ចំណាយ $0 លើ ChatGPT ! បង្កើត AI Agent ប្រើលើ Server ខ្លួនឯង (Local LLM) ជួយខ្ញុំគ្រប់គ្រង Server

ចំណាយ $0 លើ ChatGPT ! បង្កើត AI Agent ប្រើលើ Server ខ្លួនឯង (Local LLM) ជួយខ្ញុំគ្រប់គ្រង Server 🚀 ព័ត៌មានវគ្គសិក្សា Full-stack Development Bootcamp (ជំនាន់ទី៧): https://t.me/tfdTech/278 📚 វគ្គសិក្សា Data Science & AI ៖ https://t.me/tfdTech/236 🔗 ចូលរួមក្រុម Telegram របស់ខ្ញុំ៖ https://t.me/tfdTech នៅក្នុងវីដេអូនេះ ខ្ញុំនឹងបង្ហាញអ្នកទាំងអស់គ្នាអំពីរបៀបបង្កើត AI Agent ផ្ទាល់ខ្លួន ដើម្បីតាមដានសុខភាព Server (Home Lab) ដោយមិនចាំបាច់ចំណាយលុយលើ API Key របស់ ChatGPT ឬ Google Gemini ឡើយ។ In this video, I demonstrate how to build and deploy a local AI agent specifically designed to monitor Home Lab server health without spending a single cent on external API keys like ChatGPT or Gemini. យើងនឹងសិក្សាអំពីការប្រើប្រាស់ Ollama ដើម្បី Run Model តូចៗដូចជា IBM Granite 3.3 2B, Google Gemma 2 2B, និង TinyLlama 1B។ ខ្ញុំនឹងបង្ហាញពីរបៀបដែល AI យល់ពីសំណួររបស់យើង (Intent Classification) ហើយទៅទាញយកទិន្នន័យពី Server ដូចជា CPU, RAM និង Storage មកប្រាប់យើងតាមរយៈ Telegram បែបសង្ខេប និងឆ្លាតវៃ។ We explore the architecture using an AI VM running Ollama and compare three "Tiny LLMs": IBM Granite 3.3 2B, Google Gemma 2 2B, and TinyLlama 1B. You will see how we use these models as "Intent Classifiers" to turn Telegram questions into real-time server data (CPU usage, RAM, and Storage). We also discuss the challenge of "Hallucinations" in small models and why some models perform better than others for technical tasks. IG:   / darachaukh   YouTube:    / @tfdevs   Website: https://www.tfdevs.com/ Linkedin:   / qiang-cun-zhi   TikTok: https://www.tiktok.com/@chaudarakh?_r... Telegram Channel: https://t.me/tfdTech Facebook Page:   / chaudarascienceengineer   00:00 Introduction: AI summarizing Home Lab status 00:22 The benefits of AI in Home Lab monitoring 01:10 Building an AI Agent for free (No API keys) 01:36 Special Promotion: Full-stack Development Bootcamp Gen 7 02:25 Architecture: AI VM and Ollama Runtime setup 03:45 Using LLMs as "Intent Classifiers" 04:53 Workflow: From Telegram question to API data 06:19 Comparing Tiny LLMs: IBM Granite vs. Google Gemma vs. TinyLlama 07:48 Understanding AI Hallucinations in small models 08:31 Live Testing: Why TinyLlama 1B struggled 09:40 Live Testing: IBM Granite success with CPU & Storage 10:52 Live Testing: Google Gemma 2 performance 11:51 Conclusion: Privacy and cost-saving advantages