fix(decisions): load on Python 3.10 and 3.11 and bill cached tokens once under custom pricing

Decisions IR dataclasses used a mappingproxy default, which is only hashable from Python 3.12, so import litellm failed on 3.10 and 3.11. Decisions usage now reaches cost_per_token as a Usage with prompt_tokens_details, so custom pricing no longer adds cached and cache-write tokens on top of input_tokens that already include them

litellm/cost_calculator.py

@@ -9,7 +9,7 @@ from typing import TYPE_CHECKING, Any, Final, Literal, cast
 
 from httpx import Response
 from pydantic import BaseModel
-from typing_extensions import ReadOnly, TypedDict
+from typing_extensions import ReadOnly, TypedDict, assert_never
 
 import litellm
 import litellm._logging
@@ -1034,6 +1034,8 @@ def get_usage_object(
         ),
     )
 
+    if isinstance(completion_response, (DecisionsResponse, OpenAIDecisionResponse)):
+        return None if completion_response.usage is None else _decisions_usage(completion_response.usage)
     if usage_obj is None:
         return None
     if isinstance(usage_obj, Usage):
@@ -1064,15 +1066,28 @@ def get_usage_object(
         return None
 
 
-def _decisions_usage(
+def _decisions_prompt_tokens_details(usage: DecisionsUsage | OpenAIDecisionUsage) -> PromptTokensDetailsWrapper:
-    *, input_tokens: int, output_tokens: int, cached_tokens: int, cache_write_tokens: int
+    match usage:
-) -> dict[str, object]:
+        case DecisionsUsage():
-    return {
+            return PromptTokensDetailsWrapper(
-        "prompt_tokens": input_tokens,
+                cached_tokens=usage.cached_tokens, cache_write_tokens=usage.cache_write_tokens
-        "completion_tokens": output_tokens,
+            )
-        "cache_read_input_tokens": cached_tokens,
+        case OpenAIDecisionUsage():
-        "cache_creation_input_tokens": cache_write_tokens,
+            return PromptTokensDetailsWrapper(
-    }
+                cached_tokens=usage.input_tokens_details.cached_tokens,
+                cache_write_tokens=usage.input_tokens_details.cache_write_tokens,
+            )
+        case _:
+            assert_never(usage)
+
+
+def _decisions_usage(usage: DecisionsUsage | OpenAIDecisionUsage) -> Usage:
+    return Usage(
+        prompt_tokens=usage.input_tokens,
+        completion_tokens=usage.output_tokens,
+        total_tokens=usage.input_tokens + usage.output_tokens,
+        prompt_tokens_details=_decisions_prompt_tokens_details(usage),
+    )
 
 
 def _is_known_usage_objects(usage_obj):
@@ -1493,20 +1508,8 @@ def completion_cost(
                             "usage",
                             litellm.Usage(**_usage_for_dump.model_dump()),
                         )
-                    if isinstance(usage_obj, DecisionsUsage):
+                    if isinstance(usage_obj, (DecisionsUsage, OpenAIDecisionUsage)):
-                        _usage = _decisions_usage(
+                        _usage = _decisions_usage(usage_obj).model_dump()
-                            input_tokens=usage_obj.input_tokens,
-                            output_tokens=usage_obj.output_tokens,
-                            cached_tokens=usage_obj.cached_tokens,
-                            cache_write_tokens=usage_obj.cache_write_tokens,
-                        )
-                    elif isinstance(usage_obj, OpenAIDecisionUsage):
-                        _usage = _decisions_usage(
-                            input_tokens=usage_obj.input_tokens,
-                            output_tokens=usage_obj.output_tokens,
-                            cached_tokens=usage_obj.input_tokens_details.cached_tokens,
-                            cache_write_tokens=usage_obj.input_tokens_details.cache_write_tokens,
-                        )
                     elif usage_obj is None:
                         _usage = {}
                     elif isinstance(usage_obj, BaseModel):

litellm/types/decisions.py

@@ -338,7 +338,7 @@ class DecisionsIRPredicateQuestion:
     name: str | None
     instructions: DecisionsJSON | None
     criteria: NoulCriteria | None = None
-    extra: Mapping[str, object] = field(default=_NO_EXTRA)
+    extra: Mapping[str, object] = field(default_factory=lambda: _NO_EXTRA)
 
 
 @dataclass(frozen=True, slots=True)
@@ -352,7 +352,7 @@ class DecisionsIRChoiceQuestion:
     name: str | None
     instructions: DecisionsJSON | None
     choices: tuple[DecisionsIRChoiceOption, ...]
-    extra: Mapping[str, object] = field(default=_NO_EXTRA)
+    extra: Mapping[str, object] = field(default_factory=lambda: _NO_EXTRA)
 
