Chapter 35: FastAPI for ML Deployment
MODULE 6 | MLOps
Chapter 35: FastAPI for ML Deployment
Ab tak tumhara model Python file ya notebook ke andar rehta tha. Koi website, mobile app ya doosra program usay seedha nahi chala sakta. Is chapter me hum model ko ek API me lapetenge taake koi bhi use internet (ya apne computer) par bhej kar jawab le sake. Is kaam ke liye hum FastAPI use karenge.
35.1 API kya hoti hai?
JSON ki misaal
{"sepal_length": 5.1, "sepal_width": 3.5, "petal_length": 1.4, "petal_width": 0.2}
| HTTP method | Kaam | ML me misaal |
|---|---|---|
| GET | Maloomatlena | /health (server theek hai?), /symptoms (list lo) |
| POST | Databhej karkaam karwana | /predict (features bhejo, prediction lo) |
| Status code | Matlab | Kab aata hai |
|---|---|---|
| 200 | OK, sab theek | Request kamyab rahi |
| 400 | Bad Request | Request me galti (hamari apni check se) |
| 404 | Not Found | Aisa URL maujood hi nahi |
| 422 | Unprocessable Entity | Data ka format ya qeemat galat (Pydantic ki check) |
| 500 | Server Error | Server ke andar code ya model me masla |
35.2 FastAPI install aur pehla endpoint
FastAPI ek tez aur asaan Python framework hai jo API banane ke liye bana hai. Iske sath Uvicorn chalta hai, jo server hai (API ko chalane wala program). Chapter 34 me hum ne dono install kar liye the. Pehle sab se chhoti API likhte hain. Ye code hello.py me rakho:
hello.py
from fastapi import FastAPI
app = FastAPI()
@app.get("/")
def home():
return {"message": "Salam! FastAPI chal rahi hai"}
Ab terminal me server chalao (hello file ka naam hai, app us ke andar wala object):
TERMINAL
uvicorn hello:app --reload TERMINAL OUTPUT (folder ka naam aur process number tumhare yahan alag honge) INFO: Will watch for changes in these directories: ['/tmp/snip'] INFO: Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit) INFO: Started reloader process [377] using StatReload INFO: Started server process [380] INFO: Waiting for application startup. INFO: Application startup complete.
Ab browser me http://127.0.0.1:8000 kholo, ya doosre terminal me curl chalao:
TERMINAL
curl http://127.0.0.1:8000/ OUTPUT
{"message":"Salam! FastAPI chal rahi hai"}
| Code ka hissa | Matlab |
|---|---|
app = FastAPI() | Hamari API kaasli object; saare endpoints isi se jurte hain |
@app.get("/") | Decorator: batata hai ke neeche wala functionGETrequest par,**/**address par chalega |
return {...} | Pythondictionarykhud ba khudJSONme badal jati hai |
--reload | Code badlo to serverkhud dobarashuru ho (sirf development me) |
35.3 Trained model ko /predict endpoint se serve karna
Ab asli kaam: Chapter 34 me save kiya hua iris_model.joblib API me lagate hain. Do cheezein dhyan se dekho: model ek baar, shuru me load hota hai (har request par nahi), aur /predict POST hai kyunke hum data bhej rahe hain. File ka naam iris_api.py:
iris_api.py
import joblib
from fastapi import FastAPI
from pydantic import BaseModel, Field
app = FastAPI(title="Iris Predictor")
model = joblib.load("iris_model.joblib")
CLASS_NAMES = ["setosa", "versicolor", "virginica"]
class IrisInput(BaseModel):
sepal_length:
float = Field(gt=0, lt=15)
sepal_width:
float = Field(gt=0, lt=15)
petal_length:
