LLM Hub

timmy boy

Mayelana nalolu hlelo lokusebenza

I-LLM Hub ikulethela i-AI yezinga lokukhiqiza ngqo kudivayisi yakho ye-Android — iyimfihlo, iyashesha, futhi yasendaweni ngokugcwele. Qalisa ama-LLM akudivayisi esimanje (i-Gemma-3, i-Gemma-3n multimodal, i-Llama-3.2, i-Phi-4 Mini) anamafasitela amakhulu onomongo, inkumbulo yomhlaba eqhubekayo, kanye nesizukulwane sokubuyisa-ukubuyisa (RAG) esisekela izimpendulo kumadokhumenti anenkomba agcinwe kudivayisi. Dala futhi ugcine okushumekiwe kwamadokhumenti namanothi, sebenzisa ukusesha ngokufana kwe-vector endaweni, futhi unothisa izimpendulo ngosesho lwewebhu olunamandla e-DuckDuckGo lapho udinga amaqiniso abukhoma. Yonke into ebalulekile ihlala ocingweni lwakho ngaphandle kokuthi uyikhiphe ngokusobala: inkumbulo yendawo kuphela, izinkomba, nokushumeka kuvikela ubumfihlo bakho ngenkathi iletha ukuhambisana nokunemba okuphezulu.

Izici Eziyinhloko

Ukuqondiswa kwe-LLM ekudivayisi: Izimpendulo ezisheshayo, eziyimfihlo ngaphandle kokuncika kwamafu; khetha amamodeli afana nedivayisi nezidingo zakho.
I-Retrieval-Augmented Generation (RAG): Hlanganisa ukucabanga kwemodeli nezingxenyana zedokhumenti enenkomba nokushumeka ukuze ukhiqize izimpendulo ezisekelwe eqinisweni.
Inkumbulo Yomhlaba Wonke Eqhubekayo: Londoloza amaqiniso, amadokhumenti, nolwazi kumemori eqhubekayo, yedivayisi yendawo (I-Room DB) ukuze uthole ukukhumbula isikhathi eside kuzo zonke izikhathi.
Okushumekiwe Nosesho Lwe-Vector: Khiqiza okushumekiwe, okuqukethwe kwenkomba endaweni yakini, futhi ubuyise amadokhumenti afaneleka kakhulu ngokusesha okuphumelelayo okufanayo.
Ukusekelwa kwe-Multimodal: Sebenzisa amamodeli anekhono lombhalo + wesithombe (Gemma-3n) ukuze uthole ukusebenzisana okucebile uma kutholakala.
Ukuhlanganiswa Kosesho Lwewebhu: Yengeza ulwazi lwendawo ngemiphumela yewebhu enamandla ye-DuckDuckGo ukuze ulande ulwazi lwakamuva lwemibuzo ye-RAG nezimpendulo ezisheshayo.
Okungaxhumekile ku-inthanethi: Sebenza ngaphandle kokufinyelela kunethiwekhi — amamodeli, inkumbulo, nezinkomba kuyaqhubeka kudivayisi.
Ukusheshisa kwe-GPU (ongakukhetha): Zuza ekusheshisweni kwehadiwe lapho kusekelwa khona — ukuze uthole imiphumela ehamba phambili ngamamodeli amakhulu asekelwa yi-GPU sincoma amadivayisi ane-RAM okungenani engu-8GB.
Idizayini Yobumfihlo-Yokuqala: Inkumbulo, ukushumeka, nezinkomba ze-RAG zihlala zisendaweni ngokuzenzakalelayo; akukho ukulayishwa kwamafu ngaphandle kokuthi ukhethe ngokusobala ukwabelana noma ukuthekelisa ngedatha.
Ukuphatha Okuqukethwe Okude: Usekelo lwamamodeli anamafasitela amakhulu womongo ukuze umsizi akwazi ukubonisana ngemibhalo ebanzi nemilando.
Ilungele Unjiniyela: Ihlanganisa nencazelo yendawo, inkomba, kanye nezimo zokusebenzisa zokubuyisa izinhlelo zokusebenza ezidinga i-AI eyimfihlo, engaxhunyiwe ku-inthanethi.
Kungani ukhethe i-LLM Hub? I-LLM Hub yakhelwe ukuletha i-AI eyimfihlo, enembile, futhi evumelana nezimo kuselula. Ihlanganisa isivinini sokucatshangelwa kwendawo kanye nesisekelo esiyiqiniso sezinhlelo ezisekelwe ekubuyiseni kanye nokuba lula kwenkumbulo eqhubekayo - ilungele izisebenzi zolwazi, abasebenzisi abaqaphela ubumfihlo, kanye nonjiniyela abakha izici zasendaweni zokuqala ze-AI.

Amamodeli Asekelwe: Gemma-3, Gemma-3n (multimodal), Llama-3.2, Phi-4 Mini — khetha imodeli efanelana namandla edivayisi yakho kanye nezidingo zomongo.

Yini entsha

- Upgraded Phi-4 Mini Max context window to 4096 and enabled GPU backend
- Model loading configuration now remembers your last settings
- Added translation support for Italian