ChatGPT

Thought you all might like to see what's going on at https://chat.openai.com/chat

I put this mess of a script in to it:

Code: Select all

from PIL import Image
import time
from PIL import ImageFont
from PIL import ImageDraw
import matplotlib.pyplot as plt
import numpy as np
import datetime

time.sleep(0.5)
original = Image.open(f"{indigo.server.getInstallFolderPath()}/Web Assets/public/1.jpg")
#time.sleep(0.8)
draw = ImageDraw.Draw(original)
font = ImageFont.truetype("Arial", 55)
draw.text((55, 23), datetime.datetime.now().strftime("%a %d %I:%M %p text"), (255, 70, 90), font=font)
plt.subplot(1, 2, 1)
plt.title("white text")
# plt.imshow(original)
original.save(f"{indigo.server.getInstallFolderPath()}/Web Assets/public/1.jpg", optimize=True, quality=45)
original.save(f"{indigo.server.getInstallFolderPath()}/Web Assets/public/2.jpg", optimize=True, quality=45)

It gave me back

Code: Select all

from datetime import datetime
from pathlib import Path
from PIL import Image, ImageDraw, ImageFont
import matplotlib.pyplot as plt
import numpy as np
import time

image_path = Path(indigo.server.getInstallFolderPath()) / "Web Assets" / "public"

time.sleep(0.5)

with Image.open(image_path / "1.jpg") as original:
    draw = ImageDraw.Draw(original)
    font = ImageFont.truetype("Arial", 55)
    draw.text((55, 23), datetime.now().strftime("%a %d %I:%M %p text"), (255, 70, 90), font=font)
    plt.subplot(1, 2, 1)
    plt.title("white text")
    # plt.imshow(original)
    original.save(image_path / "1.jpg", optimize=True, quality=45)
    original.save(image_path / "2.jpg", optimize=True, quality=45)

I'm super impressed and imagine this will enable me to wildly expand python scripting in my setup.

It's not so great at AppleScript though. This folder action gets and error:

Code: Select all

on adding folder items to this_folder after receiving added_items
    tell application "Photos"
        set display_folder to folder "display"
        repeat with i in items of display_folder
            delete i
        end repeat
    end tell

    repeat with i in added_items
        set file_name to the name of i
        if file_name ends with ".jpg" or file_name ends with ".png" or file_name ends with ".gif" then
            tell application "Photos"
                import i to display_folder. (error is on this line)
            end tell
        end if
    end repeat
end adding folder items to

Right!? Your code chops are substantially above mine, but this is a transformative tool!

I tried to ask it some questions about Indigo but it didn't appear to have any training. But it wrote me a quick script to analyze whether a picture of my alarm panel was armed or disabled.

Code: Select all


# Import the necessary modules
import cv2
import numpy as np

# Read the image file
img = cv2.imread("/Users/ryanbuckner/Downloads/redexample.png")

# Convert the image to the HSV color space
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)

# Define the lower and upper bounds for the red and green colors
lower_red = np.array([0,100,100])
upper_red = np.array([10,255,255])
lower_green = np.array([50,100,100])
upper_green = np.array([70,255,255])

# Create masks for the red and green colors
mask_red = cv2.inRange(hsv, lower_red, upper_red)
mask_green = cv2.inRange(hsv, lower_green, upper_green)

# Check if there are any red or green pixels in the image
if np.any(mask_red):
    print("Alarm is Set")
elif np.any(mask_green):
    print("Alarm is not Set")
else:
    print("Neither red nor green light is present in the picture")



I tried to ask it some questions about Indigo but it didn't appear to have any training. But it wrote me a quick script to analyze whether a picture of my alarm panel was armed or disabled.

Code: Select all


# Import the necessary modules
import cv2
import numpy as np

# Read the image file
img = cv2.imread("/Users/ryanbuckner/Downloads/redexample.png")

# Convert the image to the HSV color space
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)

# Define the lower and upper bounds for the red and green colors
lower_red = np.array([0,100,100])
upper_red = np.array([10,255,255])
lower_green = np.array([50,100,100])
upper_green = np.array([70,255,255])

# Create masks for the red and green colors
mask_red = cv2.inRange(hsv, lower_red, upper_red)
mask_green = cv2.inRange(hsv, lower_green, upper_green)

# Check if there are any red or green pixels in the image
if np.any(mask_red):
    print("Alarm is Set")
elif np.any(mask_green):
    print("Alarm is not Set")
else:
    print("Neither red nor green light is present in the picture")



Awesome quick and dirty implementation

Check it out.

living_room.png (82.86 KiB) Viewed 1720 times

Seems like in the last day or so chatgpt digested the indigo scripting tutorial. I worked out a script today and it knew how to trigger an action group with python without me explaining, and it definitely didn’t know that when I posted my first script.

[jay (support)]

Check it out.

living_room.png

LOL - too bad that's wrong, the command is:

Code: Select all

indigo.device.turnOn('living_room_light')

And we don't recommend using the name to get the device since the name can change (but that would be harder for GPT to get right).

ha yeah I saw that it was wrong. But a few days ago it wouldn't even answer

Heh, the word is definitely out.

ChatGPT is at capacity right now

Yeah pretty much the only time I can get an answer to anything with significant input from me is late in the evenings, or on the weekends.

I tried inputting an entire "style.css" file for Hubzilla with instructions to "make it more modern" and it started, looked promising... and never finished then errored out.

Just going on about this a bit:

Here's something I've wanted to do for years, knew was possible, but could not put it all together without pulling hair out over syntax, process, etc.

I have my Nanoleaf lights react to music, and I have them changing the music light effect with each song change in Apple Music. The only way I could figure out how to do this in Indigo previously was to have a set order of triggers that looked at the current Nanoleaf scene value and changed it to another set one when triggered by a change of song. That meant a different action group and trigger with different conditions for each scene I wanted to do.

In about 10 minutes with ChatGPT, I changed all that to one trigger and one action group that runs this script:

Code: Select all

import random

# Create an array of the IDs of the action groups that you want to include in the subset (and these ids have been edited for security)
subset = [
  97870712,
  13836112,
  19560082,
  3870625,
  26797751,
  15247221,
  18884308,
  19247221,
  8937550,
  1208860,
  2653673,
  737059,
  18789262
]

# Select a random action group from the subset
selected_group = random.choice(subset)

# Trigger the selected action group
indigo.actionGroup.execute(selected_group)

It works faster, more cleanly, and is easier to modify than the previous kludge. Super basic python I know, but pulling together all the documentation to format it all correctly would have been frustrating. This script ran correctly the very first time!

[jay (support)]

ha yeah I saw that it was wrong. But a few days ago it wouldn't even answer

So the question really is, which is worse: no answer at all or an incorrect one?

Good point. For a learning system, I'd prefer some answer. Then we can fix it