LiteLLMとLangGraphで構築したAIエージェントをFastAPI・Streamlit・Dockerでデプロイしてみる

Python 3.13
litellm 1.83.14
langgraph 1.1.10
langchain 1.0.0
langchain-litellm 0.6.4
fastapi
uvicorn
streamlit

"""FastAPI で LangGraph エージェントを REST API として公開する。

エンドポイント:
- POST /chat: 1ターンの会話
- thread_id でセッションを継続
"""

from fastapi import FastAPI
from langchain.agents import create_agent
from langchain_core.tools import tool
from langchain_litellm import ChatLiteLLM
from langgraph.checkpoint.memory import InMemorySaver
from pydantic import BaseModel

@tool
def get_weather(city: str) -> str:
    """都市の天気を返します。"""
    return f"{city}: 晴れ、24°C"

llm = ChatLiteLLM(model="openai/gpt-5-mini")
checkpointer = InMemorySaver()
agent = create_agent(model=llm, tools=[get_weather], checkpointer=checkpointer)

app = FastAPI(title="LiteLLM × LangGraph Agent")

class ChatRequest(BaseModel):
    thread_id: str
    message: str

class ChatResponse(BaseModel):
    answer: str

@app.post("/chat", response_model=ChatResponse)
async def chat(req: ChatRequest) -> ChatResponse:
    config = {"configurable": {"thread_id": req.thread_id}}
    result = await agent.ainvoke(
        {"messages": [{"role": "user", "content": req.message}]}, config=config
    )
    return ChatResponse(answer=str(result["messages"][-1].content))

@app.get("/health")
def health() -> dict:
    return {"status": "ok"}

# 起動: uvicorn fastapi_server:app --host 0.0.0.0 --port 8000

$ uvicorn fastapi_server:app --reload
INFO:     Started server process [...]
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8000

# 別ターミナルで
$ curl -s http://localhost:8000/health
{"status":"ok"}

$ curl -s -X POST http://localhost:8000/chat \
       -H "Content-Type: application/json" \
       -d '{"thread_id":"demo-1","message":"東京の天気を教えて"}'
{"answer":"東京の天気は晴れ、気温は24°Cです。何か他に知りたいことはありますか？"}

"""Streamlit で LangGraph エージェントのチャットUIを提供する。

起動: streamlit run streamlit_app.py
"""

import streamlit as st
from langchain.agents import create_agent
from langchain_core.tools import tool
from langchain_litellm import ChatLiteLLM
from langgraph.checkpoint.memory import InMemorySaver

@tool
def get_weather(city: str) -> str:
    """都市の天気を返します。"""
    return f"{city}: 晴れ、24°C"

@st.cache_resource
def get_agent():
    llm = ChatLiteLLM(model="openai/gpt-5-mini")
    checkpointer = InMemorySaver()
    return create_agent(model=llm, tools=[get_weather], checkpointer=checkpointer)

agent = get_agent()
st.title("LiteLLM × LangGraph Chat")

if "thread_id" not in st.session_state:
    import uuid

    st.session_state.thread_id = str(uuid.uuid4())

if "messages" not in st.session_state:
    st.session_state.messages = []

# 既存メッセージを表示
for msg in st.session_state.messages:
    with st.chat_message(msg["role"]):
        st.markdown(msg["content"])

# 入力
if user_input := st.chat_input("メッセージを入力..."):
    st.session_state.messages.append({"role": "user", "content": user_input})
    with st.chat_message("user"):
        st.markdown(user_input)

    config = {"configurable": {"thread_id": st.session_state.thread_id}}
    with st.chat_message("assistant"):
        placeholder = st.empty()
        full_response = ""

        # ストリーミングで応答
        for event in agent.stream(
            {"messages": [{"role": "user", "content": user_input}]},
            config=config,
            stream_mode="updates",
        ):
            for _node_name, node_state in event.items():
                if "messages" in node_state:
                    last = node_state["messages"][-1]
                    if hasattr(last, "content") and isinstance(last.content, str):
                        full_response = last.content
                        placeholder.markdown(full_response)

    st.session_state.messages.append({"role": "assistant", "content": full_response})

$ streamlit run streamlit_app.py

  You can now view your Streamlit app in your browser.
  Local URL: http://localhost:8501
  Network URL: http://...:8501

FROM python:3.13-slim

WORKDIR /app

COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY . .

EXPOSE 8000

CMD ["uvicorn", "fastapi_server:app", "--host", "0.0.0.0", "--port", "8000"]

$ docker build -t litellm-agent .
$ docker run -p 8000:8000 \
       -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
       -e OPENAI_API_KEY=$OPENAI_API_KEY \
       litellm-agent