The Birth of Our Team
# AI MasterDream copilot: AMD Enhanced Artistic Rendering System
1. Introduction
2. System Requirements
3. Installation Guide
4. User Guide
5. Troubleshooting

Team Ai Master Dream copilot: 林望儒, 劉育庭

Published October 12, 2024

AI MasterDream copilot: AMD Enhanced Artistic Rendering Sys

Develop an artistic drawing optimization system based on generative AI called AI MasterDream copilot.

106

AI MasterDream copilot: AMD Enhanced Artistic Rendering Sys

Things used in this project

Hardware components

Arduino Nano 33 BLE Sense

Espressif ESP32

ESP32 and ESP32 expansion board (3.2 TFT SPI 240X320 V1.0)

TFT Touchscreen, 320x240

(SPI, ILI9341)

AMD Radeon Pro W7900 GPU

This professional GPU has 48GB of VRAM allowing for complex content creation while also supporting large LLMs such as Llama 3 locally for fast and private LLM usage. The Radeon PRO indeed caters to professionals, creators, and artists.

M5Stack ESP32 Camera Module Development Board

esp32-CAM

workstation(PC)

Software apps and online services

Arduino IDE

Install library:adafruit/Adafruit_ILI9341

python

AMD ROCm™ Software

https://rocm.docs.amd.com/

ZLUDA

LM Studio-rocm

*NOTE:remember to download the rocm version.

Hand tools and fabrication machines

Laser cutter (generic)

3D Printer (generic)

Story

https://github.com/waldolin/ai-masterdream-copilot

The Birth of Our Team

My name is Lin Wang-Ru, a young professional with a long-standing interest in AI illustration, 3D printing, design, and the Internet of Things. I've worked in various fields, including the printing industry, a musical instrument store, and government agencies. Currently, I'm continually learning industrial and IoT-related practices at a makerspace.

One day, during a networking event, I met Liu Yu-Ting, a passionate artist and designer. She has a keen interest in digital illustration and interface design, as well as a growing fascination with artificial intelligence (AI) and AI-generated art. We both had the chance to participate in the Google Cloud Skills Boost course, where we learned about AI and its potential in creative fields. We were particularly excited about the rapid advancements in AI drawing technology.

However, as we delved deeper into our studies, we encountered a significant challenge: the existing learning resources and tools were not friendly for AMD graphics card users like us. This was frustrating, and through discussion, we realized that we shared the same concern. We agreed that this was a pressing issue that needed a solution.

Based on this shared vision, we decided to develop a solution together and formed a team to participate in the AMD Pervasive AI competition. Leveraging my experience in electromechanical integration and 3D printing and her expertise in design, we aimed to create a generative AI-based drawing optimization system called AI MasterDream Copilot, specifically optimized for AMD GPUs.

Our goal with the "AI MasterDream Copilot" system is to provide a more natural, detailed, and efficient AI drawing experience, tailored specifically for AMD GPU users. AI MasterDream copilot: An Enhanced Artistic Rendering System is a system included features such as intelligent optimization, style transfer, and automatic image enhancement, addressing the current challenges AMD GPU users face in AI art creation.

We began our collaboration by combining her expertise in design and user interface with my technical background in system optimization and implementation. We studied various technical papers, read numerous books, and sought help in multiple forums. During this time, we not only enhanced our technical skills but also deepened our understanding and trust in each other as a team. By integrating the Radeon PRO W7900 GPU, we achieved efficient computational performance and used open-source platforms like PyTorch for model training and testing. It was only through real-world hardware testing that we realized the significant differences in hardware capabilities; the Radeon PRO indeed caters to professionals, creators, and artists.

Participating in this competition is not just a personal challenge but also an opportunity to showcase our journey from AI novices to creators using AI. Throughout the development process, we faced many technical challenges, particularly in tuning and optimizing the system. With a solid concept in place, we constantly referred to online resources, modified our code, and interacted with other developers in the community.

Finally, we shaped our system. To further test and refine it, we decided to continue improving it, hoping to provide a better creative tool for more creators.

The AI MasterDream Copilot not only successfully addressed the issues AMD GPU users face in AI drawing but also provided users with greater creative freedom and higher drawing efficiency. As the AI era dawns, we are both excited and confident. The AI MasterDream Copilot is not just an AI assistant tool for artistic creation; it's a platform encouraging cross-disciplinary collaboration and innovation. Through this collaboration, we have deepened our understanding of AI technology and expanded our horizons in artistic creation. We believe our efforts will bring real change for AMD users.

This is the beginning of our story and a testament to our commitment to innovation. Our team is working hard to provide every artist and creator with more possibilities, creating a better experience for those using AMD graphics cards. We hope our solution will gain significant attention and positive feedback in the competition.

# AI MasterDream copilot: AMD Enhanced Artistic Rendering System

Abstract

This project uses AMD graphics cards and TensorFlow technology to achieve efficient image generation, reasoning and text generation functions. The system uses Node-RED workflows for low-code control and runs edge computing on an Arduino development board. By combining AI tools, Angular, Arduino and andriod APP, the conversational user interface design of multiple projects has been realized.

