Merge 6558ffe749 into 9b1d69e32c

.github/workflows/test-unit.yml

@@ -999,9 +999,9 @@ jobs:
         run: |
           diff -u model_prices_and_context_window.json litellm/model_prices_and_context_window_backup.json
           if [ -z "$RUST_BRIDGE_ARTIFACT" ]; then
-            .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router --extra saml --extra caching --extra extra_proxy --extra proxy-runtime --extra utils
+            .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router --extra saml --extra extra_proxy --extra proxy-runtime --extra utils
           else
-            .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router --extra saml --extra caching --extra extra_proxy --extra proxy-runtime --extra utils --no-install-project
+            .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router --extra saml --extra extra_proxy --extra proxy-runtime --extra utils --no-install-project
             uv pip install --no-deps --python .venv/bin/python rust-bridge-dist/*.whl
             cp rust-bridge-dist/litellm/rust_bridge/_native.abi3.so litellm/rust_bridge/_native.abi3.so
             uv run --no-sync python -c "import importlib.metadata; import litellm.rust_bridge._native; print(importlib.metadata.version('litellm'))"
@@ -1437,9 +1437,9 @@ jobs:
           RUST_BRIDGE_ARTIFACT: ${{ needs.rust-bridge.outputs.artifact }}
         run: |
           if [ -z "$RUST_BRIDGE_ARTIFACT" ]; then
-            .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router --extra saml --extra caching --extra extra_proxy --extra proxy-runtime --extra utils
+            .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router --extra saml --extra extra_proxy --extra proxy-runtime --extra utils
           else
-            .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router --extra saml --extra caching --extra extra_proxy --extra proxy-runtime --extra utils --no-install-project
+            .github/scripts/uv_sync_with_retries.sh --frozen --group ci --group proxy-dev --extra google --extra proxy --extra semantic-router --extra saml --extra extra_proxy --extra proxy-runtime --extra utils --no-install-project
             uv pip install --no-deps --python .venv/bin/python rust-bridge-dist/*.whl
             cp rust-bridge-dist/litellm/rust_bridge/_native.abi3.so litellm/rust_bridge/_native.abi3.so
           fi

litellm/caching/Readme.md

@@ -8,10 +8,9 @@ The following caching mechanisms are supported:
 2. **RedisSemanticCache**
 3. **QdrantSemanticCache**
 4. **InMemoryCache**
-5. **DiskCache**
-6. **S3Cache**
-7. **AzureBlobCache**
-8. **DualCache** (updates both Redis and an in-memory cache simultaneously)
+5. **S3Cache**
+6. **AzureBlobCache**
+7. **DualCache** (updates both Redis and an in-memory cache simultaneously)
 
 ## Folder Structure
 
@@ -20,7 +19,6 @@ litellm/caching/
 ├── base_cache.py
 ├── caching.py
 ├── caching_handler.py
-├── disk_cache.py
 ├── dual_cache.py
 ├── in_memory_cache.py
 ├── qdrant_semantic_cache.py

litellm/caching/__init__.py

@@ -1,6 +1,5 @@
 from .azure_blob_cache import AzureBlobCache
 from .caching import Cache, LiteLLMCacheType
-from .disk_cache import DiskCache
 from .dual_cache import DualCache
 from .gcs_cache import GCSCache
 from .in_memory_cache import InMemoryCache

litellm/caching/caching.py

@@ -31,7 +31,6 @@ from litellm.types.utils import EmbeddingResponse, is_litellm_owned_kwarg
 
 from .azure_blob_cache import AzureBlobCache
 from .base_cache import BaseCache
-from .disk_cache import DiskCache
 from .dual_cache import DualCache  # noqa: F401  # re-exported, callers import DualCache from litellm.caching.caching
 from .gcs_cache import GCSCache
 from .in_memory_cache import InMemoryCache
@@ -132,7 +131,6 @@ class Cache:
         valkey_semantic_cache_index_name: str | None = None,
         redis_flush_size: int | None = None,
         redis_startup_nodes: list | None = None,
-        disk_cache_dir: str | None = None,
         qdrant_api_base: str | None = None,
         qdrant_api_key: str | None = None,
         qdrant_collection_name: str | None = None,
@@ -153,7 +151,7 @@ class Cache:
         Initializes the cache based on the given type.
 
