Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.memwal.ai/llms.txt

Use this file to discover all available pages before exploring further.

with_memwal_langchain and with_memwal_openai wrap an existing LLM client with automatic memory management. Before each call relevant memories are recalled and injected; after each call the user message is analyzed for new facts (fire-and-forget). Both integrations import their dependency lazily — install only what you use: 
pip install memwal[langchain]

​
LangChain

import os
from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage
from memwal import with_memwal_langchain

llm = ChatOpenAI(model="gpt-4o")
smart_llm = with_memwal_langchain(
    llm,
    key=os.environ["MEMWAL_PRIVATE_KEY"],
    account_id=os.environ["MEMWAL_ACCOUNT_ID"],
    env="prod",
    namespace="chatbot-prod",
    max_memories=5,
    min_relevance=0.3,
)

response = await smart_llm.ainvoke([HumanMessage("What are my food allergies?")])
Patches both _agenerate (async) and _generate (sync) on the model instance.

​
OpenAI SDK

Works with both openai.OpenAI (sync) and openai.AsyncOpenAI (async) — the wrapper detects which and patches chat.completions.create accordingly. 
import os
from openai import AsyncOpenAI
from memwal import with_memwal_openai

client = AsyncOpenAI()
smart_client = with_memwal_openai(
    client,
    key=os.environ["MEMWAL_PRIVATE_KEY"],
    account_id=os.environ["MEMWAL_ACCOUNT_ID"],
    env="prod",
)

response = await smart_client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "What are my food allergies?"}],
)
The JS-style alias withMemWal is exported as a shortcut for with_memwal_langchain.

​
What It Does

Before generation:
  • Reads the last user message
  • Runs recall() against MemWal
  • Filters by min_relevance (default 0.3)
  • Injects matching memories as a system message before the last user message
After generation:
  • If auto_save (default True), runs analyze() on the user message fire-and-forget
  • Extracted facts are stored asynchronously

​
Options

Both wrappers accept the same keyword arguments:
OptionDefaultDescription
server_urlhttp://localhost:8000Explicit relayer URL (wins over env)
env—Relayer preset: prod / dev / staging / local
namespace"default"Memory namespace
max_memories5Max memories injected per request
auto_saveTrueAuto-save new facts from the conversation
min_relevance0.3Minimum similarity (0–1) to include a memory
debugFalseVerbose logging via the memwal logger

​
When To Use Direct SDK Calls Instead

Use direct MemWal methods when you need precise control over when memory is stored, which text is analyzed, or how recall results are filtered and displayed.