feat: Add CacheProvider API for external distributed caching (#12056)

* feat: Add CacheProvider API for external distributed caching

Introduces a public API for external cache providers, enabling distributed
caching across multiple ComfyUI instances (e.g., Kubernetes pods).

New files:
- comfy_execution/cache_provider.py: CacheProvider ABC, CacheContext/CacheValue
  dataclasses, thread-safe provider registry, serialization utilities

Modified files:
- comfy_execution/caching.py: Add provider hooks to BasicCache (_notify_providers_store,
  _check_providers_lookup), subcache exclusion, prompt ID propagation
- execution.py: Add prompt lifecycle hooks (on_prompt_start/on_prompt_end) to
  PromptExecutor, set _current_prompt_id on caches

Key features:
- Local-first caching (check local before external for performance)
- NaN detection to prevent incorrect external cache hits
- Subcache exclusion (ephemeral subgraph results not cached externally)
- Thread-safe provider snapshot caching
- Graceful error handling (provider errors logged, never break execution)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* fix: use deterministic hash for cache keys instead of pickle

Pickle serialization is NOT deterministic across Python sessions due
to hash randomization affecting frozenset iteration order. This causes
distributed caching to fail because different pods compute different
hashes for identical cache keys.

Fix: Use _canonicalize() + JSON serialization which ensures deterministic
ordering regardless of Python's hash randomization.

This is critical for cross-pod cache key consistency in Kubernetes
deployments.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* test: add unit tests for CacheProvider API

- Add comprehensive tests for _canonicalize deterministic ordering
- Add tests for serialize_cache_key hash consistency
- Add tests for contains_nan utility
- Add tests for estimate_value_size
- Add tests for provider registry (register, unregister, clear)
- Move json import to top-level (fix inline import)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* style: remove unused imports in test_cache_provider.py

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* fix: move _torch_available before usage and use importlib.util.find_spec

Fixes ruff F821 (undefined name) and F401 (unused import) errors.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* fix: use hashable types in frozenset test and add dict test

Frozensets can only contain hashable types, so use nested frozensets
instead of dicts. Added separate test for dict handling via serialize_cache_key.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* refactor: expose CacheProvider API via comfy_api.latest.Caching

- Add Caching class to comfy_api/latest/__init__.py that re-exports
  from comfy_execution.cache_provider (source of truth)
- Fix docstring: "Skip large values" instead of "Skip small values"
  (small compute-heavy values are good cache targets)
- Maintain backward compatibility: comfy_execution.cache_provider
  imports still work

Usage:
    from comfy_api.latest import Caching

    class MyProvider(Caching.CacheProvider):
        def on_lookup(self, context): ...
        def on_store(self, context, value): ...

    Caching.register_provider(MyProvider())

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* docs: clarify should_cache filtering criteria

Change docstring from "Skip large values" to "Skip if download time > compute time"
which better captures the cost/benefit tradeoff for external caching.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* docs: make should_cache docstring implementation-agnostic

Remove prescriptive filtering suggestions - let implementations
decide their own caching logic based on their use case.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* feat: add optional ui field to CacheValue

- Add ui field to CacheValue dataclass (default None)
- Pass ui when creating CacheValue for external providers
- Use result.ui (or default {}) when returning from external cache lookup

This allows external cache implementations to store/retrieve UI data
if desired, while remaining optional for implementations that skip it.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* refactor: rename _is_cacheable_value to _is_external_cacheable_value

Clearer name since objects are also cached locally - this specifically
checks for external caching eligibility.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

* refactor: async CacheProvider API + reduce public surface

- Make on_lookup/on_store async on CacheProvider ABC
- Simplify CacheContext: replace cache_key + cache_key_bytes with
  cache_key_hash (str hex digest)
- Make registry/utility functions internal (_prefix)
- Trim comfy_api.latest.Caching exports to core API only
- Make cache get/set async throughout caching.py hierarchy
- Use asyncio.create_task for fire-and-forget on_store
- Add NaN gating before provider calls in Core
- Add await to 5 cache call sites in execution.py

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: remove unused imports (ruff) and update tests for internal API

- Remove unused CacheContext and _serialize_cache_key imports from
  caching.py (now handled by _build_context helper)
- Update test_cache_provider.py to use _-prefixed internal names
- Update tests for new CacheContext.cache_key_hash field (str)
- Make MockCacheProvider methods async to match ABC

