Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/LMCache/LMCache
/ functions
Functions
13,371 in github.com/LMCache/LMCache
⨍
Functions
13,371
◇
Types & classes
2,044
↳
Endpoints
181
↓ 15 callers
Method
contains
(self, item: Union[str, int])
lmcache/v1/utils/bloom_filter.py:79
↓ 15 callers
Function
counter_value
Look up a counter delta by metric name + attribute equality. Returns 0 if no data point matches the attribute set exactly.
tests/v1/mp_observability/subscribers/metrics/counter_helpers.py:64
↓ 15 callers
Method
destroy
Close and delete the LMCacheEngine instance by the instance ID
lmcache/v1/cache_engine.py:2155
↓ 15 callers
Method
get_current_stats
Return current aggregated stats snapshot. Thread-safe.
lmcache/cli/commands/bench/engine_bench/stats.py:129
↓ 15 callers
Function
get_spec
Return a spec instance wrapping *kv_caches* of *fmt*.
lmcache/v1/gpu_connector/kv_format/specs/registry.py:52
↓ 15 callers
Function
get_type_name_for_config
Reverse-lookup the registered type name for a config instance. Args: config: An L2 adapter config instance. Returns:
lmcache/v1/distributed/l2_adapters/config.py:127
↓ 15 callers
Function
get_zmq_context
(use_asyncio: bool = True)
lmcache/v1/rpc_utils.py:23
↓ 15 callers
Method
ipc_handle
Return an IPC handle consumable by the multiprocess server.
lmcache/v1/multiprocess/transfer_context/worker_transfer.py:96
↓ 15 callers
Method
is_key_evictable
Check if a key is eligible for eviction (not locked). This method does NOT acquire the global L1Manager lock. L1Manager.delete() will
lmcache/v1/distributed/l1_manager.py:783
↓ 15 callers
Function
make_adapter
Create a MockL2Adapter with fast bandwidth.
tests/v1/distributed/test_store_controller.py:150
↓ 15 callers
Function
make_key
Create an ObjectKey for testing.
tests/v1/multiprocess/test_unified_touch.py:31
↓ 15 callers
Method
read_prefetched_results
Read the memory objects from L1 storage that has been prefetched beforehand. Yielding an optional list of memory objects correspondin
lmcache/v1/distributed/storage_manager.py:250
↓ 15 callers
Function
register_gauge
Register an OTel observable gauge with a callback. This is a convenience wrapper that hides the OTel boilerplate. If OTel is not available, t
lmcache/v1/mp_observability/otel_init.py:162
↓ 15 callers
Method
register_handler
Register a handler for a specific RequestType. Args: request_type: The type of request to handle handler: Ha
tests/v1/multiprocess/test_mq.py:198
↓ 15 callers
Method
run_forever
Keep the engine running
lmcache/v1/standalone/__main__.py:343
↓ 15 callers
Method
run_test
Run a test by starting server and client processes. Args: request_type: Type of request to send payloads: Li
tests/v1/multiprocess/test_mq.py:216
↓ 15 callers
Method
serialize
Serialize/compress the memory object. Input: memory_obj: the memory object to be serialized/compressed. Returns
lmcache/v1/storage_backend/naive_serde/serde.py:11
↓ 15 callers
Method
start
(self)
tests/cli/commands/bench/test_server_bench.py:440
↓ 15 callers
Function
store_and_wait
(adapter: DaxL2Adapter, key: ObjectKey, obj: MemoryObj)
tests/v1/distributed/test_dax_l2_adapter.py:171
↓ 15 callers
Method
submit
Submit I/O operation - performs actual file I/O. For write operations: writes data from shared memory to file. For read operations: r
tests/conftest.py:403
↓ 15 callers
Method
submit_prefetch_task
(self, **kw: Any)
tests/cli/commands/trace/test_dispatch.py:45
↓ 15 callers
Function
wait_for_event_fd
(event_fd: int, timeout: float = 5.0)
tests/v1/distributed/test_hfbucket_l2_adapter.py:162
↓ 15 callers
Method
wait_stream
Make this stream wait until all kernels on *stream* complete. Args: stream: The stream whose pending work must finish before
lmcache/v1/platform/cpu/stub_cpu_device.py:159
↓ 15 callers
Method
with_new_worker_id
(self, new_worker_id: int)
lmcache/utils.py:545
↓ 14 callers
Method
GetOrCreate
( labels: dict, )
lmcache/v1/cache_controller/observability.py:134
