## 🧱 Step-by-Step: How I Built My LLM Lab with Chainlit on Windows OS

[ngonde1]

---

🧩 Step 1: Tell PowerShell it's okay to run scripts

Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

🗣️ “Hey PowerShell, please allow me to run my own scripts safely.”
This unlocks the ability to activate environments and run Python code.

---

🧩 Step 2: Go to the folder where my project lives

cd "C:\-----\Desktop\LLMs-from-scratch"

🗂️ “Let’s walk into the room where all my GenAI files are stored.”

---

🧩 Step 3: Build a safe Python workspace

uv venv --python 3.11

🧪 “Make me a fresh lab where I can mix Python ingredients without messing up the kitchen.”
This creates a .venv folder with Python 3.11.

---

🧩 Step 4: Step into the lab

.\.venv\Scripts\Activate.ps1

🧼 “Put on my lab coat and start working inside the safe zone.”
Now every Python command uses this clean environment.

---

🧩 Step 5: Remove broken tools from earlier

uv pip uninstall -y llms-from-scratch llms_from_scratch 2>$null

🧹 “Throw away any old or broken versions of my GenAI package.”

---

🧩 Step 6: Create a special folder for my package

New-Item -ItemType Directory -Name pkg

📦 “Make a box called pkg to hold all my GenAI chapters.”

---

🧩 Step 7: Move my code into the box

Move-Item llms_from_scratch pkg

📤 “Put my GenAI code inside the pkg folder so Python can find it.”

---

🧩 Step 8: Install my code so Python knows it

uv pip install -e .

🔧 “Tell Python: this is my GenAI package, and I might change it later—please keep watching.”

---

🧩 Step 9: Make every folder readable by Python

Get-ChildItem pkg\llms_from_scratch\ch02 -Recurse -Directory | ForEach-Object { New-Item -Path $_.FullName -Name "__init__.py" -ItemType File }
New-Item pkg\llms_from_scratch\ch02\__init__.py

🧠 “Put a magic sticker (__init__.py) on every folder so Python knows it’s part of the team.”

Repeat this for:

ch03, ch04, ch05, ch06, ch07

✅ Now every chapter and subfolder is importable.

---

🧩 Step 10: Rename folders that confuse Python

Rename-Item "pkg\llms_from_scratch\ch05\01_main_code" "main_code"

🚫 Python doesn’t like folders starting with numbers.
✅ Renaming fixed the error: “invalid decimal literal”

---

🧩 Step 11: Find the tools I need

Get-ChildItem pkg\llms_from_scratch\ch05 -Recurse -Filter *.py | Select-String "def generate"

🔍 “Where’s the generate function hiding?”
Found it in main_code/gpt_generate.py

---

🧩 Step 12: Fix broken imports

In gpt_generate.py, change:

from previous_chapters import GPTModel

to:

from llms_from_scratch.ch04.gpt import GPTModel

In app.py, change:

from llms_from_scratch.ch05 import generate

to:

from llms_from_scratch.ch05.main_code.gpt_generate import (
    generate,
    text_to_token_ids,
    token_ids_to_text,
)

🧩 “Tell Python exactly where to find my tools.”

---

🧩 Step 13: Fix the model file path

In app.py, change:

model_path = Path("..") / "01_main-chapter-code" / "gpt2-medium355M-sft.pth"

to:

model_path = Path(__file__).resolve().parent.parent / "01_main_code" / "gpt2-medium355M-sft.pth"

📁 “Show Python the correct path to my trained model file.”

---

🧩 Step 14: Launch Chainlit!

uv run chainlit run pkg/llms_from_scratch/ch07/06_user_interface/app.py

🚀 “Start the GenAI app and open the user interface!”
✅ It worked! Chainlit opened at http://localhost:8000

---

🖼️ About the Image I Uploaded

The image shows a Chainlit UI giving some answers like:

• “A mammal is a bird or mammal.”
• “A bird is a bird of prey.”

• 🧠 What can be the problem, and how can I fix it?

---

🎉 Final Result

• ✅ All chapters from ch02 to ch07 are modular and importable
• ✅ Chainlit connects to my model and functions
• ✅ I can run everything from one place