Thuta Learning
ရှာဖွေရန်
AdvancedAIbeginner

Introduction to RAG

စိတ်လျှော့ပါ။ ဒီခန်းကို စာအုပ်လိုမဟုတ်ဘဲ စကားပြောသလိုပဲ၊ နားလည်လွယ်အောင် ရှင်းပါမယ်။

ဒီခန်းပြီးရင် ဘာတတ်သွားမလဲ

  • Introduction to RAG ကို ကြောက်စရာမလိုအောင် နားလည်မယ်
  • ကိုယ်တိုင် စမ်းသုံးတတ်အောင် လုပ်မယ်
  • မှားလွယ်တဲ့နေရာကို ပြုံးပြီး ရှောင်တတ်မယ်

Retrieval-Augmented Generation က ကိုယ့် document များထဲမှ ဆိုင်ရာအချက်ရှာပြီး LLM အဖြေကို ဘယ်လို ground လုပ်ပေးသလဲ လေ့လာမည်။

ခဏလေး ဒီလိုပဲ စဉ်းစားကြည့်

RAG မှာ user မေးခွန်းနဲ့ဆိုင်တဲ့ document အပိုင်းများကို အရင်ရှာပြီး၊ တွေ့တဲ့ context ကို prompt ထဲထည့်ကာ model ကိုဖြေခိုင်းပါတယ်။ Model ကိုပြန်လေ့ကျင့်ခြင်းမဟုတ်ဘဲ ဖြေချိန်မှာလိုအပ်တဲ့အချက်ပေးခြင်း ဖြစ်ပါတယ်။

Introduction to RAG lesson illustration
RAG and Vector Search — Introduction to RAG

နေ့စဉ်ဘဝနဲ့ ချိတ်ကြည့်မယ်

Company policy၊ product docs နဲ့ knowledge base Q&A လို source အခြေပြုအဖြေမှာ RAG အသုံးဝင်ပါတယ်။ Retrieval မကောင်းရင် generation လည်းမကောင်းပါဘူး။ Answer quality တစ်ခုတည်းမစစ်ဘဲ retrieved chunks ကိုပါ စစ်ရပါတယ်။

အတူတူ လက်တွေ့စမ်းမယ်

text
Documents → Chunk → Embed → Vector store

Question → Embed → Search ────────┘

          Context + Question → LLM → Answer
You should see
RAG pipeline ကို အဆင့်လိုက်ရှင်းပြနိုင်မည်။

၅ မိနစ် စမ်းကြည့်

FAQ document ငါးခုအတွက် RAG flow ကိုဆွဲပါ။ Retrieval မတွေ့သည့်အခါ “အချက်အလက်မတွေ့ပါ” ပြန်မည့် rule ထည့်ပါ။

သတိလေးတစ်ချက်

AI output ကို အပြီးသတ်အမှန်လို့မယူပါနှင့်။ အရေးကြီးတဲ့အချက်၊ code နဲ့ user data ကို လူကပြန်စစ်ပြီးမှ အသုံးပြုပါ။

OpenAI — RetrievalOpenAI

ဒီနေရာမှာ လူအများမှားတတ်တယ်

  • RAG ကို hallucination လုံးဝပျောက်စေမည့်နည်းလို့ယူဆခြင်း
  • Source မပြဘဲ answer တစ်ခုတည်းစစ်ခြင်း

လေ့ကျင့်ခန်း

FAQ document ငါးခုအတွက် RAG flow ကိုဆွဲပါ။ Retrieval မတွေ့သည့်အခါ “အချက်အလက်မတွေ့ပါ” ပြန်မည့် rule ထည့်ပါ။

You'll know it worked when: RAG pipeline ကို အဆင့်လိုက်ရှင်းပြနိုင်မည်။