         Args:
-            type (str, optional): The type of cache to initialize. Can be "local", "redis", "redis-semantic", "qdrant-semantic", "s3" or "disk". Defaults to "local".
+            type (str, optional): The type of cache to initialize. Can be "local", "redis", "redis-semantic", "qdrant-semantic" or "s3". Defaults to "local".
 
             # Redis Cache Args
             host (str, optional): The host address for the Redis cache. Required if type is "redis".
@@ -174,9 +172,6 @@ class Cache:
             semantic_cache_scope (str, optional): "key" isolates semantic-cache buckets per key/team/org. "end_user" additionally isolates per end user (falls back to the key scope when the request carries no end-user id). Defaults to "key".
             max_messages (int, optional): Requests with more `messages` (or Responses API `input` items) than this are neither looked up nor stored, so long agent conversations never serve or create a cache entry. None disables the limit. Defaults to 4.
 
-            # Disk Cache Args
-            disk_cache_dir (str, optional): The directory for the disk cache. Defaults to None.
-
             # S3 Cache Args
             s3_bucket_name (str, optional): The bucket name for the s3 cache. Defaults to None.
             s3_region_name (str, optional): The region name for the s3 cache. Defaults to None.
@@ -306,8 +301,9 @@ class Cache:
                 account_url=azure_account_url,
                 container=azure_blob_container,
             )
-        elif type == LiteLLMCacheType.DISK:
-            self.cache = DiskCache(disk_cache_dir=disk_cache_dir)
+        else:
+            supported_types: Final = ", ".join(member.value for member in LiteLLMCacheType)
+            raise ValueError(f"Unsupported cache type {type!r}. Supported types: {supported_types}")
         if "cache" not in litellm.input_callback:
             litellm.input_callback.append("cache")
         if "cache" not in litellm.success_callback:
@@ -1028,7 +1024,7 @@ def enable_cache(
     Enable cache with the specified configuration.
 
     Args:
-        type (Optional[Literal["local", "redis", "s3", "disk"]]): The type of cache to enable. Defaults to "local".
+        type (Optional[Literal["local", "redis", "s3"]]): The type of cache to enable. Defaults to "local".
         host (Optional[str]): The host address of the cache server. Defaults to None.
         port (Optional[str]): The port number of the cache server. Defaults to None.
         password (Optional[str]): The password for the cache server. Defaults to None.
@@ -1075,7 +1071,7 @@ def update_cache(
     Update the cache for LiteLLM.
 
     Args:
-        type (Optional[Literal["local", "redis", "s3", "disk"]]): The type of cache. Defaults to "local".
+        type (Optional[Literal["local", "redis", "s3"]]): The type of cache. Defaults to "local".
         host (Optional[str]): The host of the cache. Defaults to None.
         port (Optional[str]): The port of the cache. Defaults to None.
         password (Optional[str]): The password for the cache. Defaults to None.

litellm/caching/caching_handler.py

@@ -9,7 +9,7 @@ This file is a wrapper around caching.py
 
 This class is used to handle caching logic specific for LLM API requests (completion / embedding / text_completion / transcription etc)
 
-It utilizes the (RedisCache, s3Cache, RedisSemanticCache, QdrantSemanticCache, InMemoryCache, DiskCache) based on what the user has setup
+It utilizes the (RedisCache, s3Cache, RedisSemanticCache, QdrantSemanticCache, InMemoryCache) based on what the user has setup
 
