AI Path Finder
Free consultation

Lesson · Hinglish · ⏱ 8 min padhne mein

🧰 Tool use (function calling): LLM ko action lene dena

EnglishHinglish

🏠 Asli zindagi ki misaal

Smart receptionist khud doctor ka calendar nahi dekh sakti, lekin calendar system ko call kar sakti hai: 'Kya Dr. Mehta 4 baje free hain?' Phir jawaab ke hisaab se aapko reply karti hai. Tool use LLM ko yahi power deta hai. Woh khud live data nahi la sakta, lekin aapke code se function chalwa sakta hai aur result use kar sakta hai.

💡 Aasaan shabdon mein

Aap model ko tools describe karte ho: naam, har tool kya karta hai, aur uske input parameters JSON schema mein. Jab sawaal ke liye tool chahiye, model text ki jagah tool call reply karta hai ('get_order_status', order_id='A123'). Aapka code asli function chalata hai aur result wapas bhejta hai, aur model usse final jawaab likhta hai. Chatbots aise hi orders check karte hain, tickets book karte hain ya databases query karte hain. Control aapke code ke paas rehta hai: inputs validate karo, permissions check karo, aur refund jaise risky actions ke liye insaan ki approval rakho.

🎬 Dekho kaise kaam karta hai

UserLLMYour codeOrders APIWhere is order A123?tool_call get_order_status(A123)GET /orders/A123out for delivery · 7:40 pmtool resultArriving by 7:40 pm 🛵

Model khud API call nahi karta. Woh aapke code se kehta hai, aapka code chalata hai, aur model result se jawaab likhta hai.

🪜 Step by step

  1. Ek normal function likho, jaise get_order_status(order_id).
  2. Model ko iske baare mein batao: naam, kya karta hai, aur parameters ka JSON schema.
  3. Jab user poochhe "order 4512 kahan hai?", model tool call se jawaab deta hai: get_order_status, order_id 4512.
  4. Aapka code arguments validate karta hai, permissions check karta hai, function chalata hai aur result wapas bhejta hai.
  5. Model result se user ko jawaab deta hai, ya zaroorat ho to doosra tool call karta hai.

🔑 Zaroori shabd

Tool / function
Action jo aapka code model ko deta hai, jaise search ya calculator.
JSON schema
Tool ke inputs aur unke types ka description.
Human-in-the-loop
Important actions hone se pehle insaan approve karta hai.

👀 Example dekho

text
User:   "Where is my order A123?"
Model → tool_call: get_order_status(order_id="A123")
Your code runs it → {"status": "out for delivery", "eta": "7:40 pm"}
Model:  "Your order is out for delivery and should arrive by 7:40 pm 🛵"

Tool kab use karna hai, yeh model decide karta hai. Asli kaam aapka code karta hai.

⚠️ Aam galtiyaan jinse bachna hai

  • Arguments pe aankh band karke bharosa. Model order ID ya galat amount bana sakta hai; sab validate karo.
  • Bina limits ke powerful tools dena, jaise delete ya refund bina confirmation step ke.
  • Vague descriptions wale bahut saare milte julte tools. Model galat chunta hai; tools kam aur saaf described rakho.

✅ Jaldi se check karo

Tool call asal mein kaun execute karta hai?

  1. Aapka application code
  2. User
  3. LLM khud, model ke andar
Jawaab dekho

A. Aapka application code: Model sirf call request karta hai. Code aapka chalata hai, isliye permissions aur safety aapke control mein hain.

🙋 Seekhne wale yeh poochhte hain

Kya model mera code chalata hai?

Nahi. Woh sirf batata hai kaunsa tool kin arguments ke saath call karna hai. Aapki application use chalati hai aur tay karti hai allow karna hai ya nahi.

Kya ek saath kai tools call kar sakta hai?

Kai models parallel tool calls support karte hain, jaise teen warehouses mein ek saath stock check karna. Aapka code unhe chalata hai aur saare results lautata hai.

📌 Yaad rakho

Model poochta hai, aapka code karta hai. Sab kuch validate karo.

📖 Is lesson ke shabd (AI dictionary)

Poora course, aapke roadmap ke hisaab se

Progress tracking, quick checks aur step-by-step order ke saath. Yeh lesson in courses ka hissa hai:

AI Product Engineer · LLM / Agent Engineer · AI Forward Deployed Engineer

Course mein seekho