agent_os_kernel 3.1.0__py3-none-any.whl

caas/caching.py

@@ -0,0 +1,834 @@
+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT License.
+"""
+Context Caching for LLM APIs.
+
+This module provides intelligent caching for context sent to LLM APIs,
+leveraging provider-specific caching features (Anthropic's prompt caching,
+OpenAI's predicted outputs) and local caching strategies.
+
+Key Features:
+    - Provider-agnostic caching interface
+    - Anthropic prompt caching support (cache_control breakpoints)
+    - OpenAI predicted outputs / prefix caching detection
+    - Local semantic cache for repeated queries
+    - Cache statistics and cost tracking
+    - TTL-based cache expiration
+
+Cost Savings:
+    - Anthropic: Up to 90% reduction on cached prompt tokens
+    - OpenAI: 50% reduction on cached prefix tokens
+    - Local cache: 100% reduction for exact/semantic matches
+
+Example:
+    from caas.caching import (
+        ContextCache,
+        AnthropicCacheStrategy,
+        OpenAICacheStrategy,
+        CacheConfig,
+    )
+
+    # Create cache with Anthropic strategy
+    cache = ContextCache(
+        strategy=AnthropicCacheStrategy(),
+        config=CacheConfig(ttl_seconds=3600)
+    )
+
+    # Prepare messages with cache breakpoints
+    messages = cache.prepare_messages(
+        system_prompt="You are a helpful assistant...",
+        context="Large document context here...",
+        user_message="Summarize the key points"
+    )
+
+    # Track cache hits
+    stats = cache.get_stats()
+    print(f"Cache hit rate: {stats['hit_rate']:.1%}")
+    print(f"Estimated savings: ${stats['estimated_savings']:.2f}")
+"""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import time
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from datetime import datetime, timezone
+from enum import Enum
+from typing import Any, Dict, List, Optional, Protocol, Tuple, Union
+from collections import OrderedDict
+import threading
+
+
+class CacheProvider(str, Enum):
+    """Supported LLM providers for caching."""
+    ANTHROPIC = "anthropic"
+    OPENAI = "openai"
+    LOCAL = "local"  # Local-only caching
+
+
+class CacheType(str, Enum):
+    """Types of cache hits."""
+    PROVIDER_CACHE = "provider_cache"  # Provider-side caching (Anthropic/OpenAI)
+    LOCAL_EXACT = "local_exact"  # Local exact match
+    LOCAL_SEMANTIC = "local_semantic"  # Local semantic similarity match
+    MISS = "miss"  # No cache hit
+
+
+@dataclass
+class CacheConfig:
+    """Configuration for context caching.
+    
+    Attributes:
+        ttl_seconds: Time-to-live for cached entries (default: 1 hour)
+        max_entries: Maximum number of entries in local cache
+        min_tokens_for_caching: Minimum tokens to consider caching
+        semantic_threshold: Similarity threshold for semantic cache (0-1)
+        enable_provider_cache: Whether to use provider-specific caching
+        enable_local_cache: Whether to use local caching
+        track_costs: Whether to track cost savings
+    """
+    ttl_seconds: int = 3600
+    max_entries: int = 1000
+    min_tokens_for_caching: int = 1024  # Only cache contexts >= 1024 tokens
+    semantic_threshold: float = 0.95
+    enable_provider_cache: bool = True
+    enable_local_cache: bool = True
+    track_costs: bool = True
+
+
+@dataclass
+class CacheEntry:
+    """A single cache entry.
+    
+    Attributes:
+        key: Cache key (hash of content)
+        content: The cached content
+        metadata: Additional metadata
+        created_at: When the entry was created
+        last_accessed: When the entry was last accessed
+        access_count: Number of times this entry was accessed
+        token_count: Estimated token count
+    """
+    key: str
+    content: str
+    metadata: Dict[str, Any]
+    created_at: datetime
+    last_accessed: datetime
+    access_count: int = 1
+    token_count: int = 0
+    
+    def is_expired(self, ttl_seconds: int) -> bool:
+        """Check if this entry has expired."""
+        age = (datetime.now(timezone.utc) - self.created_at).total_seconds()
+        return age > ttl_seconds
+
+
+@dataclass
+class CacheResult:
+    """Result of a cache lookup.