 In each method it will call the appropriate method from caching.py
 """

litellm/caching/disk_cache.py

@@ -1,92 +0,0 @@
-import json
-from typing import TYPE_CHECKING, Any, Final
-
-from pydantic import TypeAdapter
-
-from .base_cache import BaseCache
-
-if TYPE_CHECKING:
-    from opentelemetry.trace import Span as _Span
-
-    Span = _Span
-else:
-    Span = Any
-
-_DECODED_DISK_VALUE: Final = TypeAdapter(object)
-
-
-class DiskCache(BaseCache):
-    def __init__(self, disk_cache_dir: str | None = None):
-        try:
-            import diskcache as dc
-        except ModuleNotFoundError as e:
-            raise ModuleNotFoundError("Please install litellm with `litellm[caching]` to use disk caching.") from e
-
-        # if users don't provider one, use the default litellm cache
-        if disk_cache_dir is None:
-            self.disk_cache = dc.Cache(".litellm_cache")
-        else:
-            self.disk_cache = dc.Cache(disk_cache_dir)
-
-    def set_cache(self, key, value, **kwargs):
-        if "ttl" in kwargs:
-            self.disk_cache.set(key, value, expire=kwargs["ttl"])
-        else:
-            self.disk_cache.set(key, value)
-
-    async def async_set_cache(self, key, value, **kwargs):
-        self.set_cache(key=key, value=value, **kwargs)
-
-    async def async_set_cache_pipeline(self, cache_list, **kwargs):
-        for cache_key, cache_value in cache_list:
-            if "ttl" in kwargs:
-                self.set_cache(key=cache_key, value=cache_value, ttl=kwargs["ttl"])
-            else:
-                self.set_cache(key=cache_key, value=cache_value)
-
-    def get_cache(self, key, **kwargs):
-        original_cached_response: Final = self.disk_cache.get(key)
-        if original_cached_response:
-            try:
-                cached_response = _DECODED_DISK_VALUE.validate_python(json.loads(original_cached_response))
-            except Exception:
-                cached_response = original_cached_response
-            return cached_response
-        return None
-
-    def batch_get_cache(self, keys: list, **kwargs):
-        return_val: Final = []
-        for k in keys:
-            val = self.get_cache(key=k, **kwargs)
-            return_val.append(val)
-        return return_val
-
-    def increment_cache(self, key, value: int, **kwargs) -> int:
-        with self.disk_cache.transact():
-            cached_value: Final = self.get_cache(key=key)
-            init_value: Final = cached_value if isinstance(cached_value, int) else 0
-            new_value: Final = init_value + value
-            self.set_cache(key, new_value, **kwargs)
-            return new_value
-
-    async def async_get_cache(self, key, **kwargs):
-        return self.get_cache(key=key, **kwargs)
-
-    async def async_batch_get_cache(self, keys: list, **kwargs):
-        return_val: Final = []
-        for k in keys:
-            val = self.get_cache(key=k, **kwargs)
-            return_val.append(val)
-        return return_val
-
-    async def async_increment(self, key, value: int, **kwargs) -> int:
-        return self.increment_cache(key=key, value=value, **kwargs)
-
-    def flush_cache(self):
-        self.disk_cache.clear()
-
-    async def disconnect(self):
-        pass
-
-    def delete_cache(self, key):
-        self.disk_cache.pop(key)

litellm/router.py

@@ -1383,7 +1383,7 @@ class Router:
         self.deployment_names: list = []  # names of models under litellm_params. ex. azure/chatgpt-v-2
         self.deployment_latency_map = {}
         ### CACHING ###
-        cache_type: Literal["local", "redis", "redis-semantic", "s3", "disk"] = "local"  # default to an in-memory cache
+        cache_type: Literal["local", "redis", "redis-semantic", "s3"] = "local"  # default to an in-memory cache
         redis_cache = None
         cache_config: Final[dict[str, Any]] = {}
 

litellm/types/caching.py

@@ -13,7 +13,6 @@ class LiteLLMCacheType(str, Enum):
     REDIS_SEMANTIC = "redis-semantic"
     VALKEY_SEMANTIC = "valkey-semantic"
     S3 = "s3"
-    DISK = "disk"
     QDRANT_SEMANTIC = "qdrant-semantic"
     AZURE_BLOB = "azure-blob"
     GCS = "gcs"