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: address coderabbit review feedback

- Add try/except to _build_context, return None when hash fails
- Return None from _serialize_cache_key on total failure (no id()-based fallback)
- Replace hex-like test literal with non-secret placeholder

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: use _-prefixed imports in _notify_prompt_lifecycle

The lifecycle notification method was importing the old non-prefixed
names (has_cache_providers, get_cache_providers, logger) which no
longer exist after the API cleanup.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: add sync get_local/set_local for graph traversal

ExecutionList in graph.py calls output_cache.get() and .set() from
sync methods (is_cached, cache_link, get_cache). These cannot await
the now-async get/set. Add get_local/set_local that bypass external
providers and only access the local dict — which is all graph
traversal needs.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* chore: remove cloud-specific language from cache provider API

Make all docstrings and comments generic for the OSS codebase.
Remove references to Kubernetes, Redis, GCS, pods, and other
infrastructure-specific terminology.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* style: align documentation with codebase conventions

Strip verbose docstrings and section banners to match existing minimal
documentation style used throughout the codebase.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: add usage example to Caching class, remove pickle fallback

- Add docstring with usage example to Caching class matching the
  convention used by sibling APIs (Execution.set_progress, ComfyExtension)
- Remove non-deterministic pickle fallback from _serialize_cache_key;
  return None on JSON failure instead of producing unretrievable hashes
- Move cache_provider imports to top of execution.py (no circular dep)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* refactor: move public types to comfy_api, eager provider snapshot

Address review feedback:
- Move CacheProvider/CacheContext/CacheValue definitions to
  comfy_api/latest/_caching.py (source of truth for public API)
- comfy_execution/cache_provider.py re-exports types from there
- Build _providers_snapshot eagerly on register/unregister instead
  of lazy memoization in _get_cache_providers

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: generalize self-inequality check, fail-closed canonicalization

Address review feedback from guill:
- Rename _contains_nan to _contains_self_unequal, use not (x == x)
  instead of math.isnan to catch any self-unequal value
- Remove Unhashable and repr() fallbacks from _canonicalize; raise
  ValueError for unknown types so _serialize_cache_key returns None
  and external caching is skipped (fail-closed)
- Update tests for renamed function and new fail-closed behavior

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: suppress ruff F401 for re-exported CacheContext

CacheContext is imported from _caching and re-exported for use by
caching.py. Add noqa comment to satisfy the linter.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: enable external caching for subcache (expanded) nodes

Subcache nodes (from node expansion) now participate in external
provider store/lookup. Previously skipped to avoid duplicates, but
the cost of missing partial-expansion cache hits outweighs redundant
stores — especially with looping behavior on the horizon.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: wrap register/unregister as explicit static methods

Define register_provider and unregister_provider as wrapper functions
in the Caching class instead of re-importing. This locks the public
API signature in comfy_api/ so internal changes can't accidentally
break it.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: use debug-level logging for provider registration

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: follow ProxiedSingleton pattern for Caching class

Add Caching as a nested class inside ComfyAPI_latest inheriting from
ProxiedSingleton with async instance methods, matching the Execution
and NodeReplacement patterns. Retains standalone Caching class for
direct import convenience.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: inline registration logic in Caching class

Follow the Execution/NodeReplacement pattern — the public API methods
contain the actual logic operating on cache_provider module state,
not wrapper functions delegating to free functions.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: single Caching definition inside ComfyAPI_latest

Remove duplicate standalone Caching class. Define it once as a nested
class in ComfyAPI_latest (matching Execution/NodeReplacement pattern),
with a module-level alias for import convenience.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: remove prompt_id from CacheContext, type-safe canonicalization

Remove prompt_id from CacheContext — it's not relevant for cache
matching and added unnecessary plumbing (_current_prompt_id on every
cache). Lifecycle hooks still receive prompt_id directly.

