langgraph-goodmem

Give LangGraph agents searchable, persistent memory with GoodMem. GoodMem handles document storage, chunking, embedding, search, and optional reranking. Use it from a graph node or give an agent a search tool with access to the spaces you choose.

This package shares its tools, retriever, and ingestion functions with langchain-goodmem, so fixes reach both integrations.

Install

pip install langgraph-goodmem

Requires Python 3.10+. Configure an existing GoodMem server and space:

export GOODMEM_BASE_URL="https://your-goodmem-server.example.com"
export GOODMEM_API_KEY="your-api-key"
export GOODMEM_SPACE_ID="your-space-uuid"

Search from a graph

This complete example searches your space without an LLM. The retrieval node adds Document objects, including source metadata, to the graph's state.

import os
from typing import TypedDict

from langchain_core.documents import Document
from langchain_core.runnables import RunnableConfig
from langgraph.graph import END, START, StateGraph
from langgraph_goodmem import GoodMemRetriever

class State(TypedDict):
    question: str
    documents: list[Document]

retriever = GoodMemRetriever(space_ids=[os.environ["GOODMEM_SPACE_ID"]], k=5)

def search(state: State, config: RunnableConfig):
    return {"documents": retriever.invoke(state["question"], config=config)}

builder = StateGraph(State)
builder.add_node("search", search)
builder.add_edge(START, "search")
builder.add_edge("search", END)
graph = builder.compile()
result = graph.invoke({"question": "What is the refund policy?", "documents": []})
for document in result["documents"]:
    print(document.metadata["source"], document.page_content, sep="\n")

Use an agent

The agent example gives create_agent a scoped search tool, also usable in ToolNode. Give searches distinct names and descriptions; your code controls their spaces, filters, and rerankers.

Install langgraph-goodmem[agents] plus your chosen model provider's LangChain package, configure its credentials, and set GOODMEM_CHAT_MODEL=provider:model.

Add documents

import os

from goodmem import Goodmem
from langchain_core.documents import Document
from langgraph_goodmem import add_documents

with Goodmem(base_url=os.environ["GOODMEM_BASE_URL"],
             api_key=os.environ["GOODMEM_API_KEY"]) as client:
    memory_ids = add_documents(client, os.environ["GOODMEM_SPACE_ID"], [
        Document(page_content="Refunds are available within 30 days.",
                 metadata={"source": "https://example.com/refunds"})
    ])

Ingestion waits for the memories it created. Empty searches return immediately. GoodMemIngestionError.created_memory_ids identifies accepted writes if indexing fails, so you can check them with wait_for_memory instead of uploading again.

More options

• Set filter="CAST(val('$.department') AS TEXT) = 'support'" on the retriever for metadata filtering.
• Set reranker_id and optionally fetch_k to rerank without an LLM.
• Use retriever.invoke, ainvoke, batch, or abatch for Document results. Async calls currently run the shared synchronous SDK in a thread executor.
• Pass client=Goodmem(...) to share a connection. For local self-signed TLS, use Goodmem(..., verify=False); keep verification enabled in production.
• Administrative tools remain available for trusted workflows. GoodMemRetrieveMemories returns raw SDK events, including failure statuses. When the server reports a problem, GoodMemRetriever still returns the Documents it received, each with goodmem_partial and goodmem_statuses metadata; with no Documents it returns an empty list and emits a warning. It does not raise.
• Tools refuse any ID that is not a UUID before sending a request, so a model-supplied ID cannot reach a different endpoint.

This package provides retrieval and ingestion; it does not implement LangGraph's BaseStore or a checkpointer for graph execution state. See the 0.2 migration notes for API changes.