From a33f95f024b77128334b18da1442ca0300536526 Mon Sep 17 00:00:00 2001 From: Paillat Date: Fri, 20 Jan 2023 13:30:45 +0100 Subject: [PATCH] Fixed sme things and added dalle option --- README.md | 12 ++++--- code/imagesGeneration.py | 19 ++++++++--- code/main.py | 70 ++++++++++++++++++++-------------------- 3 files changed, 57 insertions(+), 44 deletions(-) diff --git a/README.md b/README.md index db49a8f..a00c04e 100644 --- a/README.md +++ b/README.md @@ -12,13 +12,13 @@ - We send the pdf and html files to the user # How to install -**IMPORTANT** Linux installation isn't documented yet, if anyone with a nvidia gpu wants to complete the steps for linux, feel free to pull request. +**IMPORTANT** Linux and MacOS installation isn't documented yet, if anyone wanths to complete it, feel free to do a pull request. ## Requirements - Python 3.8 - Pip - A Discord bot token - An openai api key -- (Optional) An Nvidia GPU (for image generation) +- (Optional) An Nvidia GPU (for local image generation) ## Installation - Clone the repository @@ -42,7 +42,8 @@ scoop install marp use_images = False ``` -### Image generation (optional) +## Image generation (optional) +### With Stable Diffusion UI (powerful gpu option) - Install [Stable Diffusion UI](https://github.com/cmdr2/stable-diffusion-ui) and switch to the `beta` branch. - Copy the `./image_gen_api/main.py` file to the `stable-diffusion-ui` folder - Open the file called `Dev Console.cmd` in the `stable-diffusion-ui` folder and run the following commands: @@ -50,6 +51,9 @@ use_images = False pip install uvicorn pip install fastapi ``` +- In the file `main.py`, at the first line, enable or disable the `sd` image generation. +### With DALL·E 2 (costs dalle credits) +- In the file `main.py`, at the first line, enable or disable the dalle image generation. # Running - Run the `main.py` file with : @@ -57,7 +61,7 @@ pip install fastapi python main.py ``` -### Image generation (optional) +### Local image generation (optional, only if you use the local image generation option) - Open the file called `Dev Console.cmd` in the `stable-diffusion-ui` folder and run the following commands: ``` uvicorn main:app --reload diff --git a/code/imagesGeneration.py b/code/imagesGeneration.py index 53280ca..3dbd586 100644 --- a/code/imagesGeneration.py +++ b/code/imagesGeneration.py @@ -1,7 +1,16 @@ import requests import os -async def generate(prompt,path): - r = requests.get(f"http://localhost:8000/generate_image?prompt={prompt}&path={path}") - return r.json() -#print the current working directory -print(os.getcwd()) \ No newline at end of file +import openai +async def generate(prompt, path, mode, openai_key): + #r = requests.get(f"http://localhost:8000/generate_image?prompt={prompt}&path={path}") + if mode == "sd": + r = requests.get(f"https://localhost:8000/generate_image?prompt={prompt}&path={path}") + return "image generated" + if mode == "dalle": + openai.api_key = openai_key + img = await openai.Image.acreate( + prompt=prompt, + n=1, + size="1024x1024", + ) + return img \ No newline at end of file diff --git a/code/main.py b/code/main.py index 79dfb2e..510a0a3 100644 --- a/code/main.py +++ b/code/main.py @@ -1,4 +1,7 @@ -use_images = True +#Uncomment you prefered images generation method +#use_images = "dalle" # generate images remotely with dalle 2 in your openai account +use_images = "sd" # generate images locally with stable diffusion sdkit ui (instructions in the readme) +#use_images = "No" # no images import openai # from openai import api_key import discord @@ -10,10 +13,12 @@ import asyncio import logging import datetime import base64 -if use_images: import imagesGeneration +import requests +from dotenv import load_dotenv +if use_images != "No": import imagesGeneration logging.basicConfig(level=logging.INFO) imageint = "" -if use_images: imageint = "To add an image illustration , use ![bg left:50% 70%](a-description-of-the-image.png) at the beginning of the slide, just after "---"-It's not possible to add technical images but only illustrations. The images are generated by an ai, the name of the file should be a detailed description of the image wanted. For example \" ![bg left:50% 100%](a-man-wearing-a hat-ryding-a-bicicle.png)\" but don't need to show a person necessairly." +if use_images != "No": imageint = "To add an image illustration , use ![bg left:50% 70%](a-description-of-the-image.png) at the beginning of the slide, just after \"---\". Use only .png. It's not possible to add technical images but only illustrations. The images are generated by an ai, the name of the file should be a detailed description of the image wanted. For example \" ![bg left:50% 100%](a-man-wearing-a hat-ryding-a-bicicle.png)\" but don't need to show a person