osv-scanner.toml

@@ -1,8 +1,3 @@
-[[IgnoredVulns]]
-id = "GHSA-w8v5-vhqr-4h9v"
-ignoreUntil = 2026-11-01
-reason = "diskcache has no fixed release published; remove this entry once one exists"
-
 [[IgnoredVulns]]
 id = "GHSA-v4x8-3wr5-3wq8"
 ignoreUntil = 2026-11-20

pyproject.toml

@@ -130,7 +130,6 @@ utils = [
     # Not in Docker or PyPI proxy extra.
     "numpydoc>=1.8.0,<2.0",
 ]
-caching = ["diskcache>=5.6.3,<6.0"]
 mcp = ["mcp>=2.3.0,<3", "httpx2>=2.10.0,<3", "pydantic>=2.12.0,<3"]
 # Driver for the MongoDB Atlas vector store; Atlas Vector Search has no HTTP query API.
 # The floor is 4.9 because that is the release AsyncMongoClient landed in.

tests/unit/caching/test_caching.py

@@ -250,6 +250,13 @@ def test_semantic_cache_key_reads_tenant_identity_from_litellm_metadata():
     assert key_a == key_a_in_litellm_params
 
 
+@pytest.mark.parametrize("cache_type", ["disk", "not-a-cache"])
+def test_cache_rejects_unsupported_type(cache_type: str) -> None:
+    cache_params: Final = {"type": cache_type}
+    with pytest.raises(ValueError, match=f"Unsupported cache type '{cache_type}'. Supported types: local, redis,"):
+        Cache(**cache_params)
+
+
 def test_semantic_cache_scope_rejects_unknown_value():
     with pytest.raises(ValueError, match="'team' is not a valid SemanticCacheScope"):
         _semantic_cache(semantic_cache_scope="team")