1. Introduction

1.1 Document Introdutcion

This project successfully implemented image generation, image reasoning and text generation tasks on AMD graphics cards. The system leverages TensorFlow for accelerated inference and Node-RED and workflows for low-code control and process management. In addition, for edge computing on the Arduino development board, TensorFlow and TensorFlow Lite for Microcontrollers are used to perform edge computing and integrate related APIs.

In terms of user interface design, the system combines the integration of Angular and Arduino APP with Node-RED workflows to realize the dialogue design and management of multiple projects. The project was presented and evaluated in the AMD Hackathon competition at https://www.hackster.io/.

1.2 Target Audience

This documentation is designed for users configuring their systems using AI MasterDream copilot and will benefit both beginners and experts. For beginners, this document provides detailed operating instructions and basic knowledge to help them get started quickly. For experts, the documentation covers more in-depth technical details and best practices to support efficient configuration and optimization of the system. Whether you are just starting out or have extensive experience in AI systems, this guide will help you understand how to configure and effectively use the AI MasterDream Copilot to enhance your experiences.

1.3 Conditions of use

stable diffusion: AGPL3.0

Comfyui: GPL3.0

Lm studio: MIT

AI MasterDream copilot: AFL3.0 and others

The license is based on AMD Pervasive AI Developer Contest and mentioned in license.

Models is based on the license, and add a license with age above 18.

For example, Gemma: AGP2.0

[llAMA3's license](https://github.com/meta-llama/llama-models/blob/main/models/llama3/LICENSE)

1.4

# Demonstration

ComfyUI is not as easy to use compared to SD WebUI, and you need a deeper understanding of the principles behind the graphics to use ComfyUI well. There are often problems in learning. How do I put the model in and how to use it? I have to change the code and integrate it myself.

Oftenly there is a lack of nodes. The speed is only a little faster than some fat WebUIs. When I ask questions on github, the authors will only say that you have not installed custom nodes.

You need to install them via ComfyUI-Manager and closed. A lot of people can't find the download of custom nodes. What's the node? I have to try to learn and find a solution every times. So You will like the AI MasterDream copilot.

Now you can look for the process Jason file

Then use ai to convert it to node-red for use.

and run using flow control.

## Video demonstration link:

https://github.com/waldolin/ai-masterdream-copilot

You just use the graphical selection interface, and have fun!!!

2. System Requirements

2.1 Hardware Requirements

A PC base on AMD grapic card
A phone on android
An additional arduino devices extend computating for edge AI.
An additional arduino device expanded for display.

# Run resource allocation

This project focuses on VRAM resource allocation of AMD graphics cards, using RX 6950 XT and W7900 PRO as examples. Taking the RX 6950 XT as an example, this graphics card has 16GB of VRAM resources. The following is the resource allocation of different models:

LM Studio: Loading a 2B model requires ~1GB VRAM.

Stable Diffusion (SD): Approximately 7GB VRAM is required to load the model.

ComfyUI: ~7GB VRAM required to load model.

Node-RED mainly uses RAM resources. Due to the small memory space of embedded devices, in order to ensure operating efficiency and resource allocation, the project uses TensorFlow Lite optimized model operation, which may cause a certain delay. If you have any professional suggestions or further optimization measures, please let us know.

The project also took a series of measures to adjust and optimize resource allocation, including fine-tuning, memory reclamation, system restarts, refactoring, changes, simplification and optimization. Users can choose to run the model in FP16 or FP32 mode, or choose a 2B model in Q3 or Q2 settings. If VRAM resources are sufficient, these settings have little impact; if VRAM resources are insufficient, it is recommended to consider using some or all cloud services. LM Studio can also load 2B models into RAM, but this will significantly slow down execution. You can also consider using ComfyUI or the nodes provided by this project as an alternative.

This is a process of exploring how to optimize graphics card hardware resources. While upgrading from the RX 6400 to the RX 6950 XT improves performance, it may not fully meet all needs. This is a research and exploration of spending time and hardware resources, and W7900 PRO can fully meet all needs.

# inference Memory Requirements

For inference, the memory requirements depend on the model siez and the precision of the weights.

Here is a simple table showing the qpproximate memory needed for different configurations:

8B --> 16G(FP16) 8G(FPS) 4G(INT4)

70B--> 140G(FP16) 70G(FPS) 35G(INT4)

2.2 Software Requirements

# prerequisites

Windows 10.11 above +HIP or Linux ROCm 5.7+ or ROCm 6.0

3. Installation Guide

3.1 Download and Installation

# Necessary installation

[pyhton] >=3.10.6

[thonny python]

[git]

CMake

[Rust（1.66.1 or later）]

C++ compiler

[The latest AMD Radeon software Adrenalin]

(Linux only) ROCm 5.7+ or ROCm 6.0

(Windows only) ROCm 5.7+ and ROCm 6.0[HIP SDK]

Related documents:

https://rocm.docs.amd.com/projects/HIP/en/develop/index.html

https://rocm.docs.amd.com/en/latest/

Install[ZLUDA]

Install[nodejs]

Open CMD check nodejs version

node -version

Install[node-red]

