Demos

Note

This page is a work in progress, and will be updated with more demos soon.

Snake with curses

Generate a snake game that runs in the terminal using curses, and then modify it to add color.

Steps

1. Create a snake game with curses to snake.py

2. Running fails, ask gptme to fix a bug

3. Game runs

4. Ask gptme to add color

5. Minor struggles

6. Finished game with green snake and red apple pie!

Mandelbrot with curses

Generate a program that renders mandelbrot with curses, and then modify it to add color.

Steps

1. Render mandelbrot with curses to mandelbrot_curses.py

2. Program runs

3. Add color

Fibonacci

An old demo showing off basic code execution and shell interaction.

Steps

1. Create a new dir ‘gptme-test-fib’ and git init

2. Write a fib function to fib.py, commit

3. Create a public repo and push to GitHub

Answer question from URL

Showing off basic URL loading from the prompt, and answering questions based on the content.

Steps

1. Ask who the CEO of Superuser Labs is, passing website URL

2. gptme browses the website, and answers correctly

Edit history with /edit

The /edit command allows you to directly edit the conversation history in your text editor. This is useful for:

• Fixing typos or mistakes in previous prompts

• Removing unwanted messages

• Restructuring conversation flow

• Correcting errors before they cascade

How it works:

1. The conversation is converted to TOML format

2. Your default editor ($EDITOR) opens the TOML file

3. Edit the conversation as needed (add, remove, or modify messages)

4. Save and close the editor

5. gptme validates and applies your changes

6. If there are parsing errors, you’ll get a chance to fix them

Example use cases:

Fixing a typo in a prompt:

If you made a typo that confused the assistant, use /edit to correct it. The assistant will see the corrected version.

Removing a mistake:

If the assistant misunderstood and went down the wrong path, use /edit to remove the problematic messages and restart from a good point.

Restructuring conversation:

You can reorder messages, combine prompts, or split long conversations into cleaner structure.

Tips:

• The TOML format is human-readable and easy to edit

• Each message has a role (user/assistant/system) and content

• Be careful with TOML syntax - gptme will validate before applying

• Use /undo instead if you just want to undo the last message

• Press Ctrl+C in the editor to cancel without making changes