chore: import upstream snapshot with attribution

packages/graphrag-llm/graphrag_llm/cache/create_cache_key.py

@@ -0,0 +1,71 @@
+# Copyright (c) 2024 Microsoft Corporation.
+# Licensed under the MIT License
+
+"""Create cache key."""
+
+from typing import Any
+
+from graphrag_cache import create_cache_key as default_create_cache_key
+
+_CACHE_VERSION = 4
+"""
+If there's a breaking change in what we cache, we should increment this version number to invalidate existing caches.
+
+fnllm was on cache version 2 and though we generate
+similar cache keys, the objects stored in cache by fnllm and litellm are different.
+Using litellm model providers will not be able to reuse caches generated by fnllm
+thus we start with version 3 for litellm.
+
+graphrag-llm package is now on version 4.
+This is to account for changes to the ModelConfig that affect the cache key and
+occurred when pulling this package out of graphrag.
+graphrag-llm, now that is supports metrics, also caches metrics which were not cached before.
+"""
+
+
+def create_cache_key(
+    input_args: dict[str, Any],
+) -> str:
+    """Generate a cache key based on the model configuration and input arguments.
+
+    Args
+    ____
+        input_args: dict[str, Any]
+            The input arguments for the model call.
+
+    Returns
+    -------
+        str
+            The generated cache key in the format
+            `{prefix}_{data_hash}_v{version}` if prefix is provided.
+    """
+    cache_key_parameters = _get_parameters(
+        input_args=input_args,
+    )
+    return default_create_cache_key(cache_key_parameters)
+
+
+def _get_parameters(
+    # model_config: "ModelConfig",
+    input_args: dict[str, Any],
+) -> dict[str, Any]:
+    """Pluck out the parameters that define a cache key."""
+    excluded_keys = [
+        "metrics",
+        "stream",
+        "stream_options",
+        "mock_response",
+        "timeout",
+        "base_url",
+        "api_base",
+        "api_version",
+        "api_key",
+        "azure_ad_token_provider",
+        "drop_params",
+    ]
+
+    parameters: dict[str, Any] = {
+        k: v for k, v in input_args.items() if k not in excluded_keys
+    }
+
+    return parameters