palpen / dalleepaperframe Goto Github PK

Currently, prompts_config.yml is hardcoded in the client.py module. Making it an input instead will allow for different prompt configuration across different displays (e.g. some display can have a children illustration theme whereas others could be Van Gogh oil paintings)

I'd like to give clients outside of my network ability to make a requests to the my local server that's hosting DALL-E.

The client checks the saved_images folder for text prompts for which images have already been generated. If there happens to be a Tweet prompt for which no images has been generated by the client and the server also happens to be off, the client will keep sending a request to the server for a new image and fail. Eventually, the counter for the number of screen refreshes reaches its max limit and the client's program is shut off.

This is a bug. We want the client to load images from the saved_images folder when the server is off (if there are no images, then the program should exit) even if there are new Tweet prompts.

Solution:

• Only send requests for Tweet prompts to the Twitter API when the server is on, otherwise load image from saved_images

Implementation:

• Check status of API endpoint for the server (in check_recent_tweets_and_generate_image_if_new?) before making a request to Twitter API

At present, each client is checking Twitter for a text prompt using the same hashtag. A new tweet with the trigger hashtag (currently, #dalle) will update the image in all client displays. This issue will make the changes that will assign a unique trigger hashtag to each client so that they can be controlled from Twitter independently (but also includes the option to change them all simultaneously).

Implementation:

• Move the trigger hashtag from tweets_utils.py to the config yaml files
• Add hashtag that will change all displays (probably just keep the default, #dalle)
• Update retrieve_tweets_containing_text_prompt function in tweets_utils.py to include an argument for each trigger hashtag

prompts_config.ymllives [here](https://github.com/palpen/DALLEePaperFrame/blob/main/client/prompts_config.yml.

For the official prompt engineering guide, see the DALL-E prompt book here

Code for how to do this

import logging
from logging.handlers import RotatingFileHandler

LOG_LEVEL = logging.INFO

logging.basicConfig(
    handlers=[RotatingFileHandler('client.log', maxBytes=100000, backupCount=10)],
    level=LOG_LEVEL,
    format="[%(asctime)s] %(levelname)s [%(filename)s.%(module)s.%(funcName)s:%(lineno)d] %(message)s",
    datefmt='%Y-%m-%dT%H:%M:%S'
)
logger = logging.getLogger('client_logger')


# Testing
def foo():
    logger.info("Log in function")
    return None

for i in range(1000):
    foo()
    logger.info(f"This is an info {i}")
    logger.debug('Hello, world!')

Relevant files:

• client.py
• tweets_utils.py