+    
+    Attributes:
+        cache_type: Type of cache hit (or miss)
+        cached_content: The cached content (if hit)
+        cache_key: Key used for caching
+        token_savings: Estimated token savings
+        cost_savings: Estimated cost savings in USD
+    """
+    cache_type: CacheType
+    cached_content: Optional[str] = None
+    cache_key: Optional[str] = None
+    token_savings: int = 0
+    cost_savings: float = 0.0
+
+
+@dataclass
+class CacheStats:
+    """Statistics for cache performance.
+    
+    Attributes:
+        total_requests: Total number of cache requests
+        provider_hits: Hits from provider cache
+        local_exact_hits: Hits from local exact match
+        local_semantic_hits: Hits from local semantic match
+        misses: Cache misses
+        total_tokens_saved: Total tokens saved
+        total_cost_saved: Total cost saved in USD
+        current_entries: Current number of cache entries
+    """
+    total_requests: int = 0
+    provider_hits: int = 0
+    local_exact_hits: int = 0
+    local_semantic_hits: int = 0
+    misses: int = 0
+    total_tokens_saved: int = 0
+    total_cost_saved: float = 0.0
+    current_entries: int = 0
+    
+    @property
+    def hit_rate(self) -> float:
+        """Calculate overall cache hit rate."""
+        if self.total_requests == 0:
+            return 0.0
+        hits = self.provider_hits + self.local_exact_hits + self.local_semantic_hits
+        return hits / self.total_requests
+    
+    def to_dict(self) -> Dict[str, Any]:
+        """Convert to dictionary."""
+        return {
+            "total_requests": self.total_requests,
+            "provider_hits": self.provider_hits,
+            "local_exact_hits": self.local_exact_hits,
+            "local_semantic_hits": self.local_semantic_hits,
+            "misses": self.misses,
+            "hit_rate": self.hit_rate,
+            "total_tokens_saved": self.total_tokens_saved,
+            "total_cost_saved": self.total_cost_saved,
+            "current_entries": self.current_entries,
+        }
+
+
+class CacheStrategy(ABC):
+    """Abstract base class for provider-specific caching strategies."""
+    
+    @property
+    @abstractmethod
+    def provider(self) -> CacheProvider:
+        """Return the provider this strategy is for."""
+        pass
+    
+    @abstractmethod
+    def prepare_messages(
+        self,
+        system_prompt: Optional[str],
+        context: str,
+        messages: List[Dict[str, Any]],
+    ) -> List[Dict[str, Any]]:
+        """
+        Prepare messages with caching hints.
+        
+        Args:
+            system_prompt: Optional system prompt
+            context: The context to cache
+            messages: The conversation messages
+            
+        Returns:
+            Messages formatted for the provider with cache hints
+        """
+        pass
+    
+    @abstractmethod
+    def estimate_savings(self, token_count: int, is_cached: bool) -> Tuple[int, float]:
+        """
+        Estimate token and cost savings from caching.
+        
+        Args:
+            token_count: Number of tokens in the cached content
+            is_cached: Whether the content was cached
+            
+        Returns:
+            Tuple of (tokens_saved, cost_saved_usd)
+        """
+        pass
+
+
+class AnthropicCacheStrategy(CacheStrategy):
+    """
+    Caching strategy for Anthropic API.
+    
+    Implements Anthropic's prompt caching feature which provides:
+    - 90% cost reduction on cached input tokens
+    - 5-minute TTL for cached prompts
+    - Requires min 1024 tokens for caching (2048 for Claude 3.5 Haiku)
+    
+    See: https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching
+    """
+    
+    # Anthropic pricing (as of 2024)
+    # Claude 3.5 Sonnet: $3/M input, $0.30/M cached input
+    INPUT_COST_PER_MILLION = 3.00
+    CACHED_COST_PER_MILLION = 0.30
+    CACHE_WRITE_COST_PER_MILLION = 3.75  # 25% premium for writing to cache
+    
+    MIN_TOKENS_FOR_CACHE = 1024  # Minimum tokens for caching
+    
+    @property
+    def provider(self) -> CacheProvider:
+        return CacheProvider.ANTHROPIC
+    
+    def prepare_messages(
+        self,
+        system_prompt: Optional[str],
+        context: str,
+        messages: List[Dict[str, Any]],
+    ) -> List[Dict[str, Any]]:
+        """
+        Prepare messages with Anthropic cache_control breakpoints.
