Window us sab mein baantti hai jo aap bhejte hain aur us jagah mein jo jawab ko chahiye — aur yahi doosra hissa budgets bhool jaate hain. Jo request jodne par seema se theek neeche baithti hai wo asal mein samati nahi, kyunki jawab ke liye jagah hi nahi bachti. Yahaan sab kuchh kholkar rakha jaata hai: system prompt, tool definitions, documents, ab tak ki conversation aur ye turn, saath mein jawab ke liye rakhi gayi jagah — sab model ki asli window ke saamne. Ye batata hai ki samata hai ya nahi aur kitne se, aur us hisse ko named karta hai jo sabse zyada ghera hai: wo lagbhag hamesha history hota hai, aur lagbhag hamesha sabse pehle usi par kaam banta hai.
Features
Request ka har hissa asli window ke saamne naapa gaya
Jawab ke liye rakhi gayi jagah — budget ko imaandaar yahi banati hai
Model jitna banayega usse zyada jagah rakhi ho to warning
History window par qabza kar le to warning
Ek chadhi hui patti, jo dikhati hai window asal mein kahaan jaati hai
Bacha hua ya zyada ho gaya hissa sankhya mein
Jaanch ki tithi ke saath models ki windows
Har field editable, aapki request ke hisaab se
Context Window Planner kaise use karein
Lakshit model chuniye
Request ke har hisse ka aakaar bhariye
Jawab ke liye paryapt tokens rakhiye
Padhiye ki samata hai ya nahi aur kaun ghera hai
Example
Input
32,100 sent + 4,000 reserved
Output
Fits, with 963.9K to spare
Rakhi gayi jagah bhi usi mein hai jo samana chahiye — baad ki soch nahi.
Common errors aur troubleshooting
Mere aankde seema se neeche hain phir bhi kehta hai nahi samayega. — Jawab ko bhi samana hai. Bandha hua aankda wo hai jo aap bhejte hain aur jo aapne rakha.
Mujhe apne hisson ke aakaar nahi pata. — Kisi khaas text ke liye token counter se naapiye. Ye tool budget ke aakaar ke liye hai, sateek maap ke liye nahi.
Rakhi gayi jagah bahut badi bataayi ja rahi hai. — Har model ki ek seema hai ki wo ek jawab mein kitna banayega, aur wo window se alag hai. Seema field ke neeche dikhti hai.
History window par haavi hai. — Yahi wo warning hai jis par kaam banta hai. Compression, context ka editing ya purani turns ka summary yahaan kisi bhi aur badlav se zyada karega.
Aksar pooche jaane wale sawaal
Jawab ke liye tokens kyun rakhein?
Kyunki jawab usi window mein banta hai. Input use bhar de to jawab dene ki jagah nahi bachti — nateeja kata hua jawab ya error, budget ke baare mein koi saaf sandesh nahi.
Window aur output limit mein kya farq hai?
Window ye hai ki model ek saath kitna sambhaal sakta hai. Output limit ye ki wo ek jawab mein kitna banayega, aur wo kaafi chhoti hoti hai. Dono jaanchi jaati hain, kyunki kisi ek se bhi aage jaana fail hota hai.
Kya aankde sateek hain?
Window ke aakaar prakashit hain, jaanch ki tithi ke saath. Hisson ke liye jo aap bharte hain wo aapka apna anumaan hai, isliye nateeje ko budget ka aakaar maaniye, guarantee nahi.
Na samaye to kya karein?
Dekhiye kaun-sa hissa sabse bada hai — lagbhag hamesha history ya laaye gaye documents. Kam laana aur purani turns ka summary banana system prompt chhaantne se kahin zyada madad karta hai.
Kya badi window se baat ban jaati hai?
Wo jagah khareedti hai, dakshata nahi. Window mein jo kuchh hai wo har request par chukaya jaata hai, isliye badi window bhar dene wali conversation usi hisaab se mehngi padti hai.