LangChain Integration

The FraiseQL LangChain integration exposes your GraphQL API as LangChain tools and retrievers, enabling agents to query your database as part of reasoning chains.

Installation

pip install "fraiseql[langchain]"

FraiseQLToolkit

FraiseQLToolkit introspects a FraiseQL server and generates a BaseTool for every query and mutation. Pass the tools to any LangChain agent.

import asyncio
from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_anthropic import ChatAnthropic
from fraiseql.integrations.langchain import FraiseQLToolkit

async def main() -> None:
    # Introspect schema and generate tools
    toolkit = await FraiseQLToolkit.from_url(
        "http://localhost:8080/graphql",
        auth_token="Bearer eyJ..."
    )
    tools = toolkit.get_tools(
        exclude=["adminUsers"]   # hide admin operations from the agent
    )

    llm = ChatAnthropic(model="claude-opus-4-6")
    agent = create_tool_calling_agent(llm, tools, prompt=...)
    executor = AgentExecutor(agent=agent, tools=tools)

    result = await executor.ainvoke({
        "input": "Show me all orders placed today"
    })
    print(result["output"])

asyncio.run(main())

Each tool’s description is taken from the GraphQL field description in your schema — write clear docstrings in your Python @fraiseql.query definitions and the agent will know how to use each tool.

FraiseQLRetriever

FraiseQLRetriever wraps a single query as a LangChain retriever, returning results as Document objects. Use it with RAG chains or as a memory source.

from fraiseql.client import FraiseQLClient
from fraiseql.integrations.langchain import FraiseQLRetriever

client = FraiseQLClient("http://localhost:8080/graphql")

retriever = FraiseQLRetriever(
    client=client,
    query="""
        query SearchProducts($search: String) {
            products(search: $search) { id name description price }
        }
    """,
    text_key="description",          # field used as Document.page_content
    metadata_keys=["id", "name", "price"],
)

docs = await retriever.ainvoke("wireless headphones under $100")
for doc in docs:
    print(doc.page_content)
    print(doc.metadata)