Deployment & Capstone Project

1import os
2
3# Local: .env file
4# Production: Platform secrets
5OPENAI_KEY = os.getenv("OPENAI_API_KEY")
6ENVIRONMENT = os.getenv("ENVIRONMENT", "development")
7
8if ENVIRONMENT == "production":
9    # Production settings
10    DEFAULT_MODEL = "gpt-4o-mini"  # Cost-effective
11    MAX_TOKENS = 500
12    RATE_LIMIT = 10  # requests per minute
13else:
14    # Development settings
15    DEFAULT_MODEL = "gpt-4o"
16    MAX_TOKENS = 2000
17    RATE_LIMIT = 100

1import logging
2import streamlit as st
3
4logging.basicConfig(level=logging.INFO)
5logger = logging.getLogger(__name__)
6
7def safe_ai_call(prompt):
8    try:
9        response = client.chat.completions.create(
10            model=DEFAULT_MODEL,
11            messages=[{"role": "user", "content": prompt}],
12            timeout=30
13        )
14        return response.choices[0].message.content
15    
16    except Exception as e:
17        logger.error(f"AI call failed: {e}")
18        st.error("🔧 Hệ thống đang gặp sự cố. Vui lòng thử lại sau.")
19        return None

1from datetime import datetime, timedelta
2
3class RateLimiter:
4    def __init__(self, max_requests=10, window_seconds=60):
5        self.max_requests = max_requests
6        self.window = timedelta(seconds=window_seconds)
7        self.requests = []
8    
9    def allow(self):
10        now = datetime.now()
11        self.requests = [r for r in self.requests if now - r < self.window]
12        
13        if len(self.requests) >= self.max_requests:
14            return False
15        
16        self.requests.append(now)
17        return True
18
19limiter = RateLimiter(max_requests=10, window_seconds=60)
20
21if not limiter.allow():
22    st.warning("⏳ Bạn đã gửi quá nhiều yêu cầu. Vui lòng đợi 1 phút.")

1# capstone/app.py
2import streamlit as st
3import pandas as pd
4from openai import OpenAI
5
6st.set_page_config(page_title="AI Data Analyst", page_icon="📊", layout="wide")
7st.title("📊 AI Data Analyst Assistant")
8
9client = OpenAI(api_key=st.secrets["OPENAI_API_KEY"])
10
11# Upload data
12uploaded = st.file_uploader("Upload CSV", type=["csv"])
13
14if uploaded:
15    df = pd.read_csv(uploaded)
16    
17    # Show preview
18    st.subheader("📋 Data Preview")
19    st.dataframe(df.head())
20    
21    col1, col2, col3 = st.columns(3)
22    col1.metric("Rows", len(df))
23    col2.metric("Columns", len(df.columns))
24    col3.metric("Missing", df.isnull().sum().sum())
25    
26    # Natural language query
27    question = st.chat_input("Hỏi về data...")
28    
29    if question:
30        # Convert data info to context
31        data_info = f"""
32        Columns: {list(df.columns)}
33        Dtypes: {df.dtypes.to_dict()}
34        Shape: {df.shape}
35        Sample: {df.head(3).to_dict()}
36        Stats: {df.describe().to_dict()}
37        """
38        
39        with st.chat_message("assistant"):
40            response = client.chat.completions.create(
41                model="gpt-4o",
42                messages=[
43                    {"role": "system", "content": f"""
44                        Bạn là Data Analyst expert.
45                        Dataset info: {data_info}
46                        Trả lời bằng tiếng Việt.
47                        Khi cần chart, output Python code (pandas + plotly).
48                    """},
49                    {"role": "user", "content": question}
50                ],
51                stream=True
52            )
53            st.write_stream(response)