📈 My Skool Community (ALL Resources): [ Ссылка ]
🔮 Make: [ Ссылка ] (*)
🌲 Pinecone: [ Ссылка ]
🤖 OpenAI: [ Ссылка ]
👾 Canva: [ Ссылка ]
🔥 RSS: [ Ссылка ] (*)
🧑💻 Carrd: [ Ссылка ] (*)
🥟 Dumpling: [ Ссылка ] (*)
💸 50 RAG Business Ideas 💸
[ Ссылка ]
🚦Automation Image: [ Ссылка ]
💬 Chatbot code: [ Ссылка ]
Variables:
{{replace(replace(replace(replace(replace(replace(replace(32.text; "/\n/g"; space); "/\r/g"; space); "/\t/g"; space); "/\f/g"; space); "/\//g"; "/"); "/\\/g"; "\\"); "/""/g"; "\""")}}
OpenAI Embeddings:
{
"input": "{{33.`JSON String`}}",
"model": "text-embedding-3-small",
"encoding_format": "float"
}
This is a monster automation about Retrieval Augmented Generation (RAG)
👾 Scrape ANYTHING (YouTube, entire websites, social media)
🧽 Cleanse and standardise that data
📊 Uploaded it to a effectively infinite storage base
💎 Chat with the data on a website
🚨 What problem does this solve 🚨
OpenAI Assistants handle small data (10-20 pages) and offer simple search. They have limited flexibility for updates and changes.
🧠 RAG & Vectors 🧠
Imagine we have 1 million documents and we place each document into a mystical bookshelf with 1 million draws. AI creates a unique identifier (a series of numbers) for each draw. We call this a vector. When we ask a question, the AI turns the question into a vector and finds the best 5 draws to open based on the similarity of the vectors.
⭐️ Why this is better ⭐️
Vector systems let us chat with billions of entries and only retrieve the information we need. They are more reliable and flexible. We can also add knowledge to them automatically. In our draw analogy, ChatGPT would open each draw for every question (not practical), which is why we use RAG.
TL;DR: This is HUGE.
Stamps ⌚️
00:00 Intro
01:15 Overview
01:32 Why use RAG?
03:01 Open AI Assistant vs RAG
04:33 How this works
09:10 Blueprints
09:46 Scrape & Cleanse YouTube
18:05 Scrape & Cleanse Websites
22:52 Upsert Vectors
35:00 Testing scenario 1 & 2
36:34 Chatbot
