fix(fireworks): keep prices from a matching fallback rule for an unlisted model

An unlisted Fireworks model that only a fallback generalization rule prices
has no raw cost map row, so the empty-row check sent it to the size bucket
and dropped the rule's rates. The Fireworks lookup now keeps the model info
when the row or a matching rule declares a rate.

litellm/llms/fireworks_ai/cost_calculator.py

@@ -11,9 +11,13 @@ from litellm.constants import (
     FIREWORKS_AI_56_B_MOE,
     FIREWORKS_AI_176_B_MOE,
 )
-from litellm.litellm_core_utils.llm_cost_calc.utils import cost_map_key_prices_anything, generic_cost_per_token
+from litellm.litellm_core_utils.llm_cost_calc.utils import (
+    cost_map_entry_prices_anything,
+    cost_map_key_prices_anything,
+    generic_cost_per_token,
+)
 from litellm.types.utils import ModelInfo, Usage
-from litellm.utils import get_model_info
+from litellm.utils import get_model_info, unmapped_model_generalization
 
 
 # Extract the number of billion parameters from the model name
@@ -65,7 +69,13 @@ def _resolve_model_info(model: str) -> ModelInfo:
         model_info: Final = get_model_info(model=model, custom_llm_provider="fireworks_ai")
     except Exception:
         return _resolve_routed_model_info(model)
-    return model_info if cost_map_key_prices_anything(model_info["key"]) else _resolve_routed_model_info(model)
+    declares_rates: Final = cost_map_key_prices_anything(model_info["key"]) or _rule_declares_rates(model)
+    return model_info if declares_rates else _resolve_routed_model_info(model)
+
+
+def _rule_declares_rates(model: str) -> bool:
+    rule_info: Final = unmapped_model_generalization(model, "fireworks_ai")
+    return rule_info is not None and cost_map_entry_prices_anything(rule_info)
 
 
 def _resolve_routed_model_info(model: str) -> ModelInfo:

tests/unit/llms/fireworks_ai/test_fireworks_ai_cost_calculator.py

@@ -7,6 +7,10 @@ from typing import Final
 import pytest
 
 import litellm
+from litellm.litellm_core_utils.fallback_generalizations import (
+    get_fallback_generalization_rules,
+    set_fallback_generalizations,
+)
 from litellm.litellm_core_utils.llm_cost_calc.utils import (
     calculate_prompt_caching_savings,
     generic_cost_per_token,
@@ -365,6 +369,36 @@ def test_an_unknown_fireworks_model_still_falls_back_to_the_parameter_size_bucke
     assert prompt_cost > 0
 
 
+@pytest.fixture
+def priced_rule_for_prefixed_fireworks_names() -> Generator[None, None, None]:
+    previous_rules: Final = list(get_fallback_generalization_rules())
+    try:
+        set_fallback_generalizations(
+            [
+                {
+                    "name": "fireworks-sol-ruled-priced",
+                    "pattern": r"fireworks_ai/accounts/fireworks/models/sol-ruled-[a-z]+",
+                    "model_info": {"mode": "chat", "input_cost_per_token": 3.21e-05, "output_cost_per_token": 7.65e-05},
+                }
+            ]
+        )
+        yield
+    finally:
+        set_fallback_generalizations(previous_rules)
+
+
+def test_a_model_priced_only_by_a_fallback_rule_bills_at_the_rule_rates(
+    priced_rule_for_prefixed_fireworks_names: None,
+):
+    prompt_cost, completion_cost = cost_per_token(
+        model="accounts/fireworks/models/sol-ruled-max",
+        usage=_usage(prompt_tokens=100, cached_tokens=0, completion_tokens=10),
+    )
+
+    assert prompt_cost == pytest.approx(100 * 3.21e-05)
+    assert completion_cost == pytest.approx(10 * 7.65e-05)
+
+
 _TIERED_INPUT_PATTERN: Final = re.compile(r"^input_cost_per_token_above_(\d+)k_tokens$")