fix: python 3.10 typing compatibility in compact editor and claude 3 cache pricing (#38076, #38056)

litellm/llms/anthropic/experimental_pass_through/context_management/editors/compact.py

@@ -13,8 +13,14 @@ Mirrors Anthropic's native ``compact_20260112`` for non-Anthropic providers:
 """
 
 import re
+import sys
 from collections.abc import Mapping, Sequence
-from typing import TYPE_CHECKING, Any, Final, Literal, NotRequired, Optional, TypedDict, Union, cast
+from typing import TYPE_CHECKING, Any, Final, Literal, Optional, TypedDict, Union, cast
+
+if sys.version_info >= (3, 11):
+    from typing import NotRequired
+else:
+    from typing_extensions import NotRequired
 
 from typing_extensions import ReadOnly
 

litellm/model_prices_and_context_window_backup.json

@@ -12269,7 +12269,7 @@
     },
     "claude-3-haiku-20240307": {
         "cache_creation_input_token_cost": 3e-07,
-        "cache_creation_input_token_cost_above_1hr": 6e-06,
+        "cache_creation_input_token_cost_above_1hr": 5e-07,
         "cache_read_input_token_cost": 3e-08,
         "deprecation_date": "2026-04-20",
         "input_cost_per_token": 2.5e-07,
@@ -12288,7 +12288,7 @@
     },
     "claude-3-opus-20240229": {
         "cache_creation_input_token_cost": 1.875e-05,
-        "cache_creation_input_token_cost_above_1hr": 6e-06,
+        "cache_creation_input_token_cost_above_1hr": 3e-05,
         "cache_read_input_token_cost": 1.5e-06,
         "deprecation_date": "2026-01-05",
         "input_cost_per_token": 1.5e-05,

model_prices_and_context_window.json

@@ -12269,7 +12269,7 @@
     },
     "claude-3-haiku-20240307": {
         "cache_creation_input_token_cost": 3e-07,
-        "cache_creation_input_token_cost_above_1hr": 6e-06,
+        "cache_creation_input_token_cost_above_1hr": 5e-07,
         "cache_read_input_token_cost": 3e-08,
         "deprecation_date": "2026-04-20",
         "input_cost_per_token": 2.5e-07,
@@ -12288,7 +12288,7 @@
     },
     "claude-3-opus-20240229": {
         "cache_creation_input_token_cost": 1.875e-05,
-        "cache_creation_input_token_cost_above_1hr": 6e-06,
+        "cache_creation_input_token_cost_above_1hr": 3e-05,
         "cache_read_input_token_cost": 1.5e-06,
         "deprecation_date": "2026-01-05",
         "input_cost_per_token": 1.5e-05,

tests/test_litellm/llms/anthropic/experimental_pass_through/context_management/test_compact_typing_compatibility.py

@@ -0,0 +1,49 @@
+"""
+Unit tests for Python 3.10+ typing compatibility in context management compact editor.
+
+Fixes Issue #38076:
+- Verifies that NotRequired is correctly imported either from typing (>= 3.11) or typing_extensions (< 3.11).
+- Verifies that _SummaryCallKwargs TypedDict definition is valid across Python versions.
+"""
+
+import importlib
+import sys
+
+
+def test_compact_editor_not_required_import():
+    """Verify that compact.py imports NotRequired correctly based on sys.version_info."""
+    import litellm.llms.anthropic.experimental_pass_through.context_management.editors.compact as compact_module
+
+    assert hasattr(compact_module, "NotRequired")
+    if sys.version_info >= (3, 11):
+        import typing
+
+        assert compact_module.NotRequired is typing.NotRequired
+    else:
+        import typing_extensions
+
+        assert compact_module.NotRequired is typing_extensions.NotRequired
+
+
+def test_compact_editor_summary_call_kwargs_definition():
+    """Verify that _SummaryCallKwargs TypedDict is properly constructed."""
+    from litellm.llms.anthropic.experimental_pass_through.context_management.editors.compact import (
+        _SummaryCallKwargs,
+    )
+
+    annotations = _SummaryCallKwargs.__annotations__
+    assert "model" in annotations
+    assert "messages" in annotations
+    assert "max_tokens" in annotations
+    assert "timeout" in annotations
+    assert "litellm_metadata" in annotations
+    assert "user" in annotations
+    assert "allowed_model_region" in annotations
+
+
+def test_compact_module_can_be_imported():
+    """Ensure litellm compact module imports without error on the current runtime."""
+    compact_module = importlib.import_module(
+        "litellm.llms.anthropic.experimental_pass_through.context_management.editors.compact"
+    )
+    assert compact_module is not None

tests/test_litellm/test_claude_3_cache_pricing.py

@@ -0,0 +1,53 @@
+"""
+Unit tests for Claude 3 cache creation pricing above 1 hour in model prices map.
+
+Fixes Issue #38056:
+- claude-3-haiku-20240307 cache_creation_input_token_cost_above_1hr should be 5e-07 (2x input cost of 2.5e-07)
+- claude-3-opus-20240229 cache_creation_input_token_cost_above_1hr should be 3e-05 (2x input cost of 1.5e-05)
+"""
+
+import json
+from pathlib import Path
+
+import pytest
+
+REPO_ROOT = Path(__file__).parents[2]
+MAIN_PATH = REPO_ROOT / "model_prices_and_context_window.json"
+BACKUP_PATH = REPO_ROOT / "litellm" / "model_prices_and_context_window_backup.json"
+
+
+def _load_prices(path: Path) -> dict:
+    with open(path) as f:
+        return json.load(f)
+
+
+@pytest.mark.parametrize("path", [MAIN_PATH, BACKUP_PATH], ids=["main", "backup"])
+def test_claude_3_haiku_cache_creation_cost_above_1hr(path: Path):
+    prices = _load_prices(path)
+    model = "claude-3-haiku-20240307"
+    assert model in prices, f"{model} not found in {path}"
+    model_info = prices[model]
+
+    input_cost = model_info["input_cost_per_token"]
+    assert input_cost == 2.5e-07
+
+    # Cache creation cost above 1hr is 2x standard input cost (5e-07)
+    expected_cache_creation_above_1hr = input_cost * 2
+    assert model_info["cache_creation_input_token_cost_above_1hr"] == expected_cache_creation_above_1hr
+    assert model_info["cache_creation_input_token_cost_above_1hr"] == 5e-07
+
+
+@pytest.mark.parametrize("path", [MAIN_PATH, BACKUP_PATH], ids=["main", "backup"])
+def test_claude_3_opus_cache_creation_cost_above_1hr(path: Path):
+    prices = _load_prices(path)
+    model = "claude-3-opus-20240229"
+    assert model in prices, f"{model} not found in {path}"
+    model_info = prices[model]
+
+    input_cost = model_info["input_cost_per_token"]
+    assert input_cost == 1.5e-05
+
+    # Cache creation cost above 1hr is 2x standard input cost (3e-05)
+    expected_cache_creation_above_1hr = input_cost * 2
+    assert model_info["cache_creation_input_token_cost_above_1hr"] == expected_cache_creation_above_1hr
+    assert model_info["cache_creation_input_token_cost_above_1hr"] == 3e-05