tests/unit/caching/test_disk_cache.py

@@ -1,158 +0,0 @@
-import threading
-import time
-from concurrent.futures import ThreadPoolExecutor
-
-import pytest
-
-pytest.importorskip("diskcache")
-
-from litellm.caching.disk_cache import DiskCache
-from tests._wait_helpers import await_until
-
-
-class _SlowInt(int):
-    def __add__(self, value: int) -> "_SlowInt":
-        time.sleep(0.05)
-        return _SlowInt(int(self) + value)
-
-
-@pytest.fixture
-def cache(tmp_path):
-    return DiskCache(disk_cache_dir=str(tmp_path))
-
-
-def test_increment_cache_starts_from_zero_when_key_missing(cache):
-    assert cache.increment_cache("counter", 3) == 3
-    assert cache.get_cache("counter") == 3
-
-
-def test_increment_cache_adds_to_existing_int(cache):
-    cache.set_cache("counter", 7)
-    assert cache.increment_cache("counter", 5) == 12
-    assert cache.get_cache("counter") == 12
-
-
-def test_increment_cache_treats_non_int_cached_value_as_zero(cache):
-    cache.set_cache("counter", "not-a-number")
-    assert cache.increment_cache("counter", 4) == 4
-    assert cache.get_cache("counter") == 4
-
-
-def test_increment_cache_is_atomic_under_thread_concurrency(cache):
-    seed = 1000
-    cache.set_cache("counter", _SlowInt(seed))
-    thread_count = 8
-    barrier = threading.Barrier(thread_count)
-
-    def increment(_: int) -> int:
-        barrier.wait()
-        return cache.increment_cache("counter", 1)
-
-    with ThreadPoolExecutor(max_workers=thread_count) as executor:
-        tuple(executor.map(increment, range(thread_count)))
-
-    assert cache.get_cache("counter") == seed + thread_count
-
-
-async def test_async_increment_starts_from_zero_when_key_missing(cache):
-    assert await cache.async_increment("counter", 2) == 2
-
-
-async def test_async_increment_adds_to_existing_int(cache):
-    await cache.async_set_cache("counter", 10)
-    assert await cache.async_increment("counter", 5) == 15
-
-
-async def test_async_increment_treats_non_int_cached_value_as_zero(cache):
-    await cache.async_set_cache("counter", "corrupt")
-    assert await cache.async_increment("counter", 9) == 9
-
-
-@pytest.fixture
-def disk_cache(tmp_path):
-    import litellm
-    from litellm.caching import Cache
-
-    litellm.cache = Cache(type="disk", disk_cache_dir=str(tmp_path))
-    yield litellm.cache
-    litellm.cache = None
-
-
-COMPLETION_MESSAGES = [{"role": "user", "content": "write a one sentence poem about: disk-cache-unit-key"}]
-
-
-def test_completion_is_served_from_disk_cache_until_params_change(disk_cache):
-    import litellm
-
-    first = litellm.completion(
-        "gpt-3.5-turbo", messages=COMPLETION_MESSAGES, caching=True, max_tokens=20, mock_response="first"
-    )
-    repeat = litellm.completion(
-        "gpt-3.5-turbo", messages=COMPLETION_MESSAGES, caching=True, max_tokens=20, mock_response="second"
-    )
-    changed = litellm.completion(
-        "gpt-3.5-turbo", messages=COMPLETION_MESSAGES, caching=True, temperature=0.5, mock_response="third"
-    )
-
-    assert (repeat.choices[0].message.content, repeat.id, repeat.created) == (
-        first.choices[0].message.content,
-        first.id,
-        first.created,
-    )
-    assert changed.choices[0].message.content == "third"
-
-
-async def test_async_completion_is_served_from_disk_cache_until_params_change(disk_cache):
-    import litellm
-
-    first = await litellm.acompletion(
-        "gpt-3.5-turbo", messages=COMPLETION_MESSAGES, caching=True, max_tokens=20, mock_response="first"
-    )
-    await await_until(lambda: len(disk_cache.cache.disk_cache) == 1, message="the first completion is cached")
-    repeat = await litellm.acompletion(
-        "gpt-3.5-turbo", messages=COMPLETION_MESSAGES, caching=True, max_tokens=20, mock_response="second"
-    )
-    changed = await litellm.acompletion(
-        "gpt-3.5-turbo", messages=COMPLETION_MESSAGES, caching=True, temperature=0.5, mock_response="third"
-    )
-
-    assert (repeat.choices[0].message.content, repeat.id, repeat.created) == (
-        first.choices[0].message.content,
-        first.id,
-        first.created,
-    )
-    assert changed.choices[0].message.content == "third"
-
-
-async def test_embedding_is_served_from_disk_cache_until_params_change(disk_cache, respx_mock, monkeypatch):
-    import httpx
-
-    import litellm
-
-    monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
-    input_text = ["hello disk-embedding-unit-key"]
-    embeddings_route = respx_mock.post("https://api.openai.com/v1/embeddings").mock(
-        return_value=httpx.Response(
-            200,
-            json={
-                "object": "list",
-                "data": [{"object": "embedding", "index": 0, "embedding": [0.1, 0.2, 0.3]}],
-                "model": "text-embedding-ada-002",
-                "usage": {"prompt_tokens": 2, "total_tokens": 2},
-            },
-        )
-    )
-
-    await litellm.aembedding("openai/text-embedding-ada-002", input=input_text, caching=True, api_key="sk-unit")
-    await await_until(lambda: len(disk_cache.cache.disk_cache) == 1, message="the first embedding is cached")
-    cached = await litellm.aembedding(
-        "openai/text-embedding-ada-002", input=input_text, caching=True, api_key="sk-unit"
-    )
-    assert cached._hidden_params["cache_hit"] is True
-    assert cached.data[0]["embedding"] == [0.1, 0.2, 0.3]
-    assert embeddings_route.call_count == 1
-
-    await litellm.aembedding(
-        "openai/text-embedding-ada-002", input=input_text, caching=True, user="charlie", api_key="sk-unit"
-    )
-    assert embeddings_route.call_count == 2

tests/unit/proxy/proxy_server/test_proxy_config.py

@@ -1656,7 +1656,7 @@ def test_ProxyConfig__init_cache_sets_litellm_cache(monkeypatch):
 
 def test_ProxyConfig__init_cache_invalid_params_raises():
     pc = ProxyConfig()
-    with pytest.raises(AttributeError):
+    with pytest.raises(ValueError, match="Unsupported cache type 'this-cache-type-does-not-exist'"):
         pc._init_cache(cache_params={"type": "this-cache-type-does-not-exist"})
 
