An interactive guide to picking the right AI type for any project.
___________ __ __
|__ __| / _\ \/ / A I F R A M E W O R K
| | | | |\ / / โโโโโโโโโโโโโโโโโโโโโโโโโ
| | | |__/ | | Which AI does your project
|__| \____/|_| actually need?
[ ML ] โโ patterns from labeled data
[ LLM ] โโ understand & generate language
[ GenAI ] โโ create images, audio, video
[ Agentic ] โโ plan, act, loop, automate
Start with the output shape โ not the tech. | thegr8val
TGV AI Framework is a single-page interactive reference for anyone who needs to decide which type of AI to use on a project โ before they've written a single line of code.
It's not academic. It's a decision-making tool: walk a tree, read the tradeoffs, pick your stack, and ship.
| What | Detail |
|---|---|
| ๐ Format | Single-page HTML โ no install, no dependencies |
| ๐ฏ Audience | Engineers, analysts, builders โ anyone starting an AI project |
| ๐ฎ Scope | ML, LLM, Generative AI, and Agentic AI |
| ๐จ Design | Matches thegr8val.github.io โ Space Mono + Syne, dark theme |
TGV-AIFramework/
โ
โโโ index.html โโ The full framework (single file, self-contained)
โโโ LICENSE โโ MIT
โโโ README.md โโ You are here
The framework is structured in 10 sections:
| Section | Content |
|---|---|
// 01 |
Decision Tree โ 4 questions to find your AI type |
// 02 |
The 4 Types โ ML, LLM, GenAI, Agentic at a glance |
// 03 |
Golden Rule โ start with the output shape, not the tech |
// 04 |
Common Mistakes โ 6 anti-patterns with fixes |
// 05 |
Hybrid Combinations โ ML+LLM, LLM+RAG, Agentic+RAG, ML+Agentic |
// 06 |
Tooling by Type โ concrete tools per AI category |
// 07 |
Cost & Effort Reality โ setup, inference, explainability, maintenance |
// 08 |
When to Level Up โ 5 upgrade triggers |
// 09 |
Sources & Docs โ official documentation links per type |
// 10 |
Questions? โ reach out if you're stuck |
Option A โ live version:
โ thegr8val.github.io/TGV-AIFramework
Option B โ run locally:
git clone https://github.com/TheGr8Val/TGV-AIFramework.git
cd TGV-AIFramework
open index.html # or just double-click itNo build step. No server required. One file.
Output is a label or score? โ ML
Output is text, language, or code? โ LLM
Output is something new / creative? โ GenAI
Task needs actions across steps? โ Agentic
When in doubt: start with LLM. Add complexity only when it earns its keep.
MIT โ thegr8val