Notes - terrytaylorbonn/auxdrone GitHub Wiki

(internal notes)

25.0209

  • https://www.youtube.com/@aiDotEngineer
  • GraphRAG: The Marriage of Knowledge Graphs and RAG: Emil Eifrem AI Engineer https://www.youtube.com/watch?v=knDDGYHnnSI
    • 89,920 views Aug 28, 2024 A famous poet once said "Natural language is most powerful when it can draw from a rich context." Ok fine, I said that. But that's true of both poetry, and of LLMs! Well, Knowledge Graphs excel at capturing context. How can combining Knowledge Graphs with RAG ... an emerging technique known as GraphRAG – give context to your RAG application, and lead to more accurate and complete results, accelerated development, and explainable AI decisions? This talk will go deep on the why and how of GraphRAG, and where best to apply it. You’ll get concepts, examples, and specifics on how you can get started. You’ll walk away with an understanding of how GraphRAG can improve the context you pass to the LLM and the performance of your AI applications.
  • https://www.proplaintiff.ai/

25.0131

25.0130 openapi

xxxx

25.0129

Group1 (Neupane)

Group2 google

Group3 API the docs


25.0215 (0121) Gdrive

0.0 Concepts


what ai is not: They Are Lying To You... Awesome https://www.youtube.com/watch?v=m0J8smwFvPI

but, ai could eventually avoid tool lock in: https://youtu.be/aTXdXIEFx6c?t=524

image











Content

  • 1 AI search engines are priceless
  • 2 Can AI cover Swagger / GraphQL ?
  • 3 Integrate LLM into site
  • 4 Train LLM against site docs

1 AI search engines are priceless

The binary algorithms of AI have 0 intelligence and could never have created the API sandbox (or ZiptieAI drone) wikis (AI "training" and "learning" also have nothing in common with human training and learning). But they could now "train" on the content, repackage, add TTS and STT (speech to text), OCR, etc, and then make it much easier to replicate these wikis.

These AI tools are still basically glorified search engines (with genAI to patch things together) with TTS/STT front end. Note that the Google AI websites are STILL the typical google chaos, with no AI to help get started :) . Copilot focues on code, adding all the real intelligence that went in to building compilers.

But that search engine aspect is priceless. AI does all the searching through the endless verbose and confusing official documentation so that you dont have to. AI tools simply give you

  • Answers to your questions
  • Real-world examples (based on your existing code!)

For those who repack manufacturers content into user-friendly instructions, AI is a real threat:


2 Can AI cover Swagger / GraphQL ?

  • Swagger does not provide complete API directions.. could AI do this? (I have always done this in documentation; but readers always go to the Swagger as source of truth).
  • How to train AI to provide exact GraphQL query to exact questions?

3 Integrate LLM into site

  • Searching the site ??? how?
  • Can I train an LLM against my own website docs / API-docs?
  • The goal: AI assist for the website product

4 Train LLM against site docs

https://apisjson.org/

image



image

AIEO (AI engine optimization; not seen this term anywhere, but makes sense)

How to get your product docs / API-docs optimized for AI code tools (Copilot, ChatGPT, etc).

https://apisjson.org/

As I understand such a JSON file for a site would define the API in a much more complete format than Swagger or GraphQL. THe webcrawlers discover this file, train thair AI tool on it, then users can use standard AI tools (Copilot, GPT, Gemini, Claude.ai, etc) to get detailed targetedanswers to specific API use cases for your site. BINGO!

⚠️ **GitHub.com Fallback** ⚠️