+        
+        Adds cache_control: {"type": "ephemeral"} to cacheable content blocks.
+        """
+        prepared = []
+        
+        # System prompt with caching (if provided and long enough)
+        if system_prompt:
+            system_tokens = self._estimate_tokens(system_prompt)
+            system_content: List[Dict[str, Any]] = [{"type": "text", "text": system_prompt}]
+            
+            if system_tokens >= self.MIN_TOKENS_FOR_CACHE:
+                system_content[0]["cache_control"] = {"type": "ephemeral"}
+            
+            # Add context as second block with caching
+            if context:
+                context_block: Dict[str, Any] = {"type": "text", "text": context}
+                context_tokens = self._estimate_tokens(context)
+                if context_tokens >= self.MIN_TOKENS_FOR_CACHE:
+                    context_block["cache_control"] = {"type": "ephemeral"}
+                system_content.append(context_block)
+            
+            prepared.append({
+                "role": "system",
+                "content": system_content,
+            })
+        elif context:
+            # Context only (no system prompt)
+            context_content: List[Dict[str, Any]] = [{"type": "text", "text": context}]
+            context_tokens = self._estimate_tokens(context)
+            if context_tokens >= self.MIN_TOKENS_FOR_CACHE:
+                context_content[0]["cache_control"] = {"type": "ephemeral"}
+            
+            prepared.append({
+                "role": "system",
+                "content": context_content,
+            })
+        
+        # Add conversation messages
+        prepared.extend(messages)
+        
+        return prepared
+    
+    def estimate_savings(self, token_count: int, is_cached: bool) -> Tuple[int, float]:
+        """
+        Estimate savings from Anthropic caching.
+        
+        Returns:
+            Tuple of (tokens_saved, cost_saved_usd)
+        """
+        if not is_cached or token_count < self.MIN_TOKENS_FOR_CACHE:
+            return (0, 0.0)
+        
+        # Cost without caching
+        normal_cost = (token_count / 1_000_000) * self.INPUT_COST_PER_MILLION
+        
+        # Cost with caching (90% reduction)
+        cached_cost = (token_count / 1_000_000) * self.CACHED_COST_PER_MILLION
+        
+        # Savings
+        cost_saved = normal_cost - cached_cost
+        
+        # Token savings (conceptually, same tokens but cheaper)
+        tokens_saved = int(token_count * 0.9)  # 90% "saved" in cost terms
+        
+        return (tokens_saved, cost_saved)
+    
+    def _estimate_tokens(self, text: str) -> int:
+        """Rough token estimation (4 chars per token)."""
+        return len(text) // 4
+
+
+class OpenAICacheStrategy(CacheStrategy):
+    """
+    Caching strategy for OpenAI API.
+    
+    Implements OpenAI's automatic prompt caching which provides:
+    - 50% cost reduction on cached input tokens
+    - Automatic caching for prompts > 1024 tokens
+    - Cache expires after 5-60 minutes of inactivity
+    
+    See: https://platform.openai.com/docs/guides/prompt-caching
+    """
+    
+    # OpenAI pricing (GPT-4o as of 2024)
+    INPUT_COST_PER_MILLION = 2.50
+    CACHED_COST_PER_MILLION = 1.25  # 50% discount
+    
+    MIN_TOKENS_FOR_CACHE = 1024
+    
+    @property
+    def provider(self) -> CacheProvider:
+        return CacheProvider.OPENAI
+    
+    def prepare_messages(
+        self,
+        system_prompt: Optional[str],
+        context: str,
+        messages: List[Dict[str, Any]],
+    ) -> List[Dict[str, Any]]:
+        """
+        Prepare messages for OpenAI (no special formatting needed).
+        
+        OpenAI caching is automatic - we just need to structure messages
+        with static content first for optimal cache hits.
+        """
+        prepared = []
+        
+        # Combine system prompt and context (static content first)
+        if system_prompt or context:
+            system_text = ""
+            if system_prompt:
+                system_text += system_prompt
+            if context:
+                if system_text:
+                    system_text += "\n\n---\n\n"
+                system_text += context
+            
+            prepared.append({
+                "role": "system",
+                "content": system_text,
+            })
+        
+        # Add conversation messages
+        prepared.extend(messages)
+        
+        return prepared
+    
+    def estimate_savings(self, token_count: int, is_cached: bool) -> Tuple[int, float]:
+        """
+        Estimate savings from OpenAI caching.