↓ 14 callers
Function
_create_object_key
( chunk_id: int, model_name: str = "test_model", cache_salt: str = "", )
tests/v1/distributed/test_raw_block_l2_adapter.py:77
↓ 14 callers
Function
_get_histogram_count
(name: str)
tests/v1/mp_observability/subscribers/metrics/test_l1_lifecycle.py:44
↓ 14 callers
Function
_make_workload_config
(**overrides)
tests/cli/commands/bench/engine_bench/workloads/test_long_doc_qa.py:128
↓ 14 callers
Function
_nixl_cpu_config
(pool_size: int = 0)
tests/v1/storage_backend/test_nixl_shared_pool.py:119
↓ 14 callers
Method
_notify_keys_deleted
Update byte accounting and notify listeners that ``keys`` were deleted. ``sizes[i]`` is the byte size of ``keys[i]`` (typically the sa
lmcache/v1/distributed/l2_adapters/base.py:389
↓ 14 callers
Function
_object_key_to_string
Serialize an ObjectKey to the native-connector wire format. Unsalted:: <model_name>@<kv_rank_hex>@<object_group_id_hex>@<chunk_hash_hex>
lmcache/v1/distributed/l2_adapters/native_connector_l2_adapter.py:51
↓ 14 callers
Function
_percentile
Return the *pct*-th percentile from an already-sorted list. Uses nearest-rank with no interpolation: for N samples, the p-th percentile is th
lmcache/cli/commands/trace/_stats.py:71
↓ 14 callers
Function
_read_counter_by_attrs
()
tests/v1/mp_observability/subscribers/metrics/test_engine.py:64
↓ 14 callers
Function
_total_count
(name: str)
tests/v1/mp_observability/subscribers/metrics/test_l0_l1_throughput.py:88
↓ 14 callers
Function
applyResponse
applyResponse applies the response's JSON patches to the original pod JSON and returns the mutated pod. It asserts the response is Allowed.
operator/internal/webhook/pod_injector_test.go:118
↓ 14 callers
Method
batched_get_non_blocking
( self, lookup_id: str, keys: list[CacheEngineKey], transfer_spec: Any = None,
lmcache/v1/storage_backend/p2p_backend.py:550
↓ 14 callers
Method
close
Close EC storage resources and background workers.
lmcache/v1/ec_engine.py:110
↓ 14 callers
Method
compute_aligned_size
Helper function to compute the aligned size for a given raw size. Args: raw_size: The raw size to be aligned. R
lmcache/v1/memory_management.py:1296
↓ 14 callers
Function
create_object_key
(chunk_id: int, model_name: str = "test_model")
tests/v1/distributed/test_mooncake_store_l2_adapter.py:68
↓ 14 callers
Function
create_test_key
(key_id: int = 0)
tests/v1/storage_backend/test_bigtable_connector.py:204
↓ 14 callers
Method
deserialize
Deserialize/decompress the memory object. Input: memory_obj: the memory object to be deserialized/decompressed.
lmcache/v1/storage_backend/naive_serde/serde.py:26
↓ 14 callers
Method
emit
Trigger all registered handlers.
lmcache/cli/metrics/metrics.py:144
↓ 14 callers
Method
exists_in_put_tasks
Check if key is in ongoing put tasks.
tests/v1/storage_backend/test_storage_plugin.py:76
↓ 14 callers
Method
get
Returns the LMCacheEngine instance associated with the instance ID, or None if not found.
lmcache/v1/cache_engine.py:2149
↓ 14 callers
Method
get_context
(self)
examples/frontend/chat_session.py:37
↓ 14 callers
Function
get_event_bus
Return the current global EventBus singleton.
lmcache/v1/mp_observability/event_bus.py:374
↓ 14 callers
Method
get_from_extra_config
(self, key: str, default=None)
tests/v1/test_ec_connector.py:26
↓ 14 callers
Method
get_memory_format
Get the memory format of the MemoryObj.
lmcache/v1/memory_management.py:258
↓ 14 callers
Method
handle_full_sync_start
Handle full sync start request from a worker. This is called when a worker wants to start full sync. The controller should:
lmcache/v1/cache_controller/controllers/kv_controller.py:176
↓ 14 callers
Method
has_quota
Whether ``cache_salt`` has an explicit registration. Useful for distinguishing "zero limit by default" from "explicitly registered wi
lmcache/v1/distributed/quota_manager.py:75
↓ 14 callers
Method
initialize_kvcaches_ptr
Initialize the kvcaches pointers if not already initialized.
