OllamaをStreamlitと連携させる方法｜PythonでローカルLLMの社内専用チャットUIを自作する手順

この記事のポイント

・OllamaのPOST /api/chat とrequestsでStreamlitから直接接続できる
・st.write_streamでストリーミング表示、st.session_stateで会話履歴を保持する
・GET /api/tags でモデル一覧を取得し、サイドバーで動的に切り替える
・本番公開はsystemdサービス化＋NginxのWebSocketプロキシ設定で安定運用する

# Pythonバージョン確認（3.10以上が必要）
$ python3 --version
Python 3.11.9

# プロジェクトディレクトリを作成
$ sudo mkdir -p /opt/streamlit-llm
$ sudo chown ubuntu:ubuntu /opt/streamlit-llm
$ cd /opt/streamlit-llm

# 仮想環境を作成して有効化
$ python3 -m venv venv
$ source venv/bin/activate

# pip install でパッケージを追加する
(venv) $ pip install streamlit requests
...
Successfully installed streamlit-1.36.0 requests-2.32.3

# Streamlitバージョン確認
(venv) $ streamlit --version
Streamlit, version 1.36.0

# Ollamaの疎通確認（モデル一覧を取得）
$ curl -s http://localhost:11434/api/tags | python3 -m json.tool | head -12
{
    "models": [
        {
            "name": "llama3.3:70b-instruct-q4_0",
            "modified_at": "2026-07-20T11:22:33Z",
            "size": 42523684864
        }
    ]
}

# /opt/streamlit-llm/app.py（最小構成）
import streamlit as st
import requests
import json

OLLAMA_URL = "http://localhost:11434"
MODEL = "llama3.3:70b-instruct-q4_0"

st.title("社内AIアシスタント")

def call_ollama(messages):
    resp = requests.post(
        f"{OLLAMA_URL}/api/chat",
        json={"model": MODEL, "messages": messages},
        timeout=120,
    )
    resp.raise_for_status()
    return resp.json()["message"]["content"]

prompt = st.text_area("質問を入力してください", height=100)
if st.button("送信"):
    if prompt.strip():
        with st.spinner("応答中..."):
            answer = call_ollama([{"role": "user", "content": prompt}])
        st.write(answer)

# Streamlitアプリを起動（ポート8501）
(venv) $ streamlit run app.py --server.port 8501

  You can now view your Streamlit app in your browser.

  Local URL: http://localhost:8501
  Network URL: http://192.168.1.10:8501

# /opt/streamlit-llm/app.py（ストリーミング表示版）
import streamlit as st
import requests
import json

OLLAMA_URL = "http://localhost:11434"
MODEL = "llama3.3:70b-instruct-q4_0"

st.title("社内AIアシスタント")

def stream_ollama(messages):
    with requests.post(
        f"{OLLAMA_URL}/api/chat",
        json={"model": MODEL, "messages": messages, "stream": True},
        stream=True,
        timeout=300,
    ) as resp:
        resp.raise_for_status()
        for line in resp.iter_lines():
            if line:
                data = json.loads(line)
                yield data["message"]["content"]
                if data.get("done"):
                    break

if prompt := st.chat_input("質問を入力してください"):
    st.chat_message("user").write(prompt)
    with st.chat_message("assistant"):
        st.write_stream(stream_ollama([{"role": "user", "content": prompt}]))

# /opt/streamlit-llm/app.py（会話履歴版）
import streamlit as st
import requests
import json

OLLAMA_URL = "http://localhost:11434"
MODEL = "llama3.3:70b-instruct-q4_0"

st.title("社内AIアシスタント")

# ①初回実行時だけmessagesを初期化
if "messages" not in st.session_state:
    st.session_state.messages = []

def stream_ollama(messages):
    with requests.post(
        f"{OLLAMA_URL}/api/chat",
        json={"model": MODEL, "messages": messages, "stream": True},
        stream=True,
        timeout=300,
    ) as resp:
        resp.raise_for_status()
        for line in resp.iter_lines():
            if line:
                data = json.loads(line)
                yield data["message"]["content"]
                if data.get("done"):
                    break

