The MentisDB Agent Memory Cookbook

5.2 Python: LangChain + MentisDB

Use MentisDB beside LangChain as durable long-term memory. Keep short conversational buffer memory in LangChain; store durable decisions, facts, and lessons in MentisDB.

from pymentisdb import MentisDbClient

memory = MentisDbClient("http://127.0.0.1:9472", chain_key="my-agent")

memory.append(
    agent_id="langchain-agent",
    thought_type="Decision",
    content="Use the billing API v2 endpoint for invoice export.",
    concepts=["billing", "invoice-export"],
    tags=["langchain", "production"],
)

hits = memory.ranked_search("which billing endpoint exports invoices?", limit=5)

Retrieve from MentisDB before calling the model, then include only the top relevant memories in the prompt.