Summary

The /model picker currently renders one row per custom_providers entry. When several entries share the same provider name (e.g. four ollama-cloud entries for qwen3-coder, glm-5.1, kimi-k2, minimax-m2.7), users see four separate ""Ollama Cloud"" rows in the picker, which is confusing UX — there is only one Ollama Cloud provider, so there should be one row containing four models.

This PR groups custom_providers entries that share the same provider name into a single picker row, while keeping entries with distinct names as separate rows.

Before

```
custom_providers:

• name: Ollama Cloud
  base_url: https://ollama.com/v1
  model: qwen3-coder:480b-cloud
• name: Ollama Cloud
  base_url: https://ollama.com/v1
  model: glm-5.1:cloud
• name: Ollama Cloud
  base_url: https://ollama.com/v1
  model: kimi-k2.5
• name: Ollama Cloud
  base_url: https://ollama.com/v1
  model: minimax-m2.7:cloud
• name: Moonshot
  base_url: https://api.moonshot.ai/v1
  model: kimi-k2-thinking
  ```

picker:

```

1. Ollama Cloud qwen3-coder:480b-cloud
2. Ollama Cloud glm-5.1:cloud
3. Ollama Cloud kimi-k2.5
4. Ollama Cloud minimax-m2.7:cloud
5. Moonshot kimi-k2-thinking
   ```

After

```

1. Ollama Cloud qwen3-coder:480b-cloud, glm-5.1:cloud, kimi-k2.5, minimax-m2.7:cloud
2. Moonshot kimi-k2-thinking
   ```

Each provider gets one row, and the picker drills into the model list when the user selects it — same as the curated providers (Anthropic, OpenAI, etc.) already do.

Implementation

Replaces the single-pass loop in list_authenticated_providers() (hermes_cli/model_switch.py) with a two-pass approach:

1. First pass: build an OrderedDict keyed by custom_provider_slug(name), accumulating models per group while preserving discovery order.
2. Second pass: iterate the groups and append one result row per group, skipping any slug that already appeared in an earlier provider source (the existing seen_slugs guard is preserved).

Insertion order is preserved via OrderedDict, so providers and their models still appear in the order the user listed them in custom_providers. No new dependencies.

What's preserved

• seen_slugs skip logic — providers already registered by an earlier source (curated providers, user-defined endpoints) still take precedence over custom_providers entries with the same slug.
• The not display_name or not api_url skip — entries missing either field are still ignored.
• All field parsing — base_url / url / api precedence is unchanged.
• The sort order — current provider first, then by model count descending still applies, so the user's active provider always floats to the top and the grouped row's higher total_models correctly reflects the new behaviour.

Files

• hermes_cli/model_switch.py — single function (list_authenticated_providers), 25 inserted / 10 removed.

Test plan

• Manual run of /model against a config with multiple Ollama Cloud entries shows one grouped row instead of four duplicates.
• Distinct provider names (Ollama Cloud + Moonshot) still produce separate rows.
• Insertion order is preserved across both passes.
• Existing seen_slugs precedence still applies — custom_providers entries do not override curated providers with the same slug.
• Syntax + import smoke test (ast.parse + from hermes_cli.model_switch import list_authenticated_providers).
• Happy to add a unit test against list_authenticated_providers if reviewers want one before merge — the function is pure (config in, list out), so a test against a synthetic config dict is straightforward.

Related PRs

This builds naturally on top of #7261 (custom_providers list schema support) and #7267 (hide providers with no curated LLM models). Both are still open. This PR does not depend on either being merged first — it stands alone against current main.

🤖 Generated with Claude Code