platform impact considered: yes; this only affects provider-picker candidate selection
• Tool descriptions/schemas update: N/A

[daveotero]

Updated the PR description to more explicitly match CONTRIBUTING.md / the PR template: current-main behavior, affected local LM Studio config, narrow bug scope, testing, platform tested, and duplicate-search note. This is still mergeable and intentionally limited to the model-picker visibility bugfix.

[teknium1]

Thanks for the focused regression fix. Current main still has the reported gap: hermes_cli/model_switch.py:1646-1667 probes LM Studio when LM_BASE_URL is present, but the subsequent provider-row gate at hermes_cli/model_switch.py:1795-1853 only recognizes LM_API_KEY or another credential source, so the row is skipped. The proposed condition addresses that exact divergence, and the added test exercises the interactive-picker wrapper.

Automated hermes-sweeper review.