Adapters

Chengeta AI adapters provide drop-in integrations with popular AI and agent frameworks. Each adapter implements the framework's native cache or agent interface, so you get caching without changing your existing code.

Overview

Adapters bridge the gap between Chengeta AI's cache engine and the framework-specific APIs that each AI library expects. Instead of writing custom glue code, you instantiate an adapter with a CacheManager and plug it into the framework's standard extension point.

There are two adapter styles:

• Interface adapters (LangChain, LangGraph) — subclass the framework's cache/checkpointer base class and implement its required methods.
• Wrapper adapters (all others) — wrap an agent or handler with cache logic. Non-overridden attributes proxy through to the original object via __getattr__.

---

Framework Support Matrix

Adapter	Framework	Min Version	Extra	Interface
OpenAICacheAdapter	OpenAI SDK	openai >= 1.0	pip install 'chengeta-ai[openai]'	client.chat.completions.create wrapper
AnthropicCacheAdapter	Anthropic SDK	anthropic >= 0.25	pip install 'chengeta-ai[anthropic]'	client.messages.create wrapper
GoogleADKCacheAdapter	Google ADK	google-adk >= 0.1	pip install 'chengeta-ai[google-adk]'	Agent wrapper
OpenAIAgentsCacheAdapter	OpenAI Agents SDK	openai-agents	pip install openai-agents	Runner wrapper
LlamaIndexLLMCacheAdapter	LlamaIndex	llama-index-core >= 0.10	pip install 'chengeta-ai[llamaindex]'	LLM drop-in
LlamaIndexQueryCacheAdapter	LlamaIndex	llama-index-core >= 0.10	pip install 'chengeta-ai[llamaindex]'	QueryEngine wrapper
ClaudeAgentCacheAdapter	Claude Agent SDK	claude-code-sdk	pip install claude-code-sdk	Async generator wrapper
LangChainCacheAdapter	LangChain	langchain-core >= 0.2	pip install 'chengeta-ai[langchain]'	BaseCache
LangGraphCacheAdapter	LangGraph	langgraph >= 0.1	pip install 'chengeta-ai[langgraph]'	BaseCheckpointSaver
AutoGenCacheAdapter	AutoGen	pyautogen >= 0.2 or autogen-agentchat >= 0.4	pip install 'chengeta-ai[autogen]'	Agent wrapper
CrewAICacheAdapter	CrewAI	crewai >= 0.28	pip install 'chengeta-ai[crewai]'	Crew wrapper
AgnoCacheAdapter	Agno	agno >= 0.1	pip install 'chengeta-ai[agno]'	Agent wrapper
A2ACacheAdapter	A2A	—	pip install chengeta-ai	Handler wrapper / decorator

---

Quick Start

All adapters follow the same three-step pattern:

from chengeta_ai import CacheManager, InMemoryBackend, CacheKeyBuilder

# 1. Create a CacheManager
manager = CacheManager(
    backend=InMemoryBackend(),
    key_builder=CacheKeyBuilder(namespace="myapp"),
)

# 2. Import the adapter
from chengeta_ai.adapters.langchain_adapter import LangChainCacheAdapter

# 3. Plug it in
adapter = LangChainCacheAdapter(manager)

---

Choosing the Right Adapter

If you use...	Use this adapter
OpenAI SDK directly	OpenAICacheAdapter
Anthropic SDK directly	AnthropicCacheAdapter
Google ADK agents	GoogleADKCacheAdapter
OpenAI Agents SDK	OpenAIAgentsCacheAdapter
LlamaIndex LLMs	LlamaIndexLLMCacheAdapter
LlamaIndex QueryEngine (RAG)	LlamaIndexQueryCacheAdapter
Claude Agent SDK	ClaudeAgentCacheAdapter
LangChain chat models	LangChainCacheAdapter
LangGraph state graphs	LangGraphCacheAdapter
AutoGen agents	AutoGenCacheAdapter
CrewAI crews	CrewAICacheAdapter
Agno agents	AgnoCacheAdapter
Custom A2A messaging	A2ACacheAdapter
Custom LLM functions	Middleware

---

Next Steps

• Google ADK — Cache Google ADK agent runs
• OpenAI Agents SDK — Cache OpenAI Agents SDK runner results
• LlamaIndex — Cache LLM + QueryEngine results
• Claude Agent SDK — Cache Claude agent query() streams
• LangChain — Global LLM cache
• LangGraph — Checkpoint persistence
• AutoGen — Cached agent replies
• CrewAI — Cached crew kickoff
• Agno — Cached agent runs
• A2A — Cached inter-agent messaging