LangChain Integration

LangChain + Awareness Memory

Add persistent, cross-session memory to your LangChain agents and RAG pipelines. Every decision, code change, and conversation is stored and recalled automatically across sessions.

Installation

pip install awareness-memory-cloud[langchain]

This installs the core SDK plus the LangChain adapter.

Quick Start

from memory_cloud import MemoryCloudClient
from memory_cloud.integrations.langchain import MemoryCloudLangChain
import openai

# Local mode (no API key needed)
client = MemoryCloudClient(mode="local")
mc = MemoryCloudLangChain(client=client)  # memory_id auto-managed

# Cloud mode (team collaboration, semantic search, multi-device sync)
client = MemoryCloudClient(base_url="https://awareness.market/api/v1", api_key="YOUR_API_KEY")
mc = MemoryCloudLangChain(client=client, memory_id="memory_123")

# Option 1: Wrap the LLM client for auto-recall + auto-capture
mc.wrap_llm(openai.OpenAI())

# Option 2: Use as a LangChain Retriever
retriever = mc.as_retriever()
docs = retriever._get_relevant_documents("What did we decide yesterday?")

Integration Patterns

Pattern 1: Interceptor (Zero-Code Memory)

Wrap your LLM client and all calls automatically get memory context injected before the call and stored after:

mc.wrap_llm(openai.OpenAI())

# Every LLM call now has cross-session memory context
# No changes to your existing LangChain chains needed

Pattern 2: LangChain Retriever

Use Awareness as a retriever in your RAG chains:

from langchain.chains import RetrievalQA
from langchain_openai import ChatOpenAI

retriever = mc.as_retriever()
qa_chain = RetrievalQA.from_chain_type(
    llm=ChatOpenAI(),
    retriever=retriever,
)

result = qa_chain.invoke("What's the current auth implementation?")

Pattern 3: Tool Registration

Register memory tools for your LangChain agents:

tools = mc.build_tools()

# Or get raw tool functions for manual registration
tool_functions = mc.get_tool_functions()

Use Cases

  • RAG chains with persistent memory — retriever pulls relevant past decisions and context into your chain
  • Agent tool registration — give LangChain agents the ability to read and write memory
  • Cross-session knowledge — your agent remembers what happened in previous sessions
  • Team memory — multiple agents share context via user_id and agent_role

Multi-User / Multi-Role

# Developer agent
mc_dev = MemoryCloudLangChain(
    client=client, memory_id="memory_123",
    default_metadata={"agent_role": "developer", "user_id": "alice"}
)

# Reviewer agent
mc_reviewer = MemoryCloudLangChain(
    client=client, memory_id="memory_123",
    default_metadata={"agent_role": "reviewer", "user_id": "bob"}
)

When querying with a user_id, results include both that user's data AND public (user_id=NULL) data.

Query Rewrite

All framework adapters support query rewrite via MemoryCloudBaseAdapter:

  • "rule" (default): Context-aware query + keyword extraction. Zero extra latency.
  • "llm": LLM-generated optimal queries. Best for ambiguous inputs.
  • "none": Raw user message (legacy).

Example

See examples/e2e_langchain_cloud.py in the SDK repository for a complete end-to-end example.

Next Steps

AI FrameworksCrewAI Integration