fix(inventory): avoid fresh Nous tier checks in picker payloads

hermes_cli/inventory.py

@@ -116,6 +116,7 @@ def build_models_payload(
     canonical_order: bool = False,
     pricing: bool = False,
     capabilities: bool = False,
+    force_fresh_nous_tier: bool = False,
     max_models: int = 50,
 ) -> dict:
     """Build the ``{providers, model, provider}`` shape every consumer
@@ -139,6 +140,10 @@ def build_models_payload(
       ``{model: {fast, reasoning}}`` so pickers can gate the model-options
       controls (fast toggle / reasoning) to what each model actually
       supports, instead of offering knobs the backend would reject.
+    - ``force_fresh_nous_tier``: bypass the short Nous free-tier cache when
+      selecting Portal-recommended Nous models and applying tier gating. Keep
+      this false for UI picker opens; explicit auth/model flows can opt in
+      when they need freshly-purchased credits to show up immediately.
     """
     from hermes_cli.model_switch import list_authenticated_providers
 
@@ -148,6 +153,7 @@ def build_models_payload(
         current_model=ctx.current_model,
         user_providers=ctx.user_providers,
         custom_providers=ctx.custom_providers,
+        force_fresh_nous_tier=force_fresh_nous_tier,
         max_models=max_models,
     )
 
@@ -158,7 +164,7 @@ def build_models_payload(
     if canonical_order:
         rows = _reorder_canonical(rows)
     if pricing:
-        _apply_pricing(rows)
+        _apply_pricing(rows, force_fresh_nous_tier=force_fresh_nous_tier)
     if capabilities:
         _apply_capabilities(rows)
 
@@ -293,7 +299,11 @@ def _reorder_canonical(rows: list[dict]) -> list[dict]:
     return canon + extras
 
 
-def _apply_pricing(rows: list[dict]) -> None:
+def _apply_pricing(
+    rows: list[dict],
+    *,
+    force_fresh_nous_tier: bool = False,
+) -> None:
     """Enrich each provider row with per-model pricing + Nous tier gating.
 
     Mutates ``rows`` in-place. For every row whose provider supports live
@@ -359,7 +369,9 @@ def _apply_pricing(rows: list[dict]) -> None:
         if slug == "nous":
             try:
                 if nous_free_tier is None:
-                    nous_free_tier = check_nous_free_tier(force_fresh=True)
+                    nous_free_tier = check_nous_free_tier(
+                        force_fresh=force_fresh_nous_tier
+                    )
                 row["free_tier"] = bool(nous_free_tier)
                 if nous_free_tier:
                     _selectable, unavailable = partition_nous_models_by_tier(

hermes_cli/model_switch.py

@@ -1178,6 +1178,7 @@ def list_authenticated_providers(
     current_base_url: str = "",
     user_providers: dict = None,
     custom_providers: list | None = None,
+    force_fresh_nous_tier: bool = False,
     max_models: int = 8,
     current_model: str = "",
 ) -> List[dict]:
@@ -1197,6 +1198,9 @@ def list_authenticated_providers(
       - source: str — "built-in", "models.dev", "user-config"
 
     Only includes providers that have API keys set or are user-defined endpoints.
+    ``force_fresh_nous_tier`` bypasses the short Nous tier cache for explicit
+    account-sensitive flows. UI picker opens should leave it false so they do
+    not block on fresh Portal/account checks every time.
     """
     import os
     from agent.models_dev import (
@@ -1539,7 +1543,7 @@ def list_authenticated_providers(
                     _portal = _st.get("portal_base_url", "") or ""
                 except Exception:
                     _portal = ""
-                if _nous_free(force_fresh=True):
+                if _nous_free(force_fresh=force_fresh_nous_tier):
                     model_ids, _ = _union_free(model_ids, _pricing, _portal)
                 else:
                     model_ids, _ = _union_paid(model_ids, _pricing, _portal)

tests/hermes_cli/test_inventory.py

@@ -141,6 +141,18 @@ def _list_auth_returning(rows: list[dict]):
     )
 
 
+def _nous_row(model: str = "openai/gpt-5.5") -> dict:
+    return {
+        "slug": "nous",
+        "name": "Nous",
+        "models": [model],
+        "total_models": 1,
+        "is_current": True,
+        "is_user_defined": False,
+        "source": "built-in",
+    }
+
+
 def test_build_models_payload_returns_expected_shape():
     rows = [
         {"slug": "openrouter", "name": "OpenRouter", "models": ["m1"],
@@ -173,6 +185,80 @@ def test_build_models_payload_does_not_call_provider_model_ids():
     mock_pm.assert_not_called()
 
 
+def test_build_models_payload_uses_cached_nous_tier_by_default():
+    """Picker payloads should not force fresh Nous account checks.
+
+    Desktop/status picker opens are request/response UI paths. They can hit
+    the short free-tier cache; explicit model/auth flows can still opt into a
+    fresh account check when needed.
+    """
+    ctx = _empty_ctx(provider="nous", model="openai/gpt-5.5")
+    rows = [_nous_row()]
+    with patch(
+        "hermes_cli.model_switch.list_authenticated_providers",
+        return_value=rows,
+    ) as mock_list:
+        build_models_payload(ctx)
+
+    mock_list.assert_called_once()
+    assert mock_list.call_args.kwargs["force_fresh_nous_tier"] is False
+
+
+def test_build_models_payload_can_force_fresh_nous_tier():
+    ctx = _empty_ctx(provider="nous", model="openai/gpt-5.5")
+    rows = [_nous_row()]
+    with patch(
+        "hermes_cli.model_switch.list_authenticated_providers",
+        return_value=rows,
+    ) as mock_list:
+        build_models_payload(ctx, force_fresh_nous_tier=True)
+
+    mock_list.assert_called_once()
+    assert mock_list.call_args.kwargs["force_fresh_nous_tier"] is True
+
+
+def test_pricing_uses_cached_nous_tier_by_default():
+    rows = [_nous_row()]
+    ctx = _empty_ctx(provider="nous", model="openai/gpt-5.5")
+    with (
+        _list_auth_returning(rows),
+        patch(
+            "hermes_cli.models.get_pricing_for_provider",
+            return_value={
+                "openai/gpt-5.5": {
+                    "prompt": "0.000001",
+                    "completion": "0.000002",
+                },
+            },
+        ),
+        patch("hermes_cli.models.check_nous_free_tier", return_value=False) as mock_free,
+    ):
+        build_models_payload(ctx, pricing=True)
+
+    mock_free.assert_called_once_with(force_fresh=False)
+
+
+def test_pricing_can_force_fresh_nous_tier():
+    rows = [_nous_row()]
+    ctx = _empty_ctx(provider="nous", model="openai/gpt-5.5")
+    with (
+        _list_auth_returning(rows),
+        patch(
+            "hermes_cli.models.get_pricing_for_provider",
+            return_value={
+                "openai/gpt-5.5": {
+                    "prompt": "0.000001",
+                    "completion": "0.000002",
+                },
+            },
+        ),
+        patch("hermes_cli.models.check_nous_free_tier", return_value=False) as mock_free,
+    ):
+        build_models_payload(ctx, pricing=True, force_fresh_nous_tier=True)
+
+    mock_free.assert_called_once_with(force_fresh=True)
+
+
 def test_include_unconfigured_appends_canonical_skeletons():
     """include_unconfigured=True adds CANONICAL_PROVIDERS rows that
     list_authenticated_providers didn't emit. Skeleton rows have empty