PR #11681: fix(model-picker): expose full custom provider model maps (#11677)

• Repository: NousResearch/hermes-agent
• Author: briandevans
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/11681

Description (problem / solution / changelog)

What does this PR do?

Fixes #11677: named custom_providers: entries with a populated models: map only surfaced the single default model in the shared /model picker. After this change, the picker exposes the full configured model set for those saved custom providers in both the CLI and gateway fallback list.

The root cause was in hermes_cli/model_switch.py: the custom_providers grouping path only appended entry[\"model\"] (currently around lines 1103-1105) and never iterated entry[\"models\"], even though Hermes already persists that map for saved custom providers.

This is the smallest safe fix because it only changes the read-only picker aggregation path. It does not change provider resolution, runtime routing, persistence, slug generation, or auth.

Related Issue

Fixes #11677

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• Teach list_authenticated_providers() to append model names from custom_providers[].models when building grouped rows for named custom providers.
• Preserve existing behavior for custom_providers[].model by keeping it first and deduping any overlap with the models map/list.
• Add a CLI regression test that proves a single named custom provider exposes its full models map.
• Add a gateway /model regression test that proves the fallback text list also includes the expanded model set.

Backward Compatibility / Precedence

This change only affects how the picker displays saved custom providers; it does not alter persisted config.

Narrow Scope — Why Only This Path

This PR only fixes the legacy/named custom_providers: aggregation branch in hermes_cli/model_switch.py, because that is the broken path reported in #11677.

I deliberately did not touch:

• providers: dict handling — it already has dedicated coverage in tests/hermes_cli/test_user_providers_model_switch.py and already exposes full model lists.
• runtime provider resolution — not part of the bug.
• setup/persistence flows — they already save the models map; the bug was that the picker ignored it.

If maintainers want broader cleanup across other provider shapes, that should be a separate follow-up.

How to Test

1. Configure a named custom_providers: entry with both model: and a populated models: map.
2. Run /model in the CLI or a gateway platform.
3. Confirm the provider row now lists the full configured model set instead of only the default model.

Validation

Targeted change validation:

1. source venv/bin/activate && python -m pytest tests/hermes_cli/test_model_switch_custom_providers.py -q
   • 6 passed
2. source venv/bin/activate && python -m pytest tests/gateway/test_model_command_custom_providers.py -q
   • 2 passed
3. source venv/bin/activate && python -m pytest tests/hermes_cli/test_user_providers_model_switch.py -q
   • 9 passed
4. git diff --check
   • passed
5. source venv/bin/activate && python -m py_compile hermes_cli/model_switch.py tests/hermes_cli/test_model_switch_custom_providers.py tests/gateway/test_model_command_custom_providers.py
   • passed

CI-aligned broad suite:

1. source venv/bin/activate && python -m pytest tests/ -q --ignore=tests/integration --ignore=tests/e2e --tb=short -n auto
   • local environment still has unrelated baseline failures outside this change.

Baseline-confirmed unrelated failures reproduced on clean origin/main:

• tests/gateway/test_matrix.py::TestMatrixUploadAndSend::test_upload_encrypted_room_uses_file_payload
• tests/hermes_cli/test_gateway_wsl.py::TestSupportsSystemdServicesWSL::test_wsl_with_systemd
• tests/hermes_cli/test_gateway_wsl.py::TestSupportsSystemdServicesWSL::test_native_linux
• tests/gateway/test_approve_deny_commands.py::TestBlockingApprovalE2E::test_blocking_approval_approve_once
• tests/gateway/test_approve_deny_commands.py::TestBlockingApprovalE2E::test_blocking_approval_deny
• tests/tools/test_file_staleness.py::TestStalenessCheck::test_warning_when_file_modified_externally
• tests/tools/test_file_staleness.py::TestPatchStaleness::test_patch_warns_on_stale_file
• tests/tools/test_approval_heartbeat.py::TestApprovalHeartbeat::test_heartbeat_fires_while_waiting_for_approval
• tests/tools/test_approval_heartbeat.py::TestApprovalHeartbeat::test_wait_returns_immediately_on_user_response

One additional failure, tests/gateway/test_internal_event_bypass_pairing.py::test_non_internal_event_without_user_triggers_pairing, appeared during the full-suite run but passed in isolation on clean origin/main, so it looks like an order-dependent baseline flake rather than something introduced here.

Tested platform:

• macOS (local darwin arm64 dev environment)

Adjacent surfaces checked:

• providers:-dict picker coverage in tests/hermes_cli/test_user_providers_model_switch.py
• gateway /model fallback rendering in gateway/run.py

No standalone lint command is defined; CI runs python -m pytest tests/ -q --ignore=tests/integration --ignore=tests/e2e --tb=short -n auto.

Because this touches hermes_cli/**, the Nix workflow should also run on the PR.

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 pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS (darwin arm64)

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Changed files

• hermes_cli/model_switch.py (modified, +13/-0)
• tests/gateway/test_model_command_custom_providers.py (modified, +45/-0)
• tests/hermes_cli/test_model_switch_custom_providers.py (modified, +33/-0)

PR #11696: fix: resolve duplicate custom provider (e.g., ollama) in /model picker and add live model discovery

• Repository: NousResearch/hermes-agent
• Author: arsaboo
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/11696

Description (problem / solution / changelog)

Summary

Right now, the custom provider gets duplicated and /model picker does not pick all the models in the custom provider:

Fixes: #11677, #10011

╭─ ⚙ Model Picker — Select Provider ────────────────────╮
│                                                       │
│ Current: gemini-3-flash-preview on GitHub Copilot     │
│                                                       │
│   GitHub Copilot (14 models)  ← current               │
│   OpenRouter (31 models)                              │
│   Anthropic (8 models)                                │
│   OpenAI Codex (7 models)                             │
│ ❯ Ollama Local (1 model)                              │
│   OpenAI (0 models)                                   │
│   Ollama Local (0 models)                             │
│   Cancel                                              │
│                                                       │
╰───────────────────────────────────────────────────────╯

• Fix duplicate entries in the /model TUI picker for named custom providers configured using base_url in the providers: dict — section 3 of list_authenticated_providers was creating a phantom zero-model entry because it only reads api/url keys, while section 4 correctly handles base_url
• Add live /v1/models probing for named custom providers in the /model picker — selecting a custom provider now shows all models the server has loaded, not just the one configured in config.yaml
• Document the providers: dict format alongside the legacy custom_providers: list, and clarify live model discovery behavior

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

Changes Made

• hermes_cli/model_switch.py: Guard in section 3 to skip entries with no resolvable URL; probe each custom provider group's endpoint via fetch_api_models in section 4
• cli.py: Use fetch_api_models directly for user-defined providers in the model picker selection stage instead of provider_model_ids (which doesn't handle custom:* slugs)
• tests/hermes_cli/test_model_switch_custom_providers.py: Mock fetch_api_models to prevent live network calls in tests
• website/docs/integrations/providers.md: Document both config formats, add live discovery tip
• website/docs/reference/slash-commands.md: Update /model description

Changed files

• cli.py (modified, +4/-4)
• hermes_cli/model_switch.py (modified, +22/-0)
• tests/hermes_cli/test_model_switch_custom_providers.py (modified, +3/-0)
• website/docs/integrations/providers.md (modified, +32/-8)
• website/docs/reference/slash-commands.md (modified, +2/-2)