progress-memory-langchain

LangChain integration for the Progress Memory SDK.

Provides two integration tiers:

ProgressMemoryTools ⭐ recommended for production — exposes remember, recall, and forget as LLM-callable StructuredTool objects. The LLM decides when to invoke each one — token-efficient and semantically accurate.

ProgressMemoryLangChainMemory (quick start) — automatically recalls memories before every LLM call and saves each exchange after.

Quick start — Tools (production)

from progress_memory import ProgressMemoryClient, ProgressMemoryConfig
from progress_memory_langchain import ProgressMemoryTools

config = ProgressMemoryConfig.from_env()
client = ProgressMemoryClient(config)

tools_provider = ProgressMemoryTools(client=client, user_id="user-123", agent_id="bot")
tools = tools_provider.create_tools()  # [remember, recall, forget]

Quick start — Automatic memory (demos)

from progress_memory_langchain import ProgressMemoryLangChainMemory

memory = ProgressMemoryLangChainMemory(
    client=client, user_id="user-123", agent_id="bot", auto_remember=True
)
chain_with_memory = memory.wrap(chain)
response = chain_with_memory.invoke({"input": "What are my preferences?"})