Include type name in canonicalized primitives so that int 7 and
str "7" produce distinct hashes. Also canonicalize dict keys properly
instead of str() coercion.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: address review feedback on cache provider API

- Hold references to pending store tasks to prevent "Task was destroyed
  but it is still pending" warnings (bigcat88)
- Parallel cache lookups with asyncio.gather instead of sequential
  awaits for better performance (bigcat88)
- Delegate Caching.register/unregister_provider to existing functions
  in cache_provider.py instead of reimplementing (bigcat88)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Claude <noreply@anthropic.com>
This commit is contained in:
Deep Mehta
2026-03-12 16:09:07 -07:00
committed by GitHub
parent 8d9faaa181
commit af7b4a921d
7 changed files with 859 additions and 83 deletions

View File

@@ -0,0 +1,138 @@
from typing import Any, Optional, Tuple, List
import hashlib
import json
import logging
import threading
# Public types — source of truth is comfy_api.latest._caching
from comfy_api.latest._caching import CacheProvider, CacheContext, CacheValue # noqa: F401 (re-exported)
_logger = logging.getLogger(__name__)
_providers: List[CacheProvider] = []
_providers_lock = threading.Lock()
_providers_snapshot: Tuple[CacheProvider, ...] = ()
def register_cache_provider(provider: CacheProvider) -> None:
"""Register an external cache provider. Providers are called in registration order."""
global _providers_snapshot
with _providers_lock:
if provider in _providers:
_logger.warning(f"Provider {provider.__class__.__name__} already registered")
return
_providers.append(provider)
_providers_snapshot = tuple(_providers)
_logger.debug(f"Registered cache provider: {provider.__class__.__name__}")
def unregister_cache_provider(provider: CacheProvider) -> None:
global _providers_snapshot
with _providers_lock:
try:
_providers.remove(provider)
_providers_snapshot = tuple(_providers)
_logger.debug(f"Unregistered cache provider: {provider.__class__.__name__}")
except ValueError:
_logger.warning(f"Provider {provider.__class__.__name__} was not registered")
def _get_cache_providers() -> Tuple[CacheProvider, ...]:
return _providers_snapshot
def _has_cache_providers() -> bool:
return bool(_providers_snapshot)
def _clear_cache_providers() -> None:
global _providers_snapshot
with _providers_lock:
_providers.clear()
_providers_snapshot = ()
def _canonicalize(obj: Any) -> Any:
# Convert to canonical JSON-serializable form with deterministic ordering.
# Frozensets have non-deterministic iteration order between Python sessions.
# Raises ValueError for non-cacheable types (Unhashable, unknown) so that
# _serialize_cache_key returns None and external caching is skipped.
if isinstance(obj, frozenset):
return ("__frozenset__", sorted(
[_canonicalize(item) for item in obj],
key=lambda x: json.dumps(x, sort_keys=True)
))
elif isinstance(obj, set):
return ("__set__", sorted(
[_canonicalize(item) for item in obj],
key=lambda x: json.dumps(x, sort_keys=True)
))
elif isinstance(obj, tuple):
return ("__tuple__", [_canonicalize(item) for item in obj])
elif isinstance(obj, list):
return [_canonicalize(item) for item in obj]
elif isinstance(obj, dict):
return {"__dict__": sorted(
[[_canonicalize(k), _canonicalize(v)] for k, v in obj.items()],
key=lambda x: json.dumps(x, sort_keys=True)
)}
elif isinstance(obj, (int, float, str, bool, type(None))):
return (type(obj).__name__, obj)
elif isinstance(obj, bytes):
return ("__bytes__", obj.hex())
else:
raise ValueError(f"Cannot canonicalize type: {type(obj).__name__}")
def _serialize_cache_key(cache_key: Any) -> Optional[str]:
# Returns deterministic SHA256 hex digest, or None on failure.
# Uses JSON (not pickle) because pickle is non-deterministic across sessions.
try:
canonical = _canonicalize(cache_key)
json_str = json.dumps(canonical, sort_keys=True, separators=(',', ':'))
return hashlib.sha256(json_str.encode('utf-8')).hexdigest()
except Exception as e:
_logger.warning(f"Failed to serialize cache key: {e}")
return None
def _contains_self_unequal(obj: Any) -> bool:
# Local cache matches by ==. Values where not (x == x) (NaN, etc.) will
# never hit locally, but serialized form would match externally. Skip these.
try:
if not (obj == obj):
return True
except Exception:
return True
if isinstance(obj, (frozenset, tuple, list, set)):
return any(_contains_self_unequal(item) for item in obj)
if isinstance(obj, dict):
return any(_contains_self_unequal(k) or _contains_self_unequal(v) for k, v in obj.items())
if hasattr(obj, 'value'):
return _contains_self_unequal(obj.value)
return False
def _estimate_value_size(value: CacheValue) -> int:
try:
import torch
except ImportError:
return 0
total = 0
def estimate(obj):
nonlocal total
if isinstance(obj, torch.Tensor):
total += obj.numel() * obj.element_size()
elif isinstance(obj, dict):
for v in obj.values():
estimate(v)
elif isinstance(obj, (list, tuple)):
for item in obj:
estimate(item)
for output in value.outputs:
estimate(output)
return total