 

uv.lock

@@ -1952,15 +1952,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/1e/77/dc8c558f7593132cf8fefec57c4f60c83b16941c574ac5f619abb3ae7933/dill-0.4.1-py3-none-any.whl", hash = "sha256:1e1ce33e978ae97fcfcff5638477032b801c46c7c65cf717f95fbc2248f79a9d", size = 120019, upload-time = "2026-01-19T02:36:55.663Z" },
 ]
 
-[[package]]
-name = "diskcache"
-version = "5.6.3"
-source = { registry = "https://pypi.org/simple" }
-sdist = { url = "https://files.pythonhosted.org/packages/3f/21/1c1ffc1a039ddcc459db43cc108658f32c57d271d7289a2794e401d0fdb6/diskcache-5.6.3.tar.gz", hash = "sha256:2c3a3fa2743d8535d832ec61c2054a1641f41775aa7c556758a109941e33e4fc", size = 67916, upload-time = "2023-08-31T06:12:00.316Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/3f/27/4570e78fc0bf5ea0ca45eb1de3818a23787af9b390c0b0a0033a1b8236f9/diskcache-5.6.3-py3-none-any.whl", hash = "sha256:5e31b2d5fbad117cc363ebaf6b689474db18a1f6438bc82358b024abd4c2ca19", size = 45550, upload-time = "2023-08-31T06:11:58.822Z" },
-]
-
 [[package]]
 name = "distro"
 version = "1.9.0"
@@ -4992,9 +4983,6 @@ dependencies = [
 bedrock-realtime = [
     { name = "aws-sdk-bedrock-runtime", extra = ["awscrt"], marker = "python_full_version >= '3.12'" },
 ]
-caching = [
-    { name = "diskcache" },
-]
 cli = [
     { name = "inquirerpy" },
     { name = "keyring" },
@@ -5247,7 +5235,6 @@ requires-dist = [
     { name = "cryptography", marker = "extra == 'proxy'", specifier = ">=49.0.0,<51.0" },
     { name = "ddtrace", marker = "extra == 'proxy-runtime'", specifier = ">=4.8.2,<5.0" },
     { name = "detect-secrets", marker = "extra == 'proxy-runtime'", specifier = ">=1.5.0,<2.0" },
-    { name = "diskcache", marker = "extra == 'caching'", specifier = ">=5.6.3,<6.0" },
     { name = "expression", marker = "extra == 'proxy'", specifier = ">=5.6.0,<6.0" },
     { name = "fastapi", marker = "extra == 'proxy'", specifier = ">=0.136.3,<1.0" },
     { name = "fastapi-sso", marker = "extra == 'proxy'", specifier = ">=0.19.0,<1.0" },
@@ -5340,7 +5327,7 @@ requires-dist = [
     { name = "uvloop", marker = "sys_platform != 'win32' and extra == 'proxy'", specifier = ">=0.22.1,<1.0" },
     { name = "websockets", marker = "extra == 'proxy'", specifier = ">=15.0.1,<16.0" },
 ]
-provides-extras = ["proxy", "cli", "extra-proxy", "utils", "caching", "mcp", "saml", "semantic-router", "mlflow", "grpc", "stt-vertex-chirp", "stt-nvidia-riva", "google", "bedrock-realtime", "proxy-runtime"]
+provides-extras = ["proxy", "cli", "extra-proxy", "utils", "mcp", "saml", "semantic-router", "mlflow", "grpc", "stt-vertex-chirp", "stt-nvidia-riva", "google", "bedrock-realtime", "proxy-runtime"]
 
 [package.metadata.requires-dev]
 admin-mcp = [{ name = "litellm-admin-mcp", marker = "python_full_version >= '3.12'", url = "https://github.com/BerriAI/liteadmin-mcp/archive/d35ec9c19c117d4c50cc1eccf6ce1296aac25a1a.tar.gz" }]