float = Field(gt=0, lt=15)
petal_width:
float = Field(gt=0, lt=15)
@app.get("/health")
def health():
return {"status": "ok"}
@app.post("/predict")
def predict(data: IrisInput):
features = [[data.sepal_length, data.sepal_width,
data.petal_length, data.petal_width]]
probs = model.predict_proba(features)[0]
idx = int(probs.argmax())
return {"species": CLASS_NAMES[idx], "confidence": round(float(probs[idx]), 3)}
Server chalao (uvicorn iris_api:app --reload) aur do alag phoolon ke naap bhej kar dekho:
TERMINAL
curl http://127.0.0.1:8000/health
curl -X POST http://127.0.0.1:8000/predict -H "Content-Type: application/json" \
-d '{"sepal_length":5.1,"sepal_width":3.5,"petal_length":1.4,"petal_width":0.2}'
curl -X POST http://127.0.0.1:8000/predict -H "Content-Type: application/json" \
-d '{"sepal_length":6.7,"sepal_width":3.0,"petal_length":5.2,"petal_width":2.3}'
OUTPUT
{"status":"ok"}
{"species":"setosa","confidence":1.0}
{"species":"virginica","confidence":1.0}
35.4 Pydantic se Input Validation
Agar koi -1 cm ya "abc" bhej de, to model ko kharab data milega aur API bina bataye ajeeb jawab dega. Pydantic is masle ko hal karta hai: tum IrisInput class me bata dete ho ke har field kaisa hona chahiye, aur FastAPI model tak pohanchne se pehle hi check kar leta hai.
| Pydantic ka hissa | Kya karta hai |
|---|---|
class IrisInput(BaseModel): | Input ka**dhancha (schema)**batata hai |
sepal_length: float | Is field ki qismfloathoni chahiye ("abc" nahi chalega) |
Field(gt=0, lt=15) | Qeemat0 se bariaur15 se chhotiho (gt = greater than, lt = less than) |
| Field ghayab ho | Khud ba khud422error aur saaf wazahat |
Ab galat input bhej kar dekho (pehle me negative qeemat, doosre me ek field ghayab):
TERMINAL
curl -X POST http://127.0.0.1:8000/predict -H "Content-Type: application/json" \
-d '{"sepal_length":-1,"sepal_width":3.5,"petal_length":1.4,"petal_width":0.2}'
curl -X POST http://127.0.0.1:8000/predict -H "Content-Type: application/json" \
-d '{"sepal_length":5.1,"sepal_width":3.5,"petal_length":1.4}'
OUTPUT
{"detail":[{"type":"greater_than","loc":["body","sepal_length"],
"msg":"Input should be greater than 0","input":-1,"ctx":{"gt":0.0}}]}
{"detail":[{"type":"missing","loc":["body","petal_width"],"msg":"Field required",
"input":{"sepal_length":5.1,"sepal_width":3.5,"petal_length":1.4}}]}
(Asal jawab ek hi line me aata hai; yahan parhne ke liye do lines me toda hai.) Dono baar status code 422 aaya (server ke log me 422 Unprocessable Entity dikhta hai), aur loc batata hai kis field me galti hai. Model tak ye data pohancha hi nahi. Yehi validation ka faida hai: sasta, tez aur mehfooz .
35.5 Swagger Docs (/docs) se Testing
FastAPI ki sab se pyari khasiyat: tumhari API ke liye khud ba khud ek testing page ban jata hai. Server chalte hue browser me kholo: http://127.0.0.1:8000/docs. Is page ko Swagger UI kehte hain.
-
Browser me
http://127.0.0.1:8000/docskholo. Tumhe tumhare saare endpoints ki list nazar aayegi (/health,/predict). -
POST /predictpar click karo, phir Try it out dabao. -
Request body wale box me apna JSON likho (ya diye gaye example me qeemat badlo).
-
Execute dabao. Neeche Response body me jawab aur Code (200 ya 422) nazar aayega.
