🌟 Generative AI – Complete Self-Contained Learning Guide

Promise: Study this guide from start to finish and you will understand every major concept in modern Generative AI — from first principles to working projects. No external resources needed. Every analogy, every concept, every line of code is explained here.

---

📚 What's Inside?

File	Topic	What You'll Build
01_setup.md	Environment Setup	Working API connection + diagnostic script
02a_prompt_eng.md	Prompt Engineering (Theory)	Deep understanding of all prompting techniques
02b_prompt_project.md	Prompt Engineering (Project)	Prompt Playground web app with A/B testing
03_rag.md	RAG – Retrieval-Augmented Generation	PDF chat app + multi-document knowledge base
04a_function_calling_theory.md	Function Calling & Agents (Theory)	How AI uses tools and plans actions
04b_function_calling_project.md	Function Calling & Agents (Project)	Multi-tool AI agent with weather, news, math
05a_finetuning_theory.md	Fine-Tuning (Theory)	When and why to fine-tune vs other approaches
05b_finetuning_project.md	Fine-Tuning (Project)	Custom Shakespeare model on free Colab GPU
06_evaluation.md	Evaluation & Testing	Automated AI quality scoring harness
07_multimodal.md	Multi-modal AI	Image captioning, VQA, text-to-image app
08_vector_databases.md	Vector Databases (Deep Dive)	ChromaDB + persistent knowledge store
09_llm_internals.md	How LLMs Work Internally	Transformers, attention, tokenization explained
10_production.md	Production & Best Practices	Rate limiting, caching, cost control, deployment
cheat_sheet.md	Quick Reference	Every snippet, every model, one page

---

🗺️ Learning Path

START HERE
    │
    ▼
[01] Setup ──────────────────────────────────────────────────────────► Get API working
    │
    ▼
[09] How LLMs Work (optional but highly recommended early)
    │   Understand what's happening under the hood
    ▼
[02] Prompt Engineering ─────────────────────────────────────────────► Most used skill
    │   Theory first (02a), then build the playground (02b)
    ▼
[03] RAG ────────────────────────────────────────────────────────────► Most useful pattern
    │   Making AI read YOUR documents
    ▼
[08] Vector Databases (Deep Dive) ───────────────────────────────────► Powers RAG properly
    │
    ▼
[04] Function Calling / Agents ──────────────────────────────────────► AI that takes actions
    │
    ▼
[06] Evaluation ─────────────────────────────────────────────────────► Measure your work
    │
    ▼
[05] Fine-Tuning ────────────────────────────────────────────────────► Advanced: change style
    │
    ▼
[07] Multi-Modal ────────────────────────────────────────────────────► Vision + Language
    │
    ▼
[10] Production Best Practices ──────────────────────────────────────► Ship it!

---

🧠 Core Mental Model

Before you read anything else, internalize this:

┌─────────────────────────────────────────────────────────────────────────┐
│                      THE GENERATIVE AI STACK                            │
│                                                                         │
│  ┌─────────────┐  ┌─────────────┐  ┌──────────────┐  ┌─────────────┐  │
│  │    Input    │  │  Retrieved  │  │   LLM Core   │  │   Output    │  │
│  │             │  │  Context    │  │              │  │             │  │
│  │ • Text      │─►│             │─►│ • Reasoning  │─►│ • Text      │  │
│  │ • Images    │  │ • RAG docs  │  │ • Generation │  │ • JSON      │  │
│  │ • Audio     │  │ • Tool res. │  │ • Planning   │  │ • Images    │  │
│  │ • Files     │  │ • Memory    │  │ • Tool calls │  │ • Actions   │  │
│  └─────────────┘  └─────────────┘  └──────────────┘  └─────────────┘  │
│                                                                         │
│  What YOU control:  Prompt  │  Retrieved context  │  Temperature/params │
└─────────────────────────────────────────────────────────────────────────┘

The key insight: LLMs are fixed once trained. Everything that makes them useful in your application happens in what you feed them (prompt engineering + RAG) and what tools they can call (function calling).

---

🛠️ Prerequisites

Requirement	Level
Python basics	Know how to write a function, use pip, run a script
Terminal/Command line	Know how to cd, run commands
RAM	4GB minimum (8GB recommended)
Internet	Required for API calls
Time	15 min setup + ~1 hour per topic

No ML/math background needed. This guide builds understanding through analogy first, then code.

---

💻 What You'll Need to Install (Summary)

pip install huggingface_hub streamlit chromadb sentence-transformers \
            pymupdf requests python-dotenv transformers datasets accelerate

Full instructions in 01_setup.md.

---

🧩 How the Projects Connect

Every project in this guide builds on the last:

Prompt Playground  ──►  RAG PDF Chat  ──►  AI Agent  ──►  Full Assistant
     (02b)                 (03)             (04b)          (combines all)
      │                     │                │
      │  You learn to       │  You add       │  You add
      │  control output     │  document      │  actions +
      │  quality            │  knowledge     │  tool use

By the end, you'll have the building blocks for a production-quality AI assistant that:

• Answers from your documents (RAG)
• Can take actions (function calling)
• Understands images (multi-modal)
• Is systematically tested (evaluation)

---

📝 How to Use This Guide

1. Read the theory file first for each topic — don't skip analogies, they lock in understanding.
2. Copy and run every code snippet — reading code is not the same as running it.
3. Break things on purpose — change parameters, swap models, cause errors. Errors teach you what parameters do.
4. Do the challenges at the end of each project section.
5. Keep cheat_sheet.md open in a second window as quick reference.

Let's begin → Open 01_setup.md