+        
+        Returns:
+            Tuple of (tokens_saved, cost_saved_usd)
+        """
+        if not is_cached or token_count < self.MIN_TOKENS_FOR_CACHE:
+            return (0, 0.0)
+        
+        # Cost without caching
+        normal_cost = (token_count / 1_000_000) * self.INPUT_COST_PER_MILLION
+        
+        # Cost with caching (50% reduction)
+        cached_cost = (token_count / 1_000_000) * self.CACHED_COST_PER_MILLION
+        
+        # Savings
+        cost_saved = normal_cost - cached_cost
+        
+        # Token savings (50% in cost terms)
+        tokens_saved = int(token_count * 0.5)
+        
+        return (tokens_saved, cost_saved)
+
+
+class LocalCacheStrategy(CacheStrategy):
+    """
+    Local-only caching strategy.
+    
+    Uses a local LRU cache to store and retrieve responses.
+    Provides 100% savings on exact matches.
+    """
+    
+    @property
+    def provider(self) -> CacheProvider:
+        return CacheProvider.LOCAL
+    
+    def prepare_messages(
+        self,
+        system_prompt: Optional[str],
+        context: str,
+        messages: List[Dict[str, Any]],
+    ) -> List[Dict[str, Any]]:
+        """No special preparation needed for local caching."""
+        prepared = []
+        
+        if system_prompt or context:
+            system_text = ""
+            if system_prompt:
+                system_text += system_prompt
+            if context:
+                if system_text:
+                    system_text += "\n\n"
+                system_text += context
+            
+            prepared.append({
+                "role": "system",
+                "content": system_text,
+            })
+        
+        prepared.extend(messages)
+        return prepared
+    
+    def estimate_savings(self, token_count: int, is_cached: bool) -> Tuple[int, float]:
+        """100% savings on local cache hits."""
+        if not is_cached:
+            return (0, 0.0)
+        
+        # Assume average cost of $2/M tokens
+        cost_saved = (token_count / 1_000_000) * 2.00
+        return (token_count, cost_saved)
+
+
+class LRUCache:
+    """Thread-safe LRU cache implementation."""
+    
+    def __init__(self, max_size: int = 1000):
+        self._cache: OrderedDict[str, CacheEntry] = OrderedDict()
+        self._max_size = max_size
+        self._lock = threading.RLock()
+    
+    def get(self, key: str, ttl_seconds: int) -> Optional[CacheEntry]:
+        """Get an entry from cache."""
+        with self._lock:
+            if key not in self._cache:
+                return None
+            
+            entry = self._cache[key]
+            
+            # Check expiration
+            if entry.is_expired(ttl_seconds):
+                del self._cache[key]
+                return None
+            
+            # Move to end (most recently used)
+            self._cache.move_to_end(key)
+            
+            # Update access stats
+            entry.access_count += 1
+            entry.last_accessed = datetime.now(timezone.utc)
+            
+            return entry
+    
+    def put(self, entry: CacheEntry) -> None:
+        """Add or update an entry in cache."""
+        with self._lock:
+            # Remove oldest entries if at capacity
+            while len(self._cache) >= self._max_size:
+                self._cache.popitem(last=False)
+            
+            self._cache[entry.key] = entry
+            self._cache.move_to_end(entry.key)
+    
+    def remove(self, key: str) -> bool:
+        """Remove an entry from cache."""
+        with self._lock:
+            if key in self._cache:
+                del self._cache[key]
+                return True
+            return False
+    
+    def clear(self) -> None:
+        """Clear all entries."""
+        with self._lock:
+            self._cache.clear()
+    
+    def __len__(self) -> int:
+        with self._lock:
+            return len(self._cache)
+    
+    def cleanup_expired(self, ttl_seconds: int) -> int:
+        """Remove expired entries. Returns count of removed entries."""
+        with self._lock:
+            expired_keys = [
+                key for key, entry in self._cache.items()
+                if entry.is_expired(ttl_seconds)
+            ]
+            for key in expired_keys:
+                del self._cache[key]
+            return len(expired_keys)
+
+
+class ContextCache:
+    """
+    Main context caching class.