View File

@@ -1,3 +1,4 @@
import asyncio
import bisect
import gc
import itertools
@@ -154,6 +155,7 @@ class BasicCache:
self.cache_key_set: CacheKeySet
self.cache = {}
self.subcaches = {}
self._pending_store_tasks: set = set()
async def set_prompt(self, dynprompt, node_ids, is_changed_cache):
self.dynprompt = dynprompt
@@ -196,18 +198,134 @@ class BasicCache:
def poll(self, **kwargs):
pass
def _set_immediate(self, node_id, value):
assert self.initialized
cache_key = self.cache_key_set.get_data_key(node_id)
self.cache[cache_key] = value
def _get_immediate(self, node_id):
def get_local(self, node_id):
if not self.initialized:
return None
cache_key = self.cache_key_set.get_data_key(node_id)
if cache_key in self.cache:
return self.cache[cache_key]
else:
return None
def set_local(self, node_id, value):
assert self.initialized
cache_key = self.cache_key_set.get_data_key(node_id)
self.cache[cache_key] = value
async def _set_immediate(self, node_id, value):
assert self.initialized
cache_key = self.cache_key_set.get_data_key(node_id)
self.cache[cache_key] = value
await self._notify_providers_store(node_id, cache_key, value)
async def _get_immediate(self, node_id):
if not self.initialized:
return None
cache_key = self.cache_key_set.get_data_key(node_id)
if cache_key in self.cache:
return self.cache[cache_key]
external_result = await self._check_providers_lookup(node_id, cache_key)
if external_result is not None:
self.cache[cache_key] = external_result
return external_result
return None
async def _notify_providers_store(self, node_id, cache_key, value):
from comfy_execution.cache_provider import (
_has_cache_providers, _get_cache_providers,
CacheValue, _contains_self_unequal, _logger
)
if not _has_cache_providers():
return
if not self._is_external_cacheable_value(value):
return
if _contains_self_unequal(cache_key):
return
context = self._build_context(node_id, cache_key)
if context is None:
return
cache_value = CacheValue(outputs=value.outputs, ui=value.ui)
for provider in _get_cache_providers():
try:
if provider.should_cache(context, cache_value):
task = asyncio.create_task(self._safe_provider_store(provider, context, cache_value))
self._pending_store_tasks.add(task)
task.add_done_callback(self._pending_store_tasks.discard)
except Exception as e:
_logger.warning(f"Cache provider {provider.__class__.__name__} error on store: {e}")
@staticmethod
async def _safe_provider_store(provider, context, cache_value):
from comfy_execution.cache_provider import _logger
try:
await provider.on_store(context, cache_value)
except Exception as e:
_logger.warning(f"Cache provider {provider.__class__.__name__} async store error: {e}")
async def _check_providers_lookup(self, node_id, cache_key):
from comfy_execution.cache_provider import (
_has_cache_providers, _get_cache_providers,
CacheValue, _contains_self_unequal, _logger
)
if not _has_cache_providers():
return None
if _contains_self_unequal(cache_key):
return None
context = self._build_context(node_id, cache_key)
if context is None:
return None
for provider in _get_cache_providers():
try:
if not provider.should_cache(context):
continue
result = await provider.on_lookup(context)
if result is not None:
if not isinstance(result, CacheValue):
_logger.warning(f"Provider {provider.__class__.__name__} returned invalid type")
continue
if not isinstance(result.outputs, (list, tuple)):
_logger.warning(f"Provider {provider.__class__.__name__} returned invalid outputs")
continue
from execution import CacheEntry
return CacheEntry(ui=result.ui or {}, outputs=list(result.outputs))
except Exception as e:
_logger.warning(f"Cache provider {provider.__class__.__name__} error on lookup: {e}")
return None
def _is_external_cacheable_value(self, value):
return hasattr(value, 'outputs') and hasattr(value, 'ui')
def _get_class_type(self, node_id):
if not self.initialized or not self.dynprompt:
return ''
try:
return self.dynprompt.get_node(node_id).get('class_type', '')
except Exception:
return ''