# ②過去の会話を表示
for msg in st.session_state.messages:
    st.chat_message(msg["role"]).write(msg["content"])

# ③新規入力の処理
if prompt := st.chat_input("質問を入力してください"):
    st.session_state.messages.append({"role": "user", "content": prompt})
    st.chat_message("user").write(prompt)
    with st.chat_message("assistant"):
        response = st.write_stream(
            stream_ollama(st.session_state.messages)
        )
    st.session_state.messages.append({"role": "assistant", "content": response})

# /opt/streamlit-llm/app.py（完成形：モデル切り替え対応）
import streamlit as st
import requests
import json

OLLAMA_URL = "http://localhost:11434"

st.title("社内AIアシスタント")

# モデル一覧取得（60秒キャッシュ）
@st.cache_data(ttl=60)
def get_models():
    try:
        resp = requests.get(f"{OLLAMA_URL}/api/tags", timeout=5)
        resp.raise_for_status()
        return [m["name"] for m in resp.json().get("models", [])]
    except Exception:
        return ["llama3.3:70b-instruct-q4_0"]

def stream_ollama(model, messages):
    with requests.post(
        f"{OLLAMA_URL}/api/chat",
        json={"model": model, "messages": messages, "stream": True},
        stream=True,
        timeout=300,
    ) as resp:
        resp.raise_for_status()
        for line in resp.iter_lines():
            if line:
                data = json.loads(line)
                yield data["message"]["content"]
                if data.get("done"):
                    break

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

with st.sidebar:
    st.header("設定")
    models = get_models()
    selected_model = st.selectbox("モデルを選択", models)
    if st.button("会話をリセット"):
        st.session_state.messages = []
        st.rerun()

for msg in st.session_state.messages:
    st.chat_message(msg["role"]).write(msg["content"])

if prompt := st.chat_input("質問を入力してください"):
    st.session_state.messages.append({"role": "user", "content": prompt})
    st.chat_message("user").write(prompt)
    with st.chat_message("assistant"):
        response = st.write_stream(
            stream_ollama(selected_model, st.session_state.messages)
        )
    st.session_state.messages.append({"role": "assistant", "content": response})

# /etc/systemd/system/streamlit-llm.service を作成
$ sudo tee /etc/systemd/system/streamlit-llm.service << 'EOF'
[Unit]
Description=Streamlit Local LLM Chat UI
After=network.target ollama.service

[Service]
Type=simple
User=ubuntu
WorkingDirectory=/opt/streamlit-llm
ExecStart=/opt/streamlit-llm/venv/bin/streamlit run app.py \
  --server.port 8501 \
  --server.address 127.0.0.1 \
  --server.headless true
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target
EOF

$ sudo systemctl daemon-reload
$ sudo systemctl enable streamlit-llm
Created symlink /etc/systemd/system/multi-user.target.wants/streamlit-llm.service
$ sudo systemctl start streamlit-llm
$ sudo systemctl status streamlit-llm
* streamlit-llm.service - Streamlit Local LLM Chat UI
     Loaded: loaded (/etc/systemd/system/streamlit-llm.service; enabled)
     Active: active (running) since Thu 2026-07-24 09:15:00 JST; 3s ago
   Main PID: 12345 (streamlit)

# /etc/nginx/sites-available/streamlit-llm を作成
$ sudo tee /etc/nginx/sites-available/streamlit-llm << 'EOF'
server {
    listen 80;
    server_name chat.example.internal;

    location / {
        proxy_pass http://127.0.0.1:8501;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";
        proxy_set_header Host $host;
        proxy_read_timeout 300;
        proxy_send_timeout 300;
    }
}
EOF

$ sudo ln -s /etc/nginx/sites-available/streamlit-llm /etc/nginx/sites-enabled/
$ sudo nginx -t
nginx: the configuration file /etc/nginx/nginx.conf syntax is ok
nginx: configuration file /etc/nginx/nginx.conf test is successful
$ sudo systemctl reload nginx

StreamlitとOllamaを実機で動かすハンズオンを2日間で体験する

ローカルLLMのチャットUIを自分の手で組み立て、社内サーバーに組み込むまでを体験したい方向けに、「ローカルAIマスターセミナー」を開催しています。
少人数（最大8名）ZOOMハンズオン形式で実施しています。