 
 @dataclass(frozen=True, slots=True)
@@ -366,7 +366,7 @@ class DecisionsIRScoreQuestion:
     name: str | None
     instructions: DecisionsJSON | None
     levels: tuple[DecisionsIRScoreLevel, ...]
-    extra: Mapping[str, object] = field(default=_NO_EXTRA)
+    extra: Mapping[str, object] = field(default_factory=lambda: _NO_EXTRA)
 
 
 DecisionsIRQuestion: TypeAlias = DecisionsIRPredicateQuestion | DecisionsIRChoiceQuestion | DecisionsIRScoreQuestion
@@ -382,7 +382,7 @@ class DecisionsIRRequest:
 @dataclass(frozen=True, slots=True)
 class DecisionsIRPredicateAnswer:
     probability: float
-    extra: Mapping[str, object] = field(default=_NO_EXTRA)
+    extra: Mapping[str, object] = field(default_factory=lambda: _NO_EXTRA)
 
 
 @dataclass(frozen=True, slots=True)
@@ -396,7 +396,7 @@ class DecisionsIRChoiceAnswer:
     choice: str | bool
     confidence: float
     probabilities: tuple[DecisionsIRChoiceProbability, ...]
-    extra: Mapping[str, object] = field(default=_NO_EXTRA)
+    extra: Mapping[str, object] = field(default_factory=lambda: _NO_EXTRA)
 
 
 @dataclass(frozen=True, slots=True)
@@ -411,7 +411,7 @@ class DecisionsIRScoreAnswer:
     score: float
     confidence: float
     probabilities: tuple[DecisionsIRScoreProbability, ...]
-    extra: Mapping[str, object] = field(default=_NO_EXTRA)
+    extra: Mapping[str, object] = field(default_factory=lambda: _NO_EXTRA)
 
 
 @dataclass(frozen=True, slots=True)
@@ -431,7 +431,7 @@ class DecisionsIRUsage:
     cached_tokens: int = 0
     cache_write_tokens: int = 0
     reasoning_tokens: int = 0
-    extra: Mapping[str, object] = field(default=_NO_EXTRA)
+    extra: Mapping[str, object] = field(default_factory=lambda: _NO_EXTRA)
 
 
 @dataclass(frozen=True, slots=True)
@@ -439,7 +439,7 @@ class DecisionsIRResponse:
     model: str | None
     answers: tuple[DecisionsIRAnswer, ...]
     usage: DecisionsIRUsage
-    extra: Mapping[str, object] = field(default=_NO_EXTRA)
+    extra: Mapping[str, object] = field(default_factory=lambda: _NO_EXTRA)
 
 
 @dataclass(frozen=True, slots=True)

tests/unit/decisions/test_main.py

@@ -285,6 +285,46 @@ def test_decisions_cost_uses_litellm_token_pricing() -> None:
     assert cost == pytest.approx(expected_cost)
 
 
+@pytest.mark.parametrize(
+    "response",
+    (
+        DecisionsResponse(
+            model="jev-latest",
+            answers={},
+            usage=DecisionsUsage(
+                input_tokens=_INPUT_TOKENS,
+                output_tokens=_OUTPUT_TOKENS,
+                cached_tokens=_CACHED_TOKENS,
+                cache_write_tokens=_CACHE_WRITE_TOKENS,
+            ),
+        ),
+        OpenAIDecisionResponse.model_validate(_OPENAI_RESPONSE),
+    ),
+    ids=("systemone", "openai"),
+)
+def test_custom_token_pricing_bills_cached_decisions_input_tokens_once(
+    response: DecisionsResponse | OpenAIDecisionResponse,
+) -> None:
+    cost: Final = litellm.completion_cost(
+        completion_response=response,
+        model="gpt-6-luna",
+        custom_llm_provider="openai",
+        custom_cost_per_token={
+            "input_cost_per_token": 1.0,
+            "output_cost_per_token": 2.0,
+            "cache_read_input_token_cost": 0.1,
+            "cache_creation_input_token_cost": 1.25,
+        },
+    )
+
+    assert cost == pytest.approx(
+        (_INPUT_TOKENS - _CACHED_TOKENS - _CACHE_WRITE_TOKENS) * 1.0
+        + _CACHED_TOKENS * 0.1
+        + _CACHE_WRITE_TOKENS * 1.25
+        + _OUTPUT_TOKENS * 2.0
+    )
+
+
 def test_decisions_response_hidden_params_getter_preserves_mutable_identity() -> None:
     response: Final = DecisionsResponse(model="decider", answers={}, usage=None)