def _build_context(self, node_id, cache_key):
from comfy_execution.cache_provider import CacheContext, _serialize_cache_key, _logger
try:
cache_key_hash = _serialize_cache_key(cache_key)
if cache_key_hash is None:
return None
return CacheContext(
node_id=node_id,
class_type=self._get_class_type(node_id),
cache_key_hash=cache_key_hash,
)
except Exception as e:
_logger.warning(f"Failed to build cache context for node {node_id}: {e}")
return None
async def _ensure_subcache(self, node_id, children_ids):
@@ -257,16 +375,27 @@ class HierarchicalCache(BasicCache):
return None
return cache
def get(self, node_id):
async def get(self, node_id):
cache = self._get_cache_for(node_id)
if cache is None:
return None
return cache._get_immediate(node_id)
return await cache._get_immediate(node_id)
def set(self, node_id, value):
def get_local(self, node_id):
cache = self._get_cache_for(node_id)
if cache is None:
return None
return BasicCache.get_local(cache, node_id)
async def set(self, node_id, value):
cache = self._get_cache_for(node_id)
assert cache is not None
cache._set_immediate(node_id, value)
await cache._set_immediate(node_id, value)
def set_local(self, node_id, value):
cache = self._get_cache_for(node_id)
assert cache is not None
BasicCache.set_local(cache, node_id, value)
async def ensure_subcache_for(self, node_id, children_ids):
cache = self._get_cache_for(node_id)
@@ -287,10 +416,16 @@ class NullCache:
def poll(self, **kwargs):
pass
def get(self, node_id):
async def get(self, node_id):
return None
def set(self, node_id, value):
def get_local(self, node_id):
return None
async def set(self, node_id, value):
pass
def set_local(self, node_id, value):
pass
async def ensure_subcache_for(self, node_id, children_ids):
@@ -322,18 +457,18 @@ class LRUCache(BasicCache):
del self.children[key]
self._clean_subcaches()
def get(self, node_id):
async def get(self, node_id):
self._mark_used(node_id)
return self._get_immediate(node_id)
return await self._get_immediate(node_id)
def _mark_used(self, node_id):
cache_key = self.cache_key_set.get_data_key(node_id)
if cache_key is not None:
self.used_generation[cache_key] = self.generation
def set(self, node_id, value):
async def set(self, node_id, value):
self._mark_used(node_id)
return self._set_immediate(node_id, value)
return await self._set_immediate(node_id, value)
async def ensure_subcache_for(self, node_id, children_ids):
# Just uses subcaches for tracking 'live' nodes
@@ -373,13 +508,13 @@ class RAMPressureCache(LRUCache):
def clean_unused(self):
self._clean_subcaches()
def set(self, node_id, value):
async def set(self, node_id, value):
self.timestamps[self.cache_key_set.get_data_key(node_id)] = time.time()
super().set(node_id, value)
await super().set(node_id, value)
def get(self, node_id):
async def get(self, node_id):
self.timestamps[self.cache_key_set.get_data_key(node_id)] = time.time()
return super().get(node_id)
return await super().get(node_id)
def poll(self, ram_headroom):
def _ram_gb():

View File

@@ -204,12 +204,12 @@ class ExecutionList(TopologicalSort):
self.execution_cache_listeners = {}
def is_cached(self, node_id):
return self.output_cache.get(node_id) is not None
return self.output_cache.get_local(node_id) is not None
def cache_link(self, from_node_id, to_node_id):
if to_node_id not in self.execution_cache:
self.execution_cache[to_node_id] = {}
self.execution_cache[to_node_id][from_node_id] = self.output_cache.get(from_node_id)
self.execution_cache[to_node_id][from_node_id] = self.output_cache.get_local(from_node_id)
if from_node_id not in self.execution_cache_listeners:
self.execution_cache_listeners[from_node_id] = set()
self.execution_cache_listeners[from_node_id].add(to_node_id)
@@ -221,7 +221,7 @@ class ExecutionList(TopologicalSort):
if value is None:
return None
#Write back to the main cache on touch.
self.output_cache.set(from_node_id, value)
self.output_cache.set_local(from_node_id, value)
return value
def cache_update(self, node_id, value):