lmcache/v1/gpu_connector/mock_gpu_connector.py:64
↓ 14 callers
Method
on_request_sent
Called when a request is dispatched. Increments in-flight.
lmcache/cli/commands/bench/engine_bench/progress.py:73
↓ 14 callers
Method
query
(self)
tests/v1/distributed/l2_adapters/test_p2p_l2_adapter.py:44
↓ 14 callers
Method
register_worker
Register a new worker, creating instance if needed.
lmcache/v1/cache_controller/utils.py:400
↓ 14 callers
Method
shutdown
(self)
tests/v1/mp_observability/test_event_bus.py:52
↓ 13 callers
Function
_alloc_chunks
(shape: tuple[int, ...], count: int)
tests/v1/test_python_ops_fallback.py:1833
↓ 13 callers
Method
_alloc_device_ids
Allocate ``count`` globally unique OBJ device_id values. The NIXL OBJ backend indexes registrations by device_id in a flat unordered_
lmcache/v1/storage_backend/nixl_storage_backend.py:1431
↓ 13 callers
Function
_call
Submit a request through ``MessageQueueClient`` and block. Returns the decoded response (possibly ``None`` for void replies) on success, or t
lmcache/cli/commands/bench/server_bench/helpers.py:140
↓ 13 callers
Function
_client
()
tests/v1/mp_coordinator/test_api.py:12
↓ 13 callers
Function
_make_config
(**overrides)
tests/cli/commands/bench/engine_bench/workloads/test_multi_round_chat.py:224
↓ 13 callers
Function
_roundtrip
Encode then decode through msgpack to mirror the recorder path.
tests/v1/mp_observability/trace/test_codecs.py:19
↓ 13 callers
Method
batched_get
Blocking function to get the memory objects from the storages.
lmcache/v1/storage_backend/storage_manager.py:479
↓ 13 callers
Method
batched_get_blocking
( self, keys: List[CacheEngineKey], )
lmcache/v1/storage_backend/gds_backend.py:875
↓ 13 callers
Function
batched_iteration_with_skip
Utility function to iterate over a list in batches with an initial skip. Args: lst: The list to iterate over. batch_size: The siz
lmcache/v1/multiprocess/modules/lmcache_driven_transfer.py:86
↓ 13 callers
Function
chunk_hash_windows_numba
Compute polynomial hashes over non-overlapping (chunked) windows. Unlike the rolling-hash variant, each window's hash is computed independent
lmcache/v1/multiprocess/token_hasher.py:311
↓ 13 callers
Method
close
(self)
lmcache/v1/multiprocess/mq.py:385
↓ 13 callers
Function
create_memory_obj
(size: int = 16, fill_value: float = 1.0)
tests/v1/distributed/test_s3_l2_adapter.py:339
↓ 13 callers
Function
create_test_key
Helper to create a test CacheEngineKey
tests/v1/storage_backend/test_audit_connector.py:26
↓ 13 callers
Function
findArgValue
(t *testing.T, args []string, flag string)
operator/internal/resources/resources_test.go:41
↓ 13 callers
Method
flush
Manually flush all pending messages. This method ensures all pending messages in the queue are processed before returning. It trigger
lmcache/v1/storage_backend/batched_message_sender.py:199
↓ 13 callers
Function
get_expected_count
Calculate expected token count based on save_unfull_chunk setting. Args: token_len: Total token length save_unfull_chunk: Whether
tests/v1/test_cache_engine.py:47
↓ 13 callers
Function
get_num_blocks
Return the number of blocks from ``kv_caches``. Raises: ValueError: For NBBS-fused formats with no separate block axis.
lmcache/v1/gpu_connector/utils.py:233
↓ 13 callers
Method
get_statistics
(self)
lmcache/v1/utils/bloom_filter.py:99
↓ 13 callers
Method
get_worker
Get a worker by worker_id. Uses optimistic locking (lock-free read).
lmcache/v1/cache_controller/utils.py:272
↓ 13 callers
Method
hash_tokens
Hash one chunk with rolling prefix. Returns int or bytes depending on hash_func.