-
Schema section me tumhara
IrisInputbhi dikhta hai, matlab dusron ke liye documentation tayyar hai.
| Address | Kya hai |
|---|---|
/docs | Swagger UI: interactive testing page |
/redoc | Doosra, parhne me asaan documentation page |
/openapi.json | API ka poora naqsha machine ke parhne wali shakal me |
35.6 Automatic Testing: TestClient aur pytest
Har baar haath se /docs me click karna thaka dene wala hai, aur CI/CD (Chapter 38) me ye kaam khud hona chahiye. Is liye hum tests code me likhte hain . FastAPI ka TestClient server chalaye baghair API ko call karta hai. File ka naam test_iris_api.py:
test_iris_api.py
from fastapi.testclient import TestClient
from iris_api import app
client = TestClient(app)
GOOD = {"sepal_length": 5.1, "sepal_width": 3.5,
"petal_length": 1.4, "petal_width": 0.2}
def test_health():
res = client.get("/health")
assert res.status_code == 200
assert res.json() == {"status": "ok"}
def test_predict_ok():
res = client.post("/predict", json=GOOD)
assert res.status_code == 200
assert res.json()["species"] == "setosa"
def test_predict_negative_value():
res = client.post("/predict", json={**GOOD, "sepal_length": -1})
assert res.status_code == 422
def test_predict_missing_field():
bad = {k: v for k, v in GOOD.items() if k != "petal_width"}
res = client.post("/predict", json=bad)
assert res.status_code == 422
TERMINAL
pytest -v
OUTPUT
test_iris_api.py::test_health PASSED [ 25%]
test_iris_api.py::test_predict_ok PASSED [ 50%]
test_iris_api.py::test_predict_negative_value PASSED [ 75%]
test_iris_api.py::test_predict_missing_field PASSED [100%]
========================= 4 passed, 1 warning in 1.56s =========================
35.7 Production ke liye chhoti chhoti achhi aadatein
| Aadat | Kyun |
|---|---|
/health endpoint rakho | Docker, load balancer aur monitoring isi se pata karte hain ke API zinda hai |
response_model / Pydantic sejawabka formatbhi tay karo | Jawab hamesha ek jaisa aur documented rahega |
| Modelek baarload karo | Har request par load karne se API sust ho jati hai |
Qeemton kihadd(gt, lt) lagao | Bekaar input model tak na pohanche |
| Tests likho | Har badlav ke baad API ke toot jane ka darr khatam |
35.8 Aam Galtiyan aur Hands-on
AAM GALTIYAN
-
uvicorn iris_api:appme file ka naam aur object ka naam galat likhna (iris_apifile hai,appobject). -
• Server ko
localhostpar chala kar sochna ke doosre log bhi dekh lenge. Doosron tak pohanchane ke liye deployment chahiye (Docker + hosting). -
Model ko
predict()ke andar load karna (API bohat sust ho jati hai). -
Validation na lagana aur model ko kuch bhi bhejne dena.
-
--reloadko production me chalana. Ye sirf development ke liye hai.
Hands-on
-
Apne kisi purane project (maslan Titanic ya Iris) ka model
joblibse save karo aur uski/predictAPI banao. -
Field(gt=..., lt=...)ki hadd badlo aur dekho kaun si qeemat 422 deti hai. -
/docskholo aur Swagger se 3 alag predictions karo. -
test_iris_api.pyme ek aur test likho:/healthka jawab galat hone par test fail ho (jaan boojh kar jawab badal kar dekho).
KEY TAKEAWAYS
-
API model ka waiter hai: request lekar model tak pohnchati aur JSON response wapas laati hai.
-
GET maloomat lene ke liye, POST data bhej kar prediction lene ke liye.
-
FastAPI me
@app.get(...)/@app.post(...)decorator se endpoint banta hai, aur Uvicorn server chalata hai. -
Model ek baar load karo, aur
/predictse serve karo; Pydantic galat input ko 422 de kar rok deta hai. -
/docs(Swagger) me testing asaan hai, aurTestClient+pytesttesting ko automatic bana dete hain.