Integration

Langchain

knowmind-langchain ist ein PyPI-Paket mit BaseTool-Implementierungen für Agents und einem BaseRetriever für RAG-Chains.

Installation

bash

pip install knowmind-langchain

Setup

bash

export KNOWMIND_TOKEN="kmt_…"
# optional, Default ist https://knowmind.de
export KNOWMIND_API_URL="https://knowmind.de"

Agent mit knowmind-Tools

python

from langchain_anthropic import ChatAnthropic
from langgraph.prebuilt import create_react_agent
from knowmind_langchain import KnowmindClient, knowmind_tools

client = KnowmindClient()
agent = create_react_agent(
    model=ChatAnthropic(model="claude-haiku-4-5-20251001"),
    tools=knowmind_tools(client),
)
result = agent.invoke({
    "messages": [("user", "Was wissen wir über Projekt Helios?")],
})
print(result["messages"][-1].content)

RAG mit RetrievalQA

python

from langchain.chains import RetrievalQA
from langchain_anthropic import ChatAnthropic
from knowmind_langchain import KnowmindClient, KnowmindRetriever

client = KnowmindClient()
retriever = KnowmindRetriever(client=client, k=8, hops=2)

qa = RetrievalQA.from_chain_type(
    llm=ChatAnthropic(model="claude-haiku-4-5-20251001"),
    retriever=retriever,
)
qa.invoke({"query": "Wer ist Ansprechpartner für den Maschinenbau-Kunden?"})

Tool-Übersicht

Klasse	MCP-Tool	Scope
`KnowmindRecallTool`	`knowmind_recall`	read
`KnowmindStoreTool`	`knowmind_store_memory`	write
`KnowmindLinkTool`	`knowmind_link`	write
`KnowmindStatsTool`	`knowmind_stats`	read
`KnowmindRetriever`	`knowmind_recall` (BaseRetriever)	read

Async

Alle Tools und der Retriever implementieren _arun / _aget_relevant_documents — nutzbar in LangGraph, LangChain-Server, FastAPI-Endpoints.

Quelle & Issues

Source unter Programmieren/knowmind-langchain. Apache-2.0. Fragen oder Bugs: info@schuebeler-consulting.de.