lmcache/v1/multiprocess/token_hasher.py:183
↓ 13 callers
Function
minimalCBEngine
minimalCBEngine returns a CacheBlendEngine with only the required L1 size set, then applies SetDefaults so blend/injection/chunk-size defaults are pin
operator/internal/resources/cacheblend_engine_test.go:33
↓ 13 callers
Function
parse_cache_key
Parse a key string into either a CacheEngineKey or LayerCacheEngineKey. Args: key_str: String in format: CacheEngineKey:
lmcache/utils.py:376
↓ 13 callers
Function
parse_kvcache_shape_spec
Parse a ``--kvcache-shape-spec`` string into layer groups. **Grammar** (EBNF-ish):: spec := group { ";" group } group
lmcache/v1/kv_layer_groups.py:649
↓ 13 callers
Function
set_tracing_enabled
Flip the trace gate. Called by trace recorders during ``__init__`` (on) and ``close()`` (off). Direct callers should not normally use this.
lmcache/v1/mp_observability/trace/decorator.py:49
↓ 13 callers
Method
shutdown
Sync shutdown — blocks the calling thread until shutdown is complete.
lmcache/v1/storage_backend/job_executor/pq_executor.py:125
↓ 13 callers
Method
shutdown
Optional cleanup hook. Called by ``EventBus.stop()``.
lmcache/v1/mp_observability/event_bus.py:90
↓ 13 callers
Method
start_msg
Create FullSyncStartMsg.
tests/v1/cache_controller/conftest.py:177
↓ 13 callers
Method
stop
(self, operation_name)
lmcache/v1/storage_backend/connector/eic_connector.py:50
↓ 13 callers
Function
store_range
A StoreRange with one object_key per complete chunk (mirrors publisher).
tests/v1/mp_coordinator/test_blend_directory.py:21
↓ 13 callers
Method
submit_put_task
Submit a put task to store KV cache to remote storage asynchronously. :param on_complete_callback: Optional callback invoked after t
lmcache/v1/storage_backend/remote_backend.py:221
↓ 13 callers
Function
transformPath
(path)
lmcache/lmcache_frontend/static/js/app.js:1010
↓ 13 callers
Function
write_keys_to_l1
Write keys to L1 (reserve_write + finish_write). Args: l1_manager: The L1 manager to write to. keys: Object keys to write.
tests/v1/distributed/test_store_controller.py:99
↓ 12 callers
Method
FromMessagingFuture
( raw_future: MessagingFuture[tuple[bytes, T]], device: Any | None = None, )
lmcache/v1/multiprocess/futures.py:198
↓ 12 callers
Method
_build_messages
Build chat messages for a benchmark request.
lmcache/cli/commands/bench/engine_bench/workloads/long_doc_qa.py:199
↓ 12 callers
Method
_call
( backend: Mock, keys: List[CacheEngineKey], pin: bool = False, )
tests/v1/test_nixl_batched_contains.py:186
↓ 12 callers
Function
_consume_fd
(fd)
tests/v1/native_storage_ops/test_periodic_event_notifier.py:49
↓ 12 callers
Function
_fake_stream
Async generator yielding chunks.
tests/cli/commands/bench/engine_bench/test_request_sender.py:74
↓ 12 callers
Function
_generate_vocab_pool
Generate a vocabulary pool of ``size`` unique pseudo-words. Deterministically generates synthetic words so every token is unique. Args:
lmcache/cli/commands/bench/engine_bench/workloads/long_doc_permutator.py:106
↓ 12 callers
Method
_make_backend
Create a GdsBackend with mocked allocator and fstype. Mocks are used so the tests run without cuFile / real NVMe.
tests/v1/storage_backend/test_gds_backend.py:722
↓ 12 callers
Function
_make_collector
()
tests/cli/commands/bench/engine_bench/test_progress.py:24
↓ 12 callers
Function
_make_kv_tensors
Build non-MLA per-layer KV tensors shaped ``[2, NB, BS, NH, HS]``.
tests/v1/platform/test_gpu_cache_context.py:68
↓ 12 callers
Function
_make_sm_config
Build a small StorageManagerConfig suitable for CPU testing.
tests/cli/commands/trace/test_driver.py:54
↓ 12 callers
Function
_store_completed
( task_id: int, t: float, adapter_index: int = 0, l2_name: str = "fs", bytes_transferred:
tests/v1/mp_observability/subscribers/metrics/test_l2_throughput.py:54
↓ 12 callers
Function
_write_module
(pkg_dir: Path, name: str, source: str)
tests/v1/test_subclass_discovery.py:19
↓ 12 callers
Function
_xpu_sync
()
tests/benchmarks/test_xpu_kernels_microbench.py:26
← previous
next →
401–500 of 13,371, ranked by callers