+    
+    Provides unified caching interface for LLM APIs with support for:
+    - Provider-specific caching (Anthropic, OpenAI)
+    - Local LRU cache for exact matches
+    - Statistics and cost tracking
+    
+    Example:
+        cache = ContextCache(
+            strategy=AnthropicCacheStrategy(),
+            config=CacheConfig(ttl_seconds=3600)
+        )
+        
+        # Check for cached response
+        result = cache.lookup(context_hash)
+        
+        if result.cache_type == CacheType.MISS:
+            # Make API call
+            response = api.complete(...)
+            cache.store(context_hash, response)
+    """
+    
+    def __init__(
+        self,
+        strategy: Optional[CacheStrategy] = None,
+        config: Optional[CacheConfig] = None,
+    ):
+        """
+        Initialize the context cache.
+        
+        Args:
+            strategy: Provider-specific caching strategy
+            config: Cache configuration
+        """
+        self._strategy = strategy or LocalCacheStrategy()
+        self._config = config or CacheConfig()
+        self._local_cache = LRUCache(max_size=self._config.max_entries)
+        self._stats = CacheStats()
+        self._lock = threading.RLock()
+    
+    @property
+    def strategy(self) -> CacheStrategy:
+        """Get the current caching strategy."""
+        return self._strategy
+    
+    @property
+    def config(self) -> CacheConfig:
+        """Get the cache configuration."""
+        return self._config
+    
+    def compute_key(self, content: str, metadata: Optional[Dict[str, Any]] = None) -> str:
+        """
+        Compute a cache key for content.
+        
+        Args:
+            content: The content to hash
+            metadata: Optional metadata to include in key
+            
+        Returns:
+            SHA-256 hash of the content
+        """
+        key_data = content
+        if metadata:
+            key_data += json.dumps(metadata, sort_keys=True)
+        
+        return hashlib.sha256(key_data.encode()).hexdigest()[:32]
+    
+    def lookup(
+        self,
+        context: str,
+        metadata: Optional[Dict[str, Any]] = None,
+    ) -> CacheResult:
+        """
+        Look up content in cache.
+        
+        Args:
+            context: The context to look up
+            metadata: Optional metadata for key computation
+            
+        Returns:
+            CacheResult with cache type and any cached content
+        """
+        with self._lock:
+            self._stats.total_requests += 1
+        
+        cache_key = self.compute_key(context, metadata)
+        token_count = len(context) // 4  # Rough estimate
+        
+        # Skip caching for small contexts
+        if token_count < self._config.min_tokens_for_caching:
+            with self._lock:
+                self._stats.misses += 1
+            return CacheResult(
+                cache_type=CacheType.MISS,
+                cache_key=cache_key,
+            )
+        
+        # Check local cache first
+        if self._config.enable_local_cache:
+            entry = self._local_cache.get(cache_key, self._config.ttl_seconds)
+            if entry:
+                tokens_saved, cost_saved = self._strategy.estimate_savings(
+                    entry.token_count, True
+                )
+                
+                with self._lock:
+                    self._stats.local_exact_hits += 1
+                    self._stats.total_tokens_saved += tokens_saved
+                    self._stats.total_cost_saved += cost_saved
+                
+                return CacheResult(
+                    cache_type=CacheType.LOCAL_EXACT,
+                    cached_content=entry.content,
+                    cache_key=cache_key,
+                    token_savings=tokens_saved,
+                    cost_savings=cost_saved,
+                )
+        
+        # No local hit - check if provider caching is enabled
+        if self._config.enable_provider_cache:
+            # Provider caching is handled at message preparation time
+            # We return a result indicating provider cache should be used
+            with self._lock:
+                self._stats.provider_hits += 1
+            
+            return CacheResult(
+                cache_type=CacheType.PROVIDER_CACHE,
+                cache_key=cache_key,
+            )
+        
+        # Cache miss
+        with self._lock:
+            self._stats.misses += 1
+        
+        return CacheResult(
+            cache_type=CacheType.MISS,
+            cache_key=cache_key,
+        )
+    
+    def store(
+        self,
+        context: str,
+        response: Optional[str] = None,
+        metadata: Optional[Dict[str, Any]] = None,
+    ) -> str:
+        """
+        Store content in cache.
