ledgermem-langgraph

LangGraph BaseStore backed by LedgerMem — give your agents persistent, searchable cross-thread memory in one line.

Install

pip install ledgermem-langgraph

Quickstart

from langgraph.graph import StateGraph, MessagesState
from ledgermem import LedgerMem
from langgraph_ledgermem import LedgerMemStore

store = LedgerMemStore(LedgerMem(api_key="lm_...", workspace_id="ws_..."))

# Write
store.put(("users", "u1"), "profile", {"name": "Ada", "tz": "PKT"})

# Read by exact key
item = store.get(("users", "u1"), "profile")
print(item.value)

# Semantic search across the namespace
hits = store.search(("users",), query="who lives in Pakistan?", limit=5)
for hit in hits:
    print(hit.score, hit.value)

Wire into a LangGraph workflow

graph = StateGraph(MessagesState)
# ... add nodes / edges ...
app = graph.compile(store=store)

The store is passed to every node via the standard config["configurable"]["store"] channel — your nodes can store.put(...) and store.search(...) to share long-term context across threads.

License

MIT — see LICENSE.