Lesson 02 � Foundation

PYTHON SE
AI BANAO.

Python AI engineering ki language hai � FastAPI se API banao, Pydantic se data validate karo, Uvicorn se server chalao, aur Async se speed pakdo. Sab kuch Python mein hota hai.

WHY: Python AI mein kyun hai?

Jab aapko ML model deploy karna ho, API banana ho, ya data pipeline chalani ho � Python sabse easy hai. FastAPI framework se production-ready API minutes mein ban jaati hai. Pydantic data ko safe rakhta hai. Uvicorn server ko light aur fast banata hai. Async multiple requests ko ek saath handle karta hai.

Concepts ka deep dive

FastAPI � API framework

FastAPI ek modern, fast web framework hai Python ke liye. Yeh type hints ka use karta hai automatically validation aur docs generate karne ke liye. Production mein use hota hai � Netflix, Uber, Microsoft sab use karte hain.

Pydantic � Data validation

Pydantic BaseModel se define karo ki request mein kya aana chahiye. Agar client galat data bheje toh Pydantic automatically error return karta hai. Aapko manually check nahi karna padta.

Uvicorn � ASGI Server

Uvicorn ek ASGI server hai jo FastAPI app ko run karta hai. Yeh asyncio aur uvloop use karta hai for maximum performance. Development mein --reload flag se hot reload milta hai.

Async � Non-blocking execution

Async functions (async def) non-blocking hote hain. Jab ek async function kisi external call ka wait kar raha hai (jaise database ya API call), tab server doosri requests process kar sakta hai. Isse server bahut efficient banta hai.

Code: Prediction API banana hai

Ab ek simple prediction API banate hain FastAPI se. Yeh code directly copy karke run kar sakte ho.

# FastAPI basics
from fastapi import FastAPI
from pydantic import BaseModel
import uvicorn

app = FastAPI()

class PredictionRequest(BaseModel):
 features: list[float]

class PredictionResponse(BaseModel):
 prediction: float
 confidence: float

@app.post("/predict")
async def predict(request: PredictionRequest):
 # Simple prediction
 prediction = sum(request.features) / len(request.features)
 return PredictionResponse(prediction=prediction, confidence=0.85)

@app.get("/health")
async def health():
 return {"status": "healthy"}

# Run: uvicorn main:app --reload
print("FastAPI app ready!")

How it works � step by step

• Step 1: PredictionRequest define karta hai ki client ko kya bhejna hai � ek list of floats.
• Step 2: /predict endpoint receive karta hai request, average calculate karta hai, aur PredictionResponse return karta hai.
• Step 3: /health endpoint ek simple health check hai � load balancer ke liye useful.
• Step 4: uvicorn main:app --reload se server start hota hai aur http://localhost:8000/docs pe auto-generated Swagger docs milte hain.

Async kyun zaroori hai?

Imagine karo tumhara API 100 users ko serve kar raha hai. Agar synchronous code likha toh ek request process hone tak baaki sab wait karengi. Async mein jab ek request database call kar rahi hai, tab server doosri request handle kar sakta hai � jaise ek chef jo ek saath 10 bartan pakad sake.

# Sync vs Async comparison
import time
import asyncio

# Slow synchronous function
def slow_sync(n):
 time.sleep(1) # Blocks everything
 return f"Sync done: {n}"

# Fast async function
async def fast_async(n):
 await asyncio.sleep(1) # Releases control
 return f"Async done: {n}"

# Run 3 tasks
async def main():
 # Async: all 3 run in parallel (~1 sec total)
 results = await asyncio.gather(
 fast_async(1), fast_async(2), fast_async(3)
 )
 print(results)

asyncio.run(main())

Exercise: Test your knowledge

Pydantic � Data safety net

AI systems mein data galat aana common hai. Client bhejde string jahan float hona chahiye, ya required field miss ho jaaye. Pydantic ye sab automatically handle karta hai � aapko manually check nahi karna padta.

# Advanced Pydantic usage for AI
from pydantic import BaseModel, Field
from typing import Optional

class ModelConfig(BaseModel):
 model_name: str = Field(..., description="Model ka naam")
 temperature: float = Field(0.7, ge=0.0, le=2.0)
 max_tokens: int = Field(100, gt=0)
 system_prompt: Optional[str] = None

# Valid config
config = ModelConfig(model_name="gpt-4", temperature=0.8)
print(f"Model: {config.model_name}")
print(f"Temp: {config.temperature}")

# Invalid: temperature 5.0 > 2.0 limit
# config = ModelConfig(model_name="gpt-4", temperature=5.0)
# PydanticError: value is not <= 2.0

AI Engineering Python Best Practices

• Type hints use karo: def predict(features: list[float]) -> dict: � Python ko samjhata hai ki kya expect karna hai.
• Async: Database calls, API calls, file operations � sab async mein karo for better performance.
• Pydantic for everything: Request, response, config � sab BaseModel se define karo.
• Error handling: Try-except blocks lagao, especially ML model predictions mein.
• Logging: Print ki jagah logging module use karo � production mein debugging easy hogi.