Lasted edited: March 6, 2025

A work in progress. Maybe.

I had this 1200 line main.py file that needed refactoring. The app, a media asset manager, was nowhere near completion. It was only gonna get longer. So, I thought, refactoring with AI shouldn’t take more than a day. A week later my main.py is 170 lines and the app ain’t exactly working like before. It’s worse. But I remain optimistic that I’m on the right track. Here’s what I would have done differently.

Make Sure You Know Git

Commit. Commit. Commit.

Make branches.

Commit. Commit. Commit.

Don’t start until you know git.

Commit. Checkout. Branch. Status. Tag. Push. Log. Stash. Pop. Commit. Commit. Commit.

If you don’t know what these commands mean, do not pass go. Go directly to git boot camp.

If you got a git question, ask the AI for advice. One thing is for sure, AI may struggle with refactoring 1200 lines of code, but AI knows git. Ask the AI questions about using git before you mess it up. If using cursor, or something other ide that uses AI, have the AI write the commit. It’ll be more thorough than you’ll ever be.

Remember git the secret sauce.

Small Bites Will Be Much Faster

AI may be great but once you start getting to large files, it runs out of tokens and fails.

It’ll be in the middle of writing a file and you’ll get the dreaded AI middle finger, “I’m done. Finish it yourself.”

So, if you think that the AI is gonna run out of tokens when writing a file, ask it to write the file in 4 sections, stopping after each section, until you give it the okay.

But that may the least of the problems.

Don’t Let It Change Your Code Automatically

Yeah, you’re starting out with Cursor AI and you’re rolling. The AI is typing away and all looks great until you try to run it.

Now, I’m not saying that AI can’t write the simple CRUD app. But when you’re working on refactoring a large file, you’ve now crossed the line.

Make sure to tell the AI, “DO NOT CHANGE ANYTHING> JUST TELL ME FIRST WHAT YOU PROPOSE AND WHY.”

Read over its response and you might be shocked. It’ll say things, “Assuming the function is named’ “happy-days”, here’s some code. It will write functions to log errors and put in fixes without first checking the logs. It will add files that duplicate functions.

Have the AI write the prompt for itself.

Before even starting with the refactoring process, have a conversation with the AI. Explain that you’re concerned about it running out of tokens.

Ask it how base to proceed. What files it needs to review. Go slow.

Don’t let the AI take you for a joyride

Keep constant control by using small incremental changes. By looking at the files created. If you don’t understand the code, ask the AI to explain the code.

Be careful of getting second opinions from different AI

It’s tempting to take a file from Claude and give it Char

#Other Problems

The AI will assume file-structure/file-names/function-names and files will get written that way.

Some Git Stuff to make using AI less painful

When first creating a branch, it’s best to push the branch immediately to the remote. This is because the first push of a new branch must include the flag -u to tell git to track the upstream branch. If you forget, it complicates stuff later.

So do a git push -u right after first creating a branch

Create branch

git checkout -b new-branch

Commit to new-branch with no changes (i.e. - “empty”) commit

git commit --allow-empty -m "Initial commit"

Now push with the -u flag (upstream)

git push -u origin new-branch

Now, it’s all good to go.

Some PostgreSQL Commands that might be useful

When changes are made to the models

Generate a new migration

alembic revision --autogenerate -m "Description of your changes"

Apply the migration

alembic upgrade head

To get into database with password

psql -U bread_user -d bread_db -h localhost

psql -U user_name -d databasename_db -h localhost

Common PostgreSQL Commands (Terminal/psql)

To switch to superuser from linux command line

sudo -u postgres psql

Basic Connection and Navigation:

  • psql: Opens the psql interactive terminal (default connection).
  • psql -U <username> -d <database_name>: Connect to a specific database as a user.
  • psql -h <hostname> -p <port> -U <username> -d <database_name>: Connect to a remote server.
  • \q or \quit: Exit psql.
  • \l or \list: List databases.
  • \c <database_name> or \connect <database_name>: Connect to a database.
  • \dt: List tables.
  • \dv: List views.
  • \df: List functions.
  • \du: List users (roles).
  • \h: Help for SQL commands.
  • \?: Help for psql meta-commands.
  • \e: Open query in editor.
  • \s: Display query history.
  • \timing: Toggle query timing.

Data Manipulation and Queries:

  • Don’t forget a semi-colon (;) at the end of the line/command

  • SELECT * FROM <table_name>;: Retrieve all data.
  • SELECT column1, column2 FROM <table_name> WHERE condition;: Retrieve specific columns.
  • INSERT INTO <table_name> (column1, column2) VALUES (value1, value2);: Insert data.
  • UPDATE <table_name> SET column1 = value1 WHERE condition;: Update data.
  • DELETE FROM <table_name> WHERE condition;: Delete data.
  • CREATE TABLE <table_name> (column1 type, column2 type, ...);: Create a table.
  • DROP TABLE <table_name>;: Delete a table.
  • ALTER TABLE <table_name> ADD COLUMN column3 type;: Add a column.
  • ALTER TABLE <table_name> DROP COLUMN column3;: Delete a column.
  • CREATE INDEX index_name ON table_name (column_name);: Create an index.
  • EXPLAIN SELECT * FROM table_name WHERE condition;: Show query plan.
  • BEGIN; ... COMMIT;: Start and commit a transaction.
  • BEGIN; ... ROLLBACK;: Start and rollback a transaction.

User and Role Management:

  • CREATE USER <username> WITH PASSWORD '<password>';: Create a user.
  • ALTER USER <username> WITH PASSWORD '<new_password>';: Change user password.
  • DROP USER <username>;: Delete a user.
  • GRANT <privilege> ON <object> TO <user/role>;: Grant privileges.
  • REVOKE <privilege> ON <object> FROM <user/role>;: Revoke privileges.
  • CREATE ROLE <role_name> WITH LOGIN;: create a role that can login.
  • GRANT <role_name> to <user_name>;: Grants a role to a user.

Database Management:

  • CREATE DATABASE <database_name>;: Create a database.
  • DROP DATABASE <database_name>;: Delete a database.
  • \set AUTOCOMMIT off: Disable autocommit.
  • \set AUTOCOMMIT on: Enable autocommit.

React

Rebuild frontend changes

cd ~/projects/bread/frontend

npm run build

Remove old files (optional, but recommended)

sudo rm -rf /var/www/recipes/html/*

Copy new files

sudo cp -r dist/* /var/www/recipes/html/

Set permissions

sudo chown -R www-data:www-data /var/www/recipes/html/

sudo chmod -R 755 /var/www/recipes/html/

Restarting Server

Kill any existing uvicorn processes

pkill -f uvicorn

Start uvicorn with explicit host binding

cd ~/projects/bread/backend

source venv/bin/activate

nohup uvicorn app.main:app --host 127.0.0.1 --port 8000 --forwarded-allow-ips='*' &

Visit Emlekezik.com