+        
+        Args:
+            context: The context to cache
+            response: Optional response to cache with context
+            metadata: Optional metadata
+            
+        Returns:
+            Cache key
+        """
+        cache_key = self.compute_key(context, metadata)
+        token_count = len(context) // 4
+        
+        # Don't cache small contexts
+        if token_count < self._config.min_tokens_for_caching:
+            return cache_key
+        
+        if self._config.enable_local_cache:
+            now = datetime.now(timezone.utc)
+            entry = CacheEntry(
+                key=cache_key,
+                content=response or context,
+                metadata=metadata or {},
+                created_at=now,
+                last_accessed=now,
+                token_count=token_count,
+            )
+            self._local_cache.put(entry)
+            
+            with self._lock:
+                self._stats.current_entries = len(self._local_cache)
+        
+        return cache_key
+    
+    def prepare_messages(
+        self,
+        system_prompt: Optional[str] = None,
+        context: str = "",
+        messages: Optional[List[Dict[str, Any]]] = None,
+    ) -> List[Dict[str, Any]]:
+        """
+        Prepare messages with caching hints.
+        
+        Uses the configured strategy to format messages for optimal caching.
+        
+        Args:
+            system_prompt: Optional system prompt
+            context: Context to include (will be cached if large enough)
+            messages: Conversation messages
+            
+        Returns:
+            List of messages formatted for the provider
+        """
+        return self._strategy.prepare_messages(
+            system_prompt=system_prompt,
+            context=context,
+            messages=messages or [],
+        )
+    
+    def invalidate(self, cache_key: str) -> bool:
+        """
+        Invalidate a cache entry.
+        
+        Args:
+            cache_key: Key of the entry to invalidate
+            
+        Returns:
+            True if entry was removed, False if not found
+        """
+        removed = self._local_cache.remove(cache_key)
+        if removed:
+            with self._lock:
+                self._stats.current_entries = len(self._local_cache)
+        return removed
+    
+    def clear(self) -> None:
+        """Clear all cache entries."""
+        self._local_cache.clear()
+        with self._lock:
+            self._stats.current_entries = 0
+    
+    def cleanup(self) -> int:
+        """
+        Remove expired entries.
+        
+        Returns:
+            Number of entries removed
+        """
+        removed = self._local_cache.cleanup_expired(self._config.ttl_seconds)
+        with self._lock:
+            self._stats.current_entries = len(self._local_cache)
+        return removed
+    
+    def get_stats(self) -> Dict[str, Any]:
+        """
+        Get cache statistics.
+        
+        Returns:
+            Dictionary with cache statistics
+        """
+        with self._lock:
+            self._stats.current_entries = len(self._local_cache)
+            return self._stats.to_dict()
+    
+    def reset_stats(self) -> None:
+        """Reset cache statistics."""
+        with self._lock:
+            current_entries = len(self._local_cache)
+            self._stats = CacheStats(current_entries=current_entries)
+
+
+# Convenience function for creating caches
+def create_cache(
+    provider: Union[str, CacheProvider] = CacheProvider.LOCAL,
+    **kwargs,
+) -> ContextCache:
+    """
+    Create a context cache for a specific provider.
+    
+    Args:
+        provider: Provider name or CacheProvider enum
+        **kwargs: Additional config options passed to CacheConfig
+        
+    Returns:
+        Configured ContextCache instance
+    
+    Example:
+        # Anthropic cache
+        cache = create_cache("anthropic", ttl_seconds=3600)
+        
+        # OpenAI cache
+        cache = create_cache("openai", max_entries=500)
+        
+        # Local-only cache
+        cache = create_cache("local")
+    """
+    if isinstance(provider, str):
+        provider = CacheProvider(provider.lower())
+    
+    strategy: CacheStrategy
+    if provider == CacheProvider.ANTHROPIC:
+        strategy = AnthropicCacheStrategy()
+    elif provider == CacheProvider.OPENAI:
+        strategy = OpenAICacheStrategy()
+    else:
+        strategy = LocalCacheStrategy()
+    
+    config = CacheConfig(**kwargs)
+    
+    return ContextCache(strategy=strategy, config=config)