necessairly." intstructions = f'''Here is a presentation with marp. It's not possible to make slides longer than 200 characters. to separate slides, " @@ -43,24 +48,30 @@ async def present(ctx: discord.ApplicationContext, subject: str, style: str = "d await ctx.defer() date = datetime.datetime.now() date = date.strftime("%Y-%m-%d-%H-%M-%S") - prompt = f"{intstructions} {indications} The subject of the presentation is: {subject} The Language is: {language} <|endofprompt|> \n" - #replace the spaces in the szbject with dashes - + marp = f'''--- +marp: true +theme: {styles[styles.index(style)]} +class: + - lead +''' + if style in darkstyles: marp = marp + f" - invert\n---" + else: marp = marp + "\n---" + prompt = f"{intstructions} {indications} The subject of the presentation is: {subject} The Language is: {language} <|endofprompt|> \n {marp}" subject2 = subject subject = subject.replace(" ", "-") #we save teh subject in base64 in a variable b64 = base64.urlsafe_b64encode(subject.encode("utf-8")) #if dosen't exist, create a directory called "userid" where the userid is the id of the user who called the command uid = str(ctx.author.id) - if not os.path.exists("data/"+uid): - os.mkdir("data/"+uid) + if not os.path.exists("./data/"+uid): + os.mkdir("./data/"+uid) datenow = datetime.datetime.now() datenow = datenow.strftime("%Y-%m-%d-%H-%M-%S") - os.mkdir(f"data/{uid}/{b64}{datenow}") + os.mkdir(f"./data/{uid}/{b64}{datenow}") response = await openai.Completion.acreate( engine="text-davinci-003", prompt=prompt, - temperature=0.7, + temperature=0.6, max_tokens=1024, top_p=1, frequency_penalty=0, @@ -69,19 +80,6 @@ async def present(ctx: discord.ApplicationContext, subject: str, style: str = "d ) #we save the output in a variable output = response["choices"][0]["text"] - #if the output dosent start with --- or with \n--- or with \n\n--- we add it at the beginning of the output - - #we add the marp header - marp = f'''--- -marp: true -theme: {styles[styles.index(style)]} -class: - - lead -''' - if style in darkstyles: - marp = marp + f" - invert\n" -# if not output.startswith("---") and not output.startswith("\n---") and not output.startswith("\n\n---"): -# output = "---\n" + output present = marp + output ##we save the output in a file called "subject.md" matches = re.finditer(r'!\[.*?\]\((.*?)\)', present) @@ -89,23 +87,25 @@ class: for match in matches: image_filenames.append(match.group(1)) #we create a text file with the image names and a md file for the presentation with utf8 encoding - with open(f"./data/{uid}/{b64}{datenow}/{subject}.md", "w", encoding="utf8") as f: - f.write(present) with open(f"./data/{uid}/{b64}{datenow}/{subject}-images.txt", "w", encoding="utf8") as f: for image in image_filenames: f.write(image + "\n") #now we generate the images, if there are any - if len(image_filenames) > 0 and use_images: + if len(image_filenames) > 0 and use_images!="no": #now we first remove the extension from the image filenames by removing the last 4 characters image_filenames = [image[:-4] for image in image_filenames] print(image_filenames) for images in image_filenames: #we download the image - await imagesGeneration.generate(images, f"{os.getcwd()}\\data\\{uid}\\{b64}{datenow}\\") - #now we rename the image to remove the _0 from the end of the filename - os.rename(f"{os.getcwd()}\\data\\{uid}\\{b64}{datenow}\\{images}_0.png", f"{os.getcwd()}\\data\\{uid}\\{b64}{datenow}\\{images}.png") - #now we whait 10 seconds for discord to resume the websocket connection - # await asyncio.sleep(10) + print ("generating image" + images) + r = await imagesGeneration.generate(images, f"{os.getcwd()}\\data\\{uid}\\{b64}{datenow}\\", use_images, apikey) + if use_images == "sd": os.rename(f"{os.getcwd()}\\.\\data\\{uid}\\{b64}{datenow}\\{images}_0.png", f"{os.getcwd()}\\data\\{uid}\\{b64}{datenow}\\{images}.png") + if use_images == "dalle": + image_url = r['data'][0]['url'] + img_data = requests.get(image_url).content + with open(f'./data/{uid}/{b64}{datenow}/{images}.png', 'wb') as handler: + handler.write(img_data) + with open(f"./data/{uid}/{b64}{datenow}/{subject}.md", "w", encoding="utf8") as f: f.write(present) #we execute the command to convert the markdown file to a pdf and html file and also generate the first slide image cmd = f"marp --pdf --allow-local-files ./data/{uid}/{b64}{datenow}/{subject}.md" os.system(cmd) @@ -162,9 +162,9 @@ async def on_ready(): if not os.path.exists("data"): os.mkdir("data") #get the openai key drom he key.env file -with open("key.env", "r") as f: - apikey = f.read() +load_dotenv() +token = os.getenv("TOKEN") +apikey = os.getenv("OPENAI") openai.api_key = apikey -with open("token.env", "r") as f: - token = f.read() +print(token) bot.run(token) \ No newline at end of file