Capstone & Industry Sim

🛠 Capstone Project: SmartReview API (5–25 min)

💻 Full Capstone Code (25–45 min)

># smart_review/api.py
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel, Field
from typing import List
import pickle, re

app = FastAPI(title="SmartReview API", version="1.0")
CONFIDENCE_THRESHOLD = 0.70          # below = uncertain

class ReviewIn(BaseModel):
    text: str = Field(..., min_length=3, max_length=1000)

class ReviewOut(BaseModel):
    sentiment:  str    # positive | negative | uncertain
    confidence: float
    word_count: int

def analyse(text: str, vectorizer, model) -> ReviewOut:
    vec   = vectorizer.transform([text])
    pred  = model.predict(vec)[0]
    proba = model.predict_proba(vec)[0].max()
    if proba < CONFIDENCE_THRESHOLD:
        label = "uncertain"
    else:
        label = "positive" if pred == 1 else "negative"
    return ReviewOut(sentiment=label, confidence=round(float(proba), 3),
                      word_count=len(text.split()))

@app.post("/predict", response_model=ReviewOut)
def predict(req: ReviewIn):
    return analyse(req.text, vectorizer, model)

@app.post("/batch", response_model=List[ReviewOut])
def batch_predict(reviews: List[ReviewIn]):
    if len(reviews) > 50:
        raise HTTPException(400, "Max 50 reviews per batch")
    return [analyse(r.text, vectorizer, model) for r in reviews]

># tests/test_api.py
from fastapi.testclient import TestClient
from smart_review.api import app

client = TestClient(app)

def test_positive_review():
    r = client.post("/predict", json={"text": "Excellent! Highly recommend this product."})
    assert r.status_code == 200
    assert r.json()["sentiment"] in ["positive", "uncertain"]

def test_negative_review():
    r = client.post("/predict", json={"text": "Terrible, waste of money, very bad."})
    assert r.status_code == 200

def test_text_too_short():
    r = client.post("/predict", json={"text": "ok"})
    assert r.status_code == 422           # Pydantic min_length=3

def test_batch_limit():
    big_batch = [{"text": "great product"}] * 51
    r = client.post("/batch", json=big_batch)
    assert r.status_code == 400

def test_confidence_field():
    r = client.post("/predict", json={"text": "Amazing quality, super fast delivery!"})
    data = r.json()
    assert 0.0 <= data["confidence"] <= 1.0

def test_word_count():
    text = "This is a five word review"
    r = client.post("/predict", json={"text": text})
    assert r.json()["word_count"] == 6

🚀 Git-Flow — Ship It Like a Pro (25–45 min)

># Industry standard: never commit directly to main

git checkout -b feature/confidence-threshold

# ... write code and tests ...

git add .
git commit -m "feat: add confidence threshold to sentiment API"

git push origin feature/confidence-threshold

# GitHub: Create Pull Request
# PR title: "feat: add confidence threshold"
# Description: What changed, why, how to test
# Request review from a classmate

# After approval: merge to main
git checkout main
git pull
git branch -d feature/confidence-threshold