Jo model browser mein chal sake wo itna chhota hota hai ki do sau page apne context mein rakh hi nahin sakta, isliye poora document dena vikalp hai hi nahin. Uske badle document overlapping hisson mein kaata jaata hai, aur har sawaal ke liye wo hisse dhoondhe jaate hain jinme sawaal ke sabse khaas shabd milte hain, aur sirf wahi padhe jaate hain. Har jawab un pages ko ginaata hai jinse wo aaya, aur wo hisse neeche dikhte hain, taaki daawa document se jaancha ja sake. Model WebGPU se aapke apne hardware par chalta hai.
Features
Document overlapping hisson mein kata jaata hai, page numbers ke saath
Khoj kam istemaal hone wale shabdon ko zyada bhaar deti hai
Har jawab istemaal kiye pages cite karta hai, hisse neeche dikhte hain
Nau models, 350 MB se 2.2 GB tak, pehli download ke baad cached
Aage ke sawaalon mein haal ki baatcheet saamne rehti hai
Chetavani jab document mein poochhne layak text hi bahut kam ho
Poori baatcheet text ki tarah copy karein
PDF se baat karein kaise use karein
PDF ko page par drop karein
Model chunein — pehla istemaal use ek baar download karta hai
Sawaal poochein, ya diye gaye sujhaavon se shuru karein
lease.pdf — "What notice period does either party have to give?"
Output
Two months in writing, by either party. (p. 3) · Answered from Page 3
Istemaal kiya hissa neeche dikhta hai taaki daawa jaancha ja sake.
Common errors aur troubleshooting
Browser model nahin chala paa raha. — Use WebGPU chahiye, jo desktop Chrome aur Edge dete hain. Uske bina PDF Summarizer chalta hai aur Extract Text shabd de deta hai jinhe kahin aur istemaal kar sakte hain.
Bahut kam text padha ja saka. — PDF ek scan hai. Pehle text recognition chalayein; bina text layer wale page se nikaalne ko kuch hai hi nahin.
Jawab galat hai par pakka lag raha hai. — Chhote models aisa karte hain. Cite kiye hisse kholein — jawab unme nahin hai to wo banaya gaya tha, aur bada model aam taur par madad karta hai.
Pehla sawaal bahut dheema hai. — Model download ho raha hai, jo ek hi baar hota hai aur phir cache ho jaata hai. Iske baad ke sawaal turant shuru hote hain.
Aksar pooche jaane wale sawaal
Kya mera document kisi server par jaata hai?
Nahin. Text isi tab mein nikalta hai, khoj yahin chalti hai, aur model aapke apne GPU par chalta hai. Model ke weights ek baar sarvajanik host se aate hain aur browser unhe cache karta hai.
Wo sirf kuch hisse hi kyun padhta hai?
Kyunki is naap ke model ka context seemit hota hai. Chaar sabse mel khaate hisse dena document ko kisi manmaani jagah se kaat dene se behtar jawab deta hai.
Chat ke liye kaunsa model theek hai?
1B wale sujhaaye gaye se shuru karein — lagbhag 750 MB, seedhe sawaalon ke liye kaafi. Jawab patle lagein to upar jaayein, aur dhyaan rakhein ki bade models ko sakshham GPU chahiye.
Kya wo poore document par sawaal ka jawab de sakta hai?
Kamzor tareeke se. Khoj hisse dhoondti hai, isliye «clause 7 kya kehta hai» achha chalta hai aur «200 page ki har zimmedari ka saaraansh do» nahin. Doosre ke liye summarizer istemaal karein.
Kya citation garantee hai ki jawab sahi hai?
Wo aapko jaanchne deta hai, jo asli maqsad hai. Chhote models diye gaye hisse ko bhi galat padh lete hain, isliye citation sabooti nahin, jaanchne ki shuruaat hai.
Related tools
PDF summarizer — Lambe PDF ka section-dar-section sar apne device par banayein.
PDF se Text — PDF se selectable text plain text ya Markdown ke roop mein extract karein.
PDF OCR — Scanned PDF ka text pehchaan kar use searchable banayein.
Audio se PDF — Recording ko device par transcribe karein aur PDF, SRT ya VTT mein lein.
PDF Compare karein — Do PDFs ke extracted text ko line-by-line compare karein.