Databricks Apps

# 앱 Service Principal 인증 (기본)
from databricks.sdk import WorkspaceClient

# 앱 내부에서는 자동으로 앱의 Service Principal로 인증됩니다
w = WorkspaceClient()
tables = w.tables.list(catalog_name="main", schema_name="default")

# 사용자 대리 인증
from databricks.sdk import WorkspaceClient

def get_user_client(request_headers):
    """요청 헤더에서 사용자 토큰을 추출하여 클라이언트 생성"""
    w = WorkspaceClient(
        host=os.environ["DATABRICKS_HOST"],
        token=request_headers.get("X-Forwarded-Access-Token")
    )
    return w

my-app/
| 파일 | 설명 |
|------|------|
| `app.py` | 메인 앱 코드 |
| `app.yaml` | Databricks Apps 설정 |
| `requirements.txt` | 추가 Python 패키지 |
| `static/logo.png` | 정적 파일 (선택) |

# 1. 앱 생성 (최초 1회)
databricks apps create my-dashboard \
  --description "매출 분석 대시보드"

# 2. 앱 배포 (코드 업로드 + 시작)
databricks apps deploy my-dashboard \
  --source-code-path ./my-app

# 3. 앱 상태 확인
databricks apps get my-dashboard

# 4. 앱 로그 확인
databricks apps get-logs my-dashboard

# databricks.yml
bundle:
  name: my-dashboard

resources:
  apps:
    my-dashboard:
      name: "my-dashboard"
      description: "매출 분석 대시보드"
      source_code_path: ./src
      config:
        command:
          - "streamlit"
          - "run"
          - "app.py"
          - "--server.port"
          - "8000"
        resources:
          - name: "warehouse"
            sql_warehouse:
              id: "${var.warehouse_id}"
              permission: "CAN_USE"

import streamlit as st
from databricks import sql
import os

# SQL Warehouse 연결 (리소스 바인딩 활용)
connection = sql.connect(
    server_hostname=os.environ["DATABRICKS_HOST"],
    http_path=os.environ["DATABRICKS_SQL_WAREHOUSE_HTTP_PATH"],
    credentials_provider=lambda: {
        "Authorization": f"Bearer {os.environ['DATABRICKS_TOKEN']}"
    }
)

cursor = connection.cursor()
cursor.execute("SELECT * FROM main.sales.daily_revenue LIMIT 100")
df = cursor.fetchall_arrow().to_pandas()

st.dataframe(df)

from databricks.sdk import WorkspaceClient

w = WorkspaceClient()

# Unity Catalog 테이블 목록 조회
for table in w.tables.list(catalog_name="main", schema_name="sales"):
    print(f"테이블: {table.name}")

# app.py — Streamlit LLM 챗봇
import streamlit as st
from databricks.sdk import WorkspaceClient

st.title("🤖 AI 어시스턴트")

w = WorkspaceClient()

# 채팅 히스토리 관리
if "messages" not in st.session_state:
    st.session_state.messages = []

# 이전 메시지 표시
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.markdown(message["content"])

# 사용자 입력
if prompt := st.chat_input("질문을 입력하세요"):
    st.session_state.messages.append({"role": "user", "content": prompt})

    with st.chat_message("user"):
        st.markdown(prompt)

    # Model Serving Endpoint 호출
    response = w.serving_endpoints.query(
        name="llm-endpoint",
        messages=[
            {"role": "system", "content": "당신은 데이터 분석 전문가입니다."},
            *st.session_state.messages
        ]
    )

    assistant_message = response.choices[0].message.content

    with st.chat_message("assistant"):
        st.markdown(assistant_message)

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

command:
  - "streamlit"
  - "run"
  - "app.py"
  - "--server.port"
  - "8000"

resources:
  - name: "llm-endpoint"
    serving_endpoint:
      name: "llm-endpoint"
      permission: "CAN_QUERY"

mkdir sales-dashboard && cd sales-dashboard

# app.py
import streamlit as st
from databricks import sql
import os
import pandas as pd

st.set_page_config(page_title="매출 대시보드", layout="wide")
st.title("📊 일별 매출 대시보드")

# SQL Warehouse 연결
@st.cache_resource
def get_connection():
    return sql.connect(
        server_hostname=os.environ["DATABRICKS_HOST"],
        http_path=os.environ["DATABRICKS_SQL_WAREHOUSE_HTTP_PATH"],
        credentials_provider=lambda: {
            "Authorization": f"Bearer {os.environ['DATABRICKS_TOKEN']}"
        }
    )

conn = get_connection()

# 데이터 조회
@st.cache_data(ttl=300)  # 5분 캐시
def load_data():
    cursor = conn.cursor()
    cursor.execute("""
        SELECT sale_date, total_revenue, total_orders, unique_customers
        FROM main.ecommerce.gold_daily_kpi
        ORDER BY sale_date DESC
        LIMIT 90
    """)
    return cursor.fetchall_arrow().to_pandas()

df = load_data()

# KPI 카드
col1, col2, col3 = st.columns(3)
col1.metric("총 매출", f"₩{df['total_revenue'].sum():,.0f}")
col2.metric("총 주문", f"{df['total_orders'].sum():,}건")
col3.metric("고유 고객", f"{df['unique_customers'].sum():,}명")

# 차트
st.line_chart(df.set_index("sale_date")["total_revenue"])
st.dataframe(df, use_container_width=True)

command:
  - "streamlit"
  - "run"
  - "app.py"
  - "--server.port"
  - "8000"

resources:
  - name: "sql-warehouse"
    sql_warehouse:
      id: "${DATABRICKS_WAREHOUSE_ID}"
      permission: "CAN_USE"

streamlit
databricks-sql-connector
pandas
pyarrow

# 앱 생성 및 배포
databricks apps create sales-dashboard --description "일별 매출 대시보드"
databricks apps deploy sales-dashboard --source-code-path .

# 배포 상태 확인
databricks apps get sales-dashboard