LLM Hub - Local AI Assistant

timmy boy

Mayelana nalolu hlelo lokusebenza

🤖 Ihabhu ye-LLM - Umsizi Wakho Oyimfihlo we-AI

I-Advanced AI isebenza ngokuphelele ocingweni lwakho - ayidingeki i-inthanethi, ayikho idatha eyabiwe, ubumfihlo obuphelele buqinisekisiwe. Amamodeli Olimi Asezingeni eliphezulu ephaketheni lakho!

✨ AMATHULUSI E-AI AMAHLANU ANAMANDLA

📝 Ingxoxo Ehlakaniphile
Izingxoxo eziningi zokuqwashisa ngomongo, inkumbulo ye-RAG, ukusesha kwewebhu ongakukhetha, okukhiphayo kokuguqula umbhalo ube inkulumo ngokufunda ngokuzenzakalela, nokusekelwa kombhalo, izithombe, nokufaka umsindo.

✍️ USIZO LOKUBHALA
Fingqa amadokhumenti, wandise imibono, bhala kabusha okuqukethwe, uthuthukise uhlelo lolimi, futhi ukhiqize ikhodi ngezincazelo zolimi lwemvelo.

🌍 UMhumushi (50+ Izilimi)
Humusha umbhalo, izithombe (OCR), nomsindo ngesikhathi sangempela. Isebenza ngokungaxhunyiwe ku-inthanethi ngokuphelele ngosekelo lwe-bidirectional.

🎙️ TRANSCRIBER
Guqula umsindo ube umbhalo ngokunemba okuphezulu. Isekela i-WAV. Konke kucutshungulwa kudivayisi.

🛡️ I-SCAM DETECTOR
Hlaziya imilayezo nezithombe ezisolisayo ngemizamo yobugebengu bokweba imininingwane ebucayi. Thola ukuhlolwa okucacile kobungozi kanye nezincazelo ezinemininingwane.

🚀 AMAMODELI E-AI-CUTTING-EDGE

• I-Gemma-3 1B (Google) - Iyashesha futhi iyasebenza
• I-Gemma-3n E2B/E4B (Google) - I-Multimodal: umbhalo, umbono, umsindo
• I-Llama-3.2 1B/3B (Meta) - Umthombo ovulekile onamandla
• I-Phi-4 Mini (Microsoft) - Ilungiselelwe iselula

Wonke amamodeli asebenzisa u-100% kudivayisi esebenzisa ubuchwepheshe beMediaPipe neLiteRT.

🔐 UBUMFIHLO NOKUPHEPHA

• Ukuqoqwa kwedatha engu-Zero - izingxoxo zihlala kudivayisi yakho
• Ayikho i-inthanethi edingekayo ukuze kucatshangwe nge-AI
• Awekho ama-akhawunti, akukho ukulandelela, akukho okulayishwayo ngamafu
• Umthombo ovulekile futhi obala
• Qedela ukusebenza okungaxhunyiwe ku-inthanethi

⚡ IZICI EZIPHUMA

• Umbhalo-kuya-enkulumweni ngokufunda okuzenzakalelayo kokulalela kwe-hands-free
• Ukusheshisa kwe-GPU ukuze usebenze ngokushesha
• I-Multimodal: umbhalo, izithombe, nomsindo
• I-RAG enenkumbulo yomhlaba wonke ukuze uthole izimpendulo ezithuthukisiwe
• Ngenisa ngokwezifiso .task noma .litertlm amamodeli
• Ukulanda okuqondile okuvela ku-HuggingFace
• I-UI Yedizayini Yezinto Ezinhle
• 13 izilimi zokusebenzelana

📱 IZIDINGO

Okungenani: I-Android 8.0+, 2GB RAM, 1-5GB isitoreji
Okunconyiwe: 6GB+ RAM, 8GB+ yokusheshisa i-GPU

💡 INDLELA ESEBENZA NGAYO

1. Landa amamodeli e-AI kuhlelo lokusebenza (isikhathi esisodwa)
2. Khetha ithuluzi lakho: Ingxoxo, Usizo Lokubhala, Umhumushi, Umlobi, noma Umtshina Womkhonyovu
3. Sebenzisa i-AI ungaxhunyiwe ku-inthanethi ngokuphelele ngobumfihlo obugcwele
4. Ongakukhetha: Nika amandla ukusesha kwewebhu, layisha izithombe/umsindo, noma sebenzisa umbhalo ube inkulumo

🌟 PERFECT FOR

• Abasebenzisi abaqaphela ubumfihlo abazisa ukuphepha kwedatha
• Ochwepheshe abadinga usizo lwe-AI olungaxhunyiwe ku-inthanethi
• Abasebenzisi abafuna izimpendulo zezwi ze-handsfree
• Abafundi abasebenza ngokubhala nokucwaninga
• Abahambi abadinga ukuhumusha okungaxhunyiwe ku-inthanethi
• Noma ubani ovikela imikhonyovu
• Onjiniyela abazama nge-AI ekudivayisi
• Abadali bokuqukethwe nababhali

📖 OPEN SOURCE (MIT License)
github.com/timmyy123/LLM-Hub

🏆 IMANDLA YI
I-Google MediaPipe & LiteRT, Gemma, Llama, Phi models, HuggingFace

Izwa ikusasa losizo lwe-AI oluyimfihlo, olukudivayisi!

Yini entsha

Fixed an issue where the response was getting stuck with no content.