Open CMD

npm install -g --unsafe-perm node-red
npm -g install
@angular/cli@18.0.1
@vue/cli@5.0.8
create-react-app@5.0.1
express@4.19.2
http-server@14.1.1
node-sd-webui@0.0.8
npm@10.8.1
vercel@29.3.3
replicate@0.31.1

Please make sure you have installed the following software:

install[stable diffusion]

https://github.com/lshqqytiger/stable-diffusion-webui-amdgpu

https://github.com/lshqqytiger/stable-diffusion-webui-amdgpu-forge

and related plug-ins

Or use google colab

https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/stable_diffusion.ipynb

(optional)

for example,
GIT CLONE https://github.com/lshqqytiger/stable-diffusion-webui-amdgpu
pip install -r requirements.txt

And

set COMMANDLINE_ARGS=--api

(necessary)

Related documents:

https://forum.gamer.com.tw/C.php?bsn=60030&snA=635965

Install[comfyUI]

https://github.com/patientx/ComfyUI-Zluda

https://github.com/comfyanonymous/ComfyUI

And related plug-ins (which can provide subtle control over the drawing process)

https://github.com/HomunMage/CrewAI-GUI

(optional)

Related documents:

https://comfyanonymous.github.io/ComfyUI_tutorial_vn/

https://hackmd.io/@Sai/comfyUI?print-pdf#/

install[LM Studio-rocm]

(https://lmstudio.ai/rocm)

(run locally)

*NOTE:remember to download the rocm version.

Or ollama

Or LangChain

Or ollama

Or llama-cpp

Or vllm

Or GOOGLE AI API or OPENAI API (requires key and fee)

Install[tensorflow]

pip install tensorflow==2.14.0
pip install tf-nightly

ROCm version //TensorFlow version

6.1.x //2.13.1、2.14.0、2.15.0

6.0.x //2.12.1、2.13.1、2.14.0

Install tensorflow lite in arduino (the guides)

pip install tflite-runtime

Use Arduino Nano 33 BLE Sense

git clone https://github.com/tensorflow/tflite-micro-arduino-examples Arduino_TensorFlowLite
cd Arduino_TensorFlowLite
git pull

Optional installation (recommended installation)

[uTensor](https://github.com/uTensor/uTensor)

[Pytorch](https://github.com/pytorch/pytorch.git) It is recommended to use Linux for training.

pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm5.7

Related documents:

https://rocm.docs.amd.com/projects/install-on-linux/en/develop/how-to/3rd-party/pytorch-install.html

[DirectML](https://github.com/microsoft/DirectML)

(Linux only)[Scale](https://github.com/spectral-compute/scale-docs?tab=readme-ov-file)

[MicroPython](https://micropython.org/)

Related documents:

https://docs.arduino.cc/micropython/

https://docs.arduino.cc/learn/programming/arduino-and-python

https://docs.arduino.cc/micropython/basics/code-editors

[Clip-interrogator](https://replicate.com/pharmapsychotic/clip-interrogator?output=json)

Document:https://replicate.com/docs

3.2.1 Installation Steps

Install the SDK

### AI MasterDream Copilot App User Guide

#### start using

1. **Launch the app:** Open the AI MasterDream Copilot app (under construction), follow the prompts to set your age and register a Google account (in-app purchase).

2. **Age Restriction:** This application is prohibited from being used by users under the age of 18. Please confirm that your age meets the requirements before use.

#### Select function

**Usage Suggestions:** You can choose to use the original Stable Diffusion (XL) or ComfyUI and other AI drawing software, and purchase this product additionally to use functions such as special graphical prompt design and AI drawing fast teleprompter.
**AI MasterDream Copilot Drawing Assistant:** Improve drawing efficiency and create a new customer experience with the AI MasterDream Copilot Drawing Assistant. We provide solutions that balance cost and service quality, and provide feedback to users who originally used various open source AI drawing software to improve their experience.
**Version Selection:** Detailed online subscription version and offline buyout version plans will be announced in the future, and different customized services and updated warranty services will be provided.

#### System Configuration

1. **Prepare API:** Before running this project, please prepare the following API:

LM Studio's Local Inference Server API (or Google AI API, OpenAI API alternative)
Stable Diffusion (XL) API
ComfyUI API

2. **Configuration File:** Fill in the details of the above API in the relevant config.json file. If you have a Google AI or Azure OpenAI API key, you can also fill it in config.json, but this is not mandatory. Please be careful to avoid leaking API keys and do not share unauthorized software programs.

3. **System Operation:** Please run the Node-RED and AI MasterDream Copilot applications/software. If there is an error during installation or update, please report and inquire to the original publisher.

#### Precautions

**Version Updates:** When Stable Diffusion or ComfyUI is running properly, please avoid updating until AI MasterDream Copilot updates the version requirements. Avoid issues with bugs or API changes that may occur after updating, but this is not mandatory.
**Problem Report:** From 2022 to the present, error problems have occurred when running with AMD graphics cards. Please seek answers and report errors. If some operational errors do not affect this function, ComfyUI can be used interchangeably with Node-RED.

We are committed to providing the best user experience. If you have any questions, please feel free to contact our support team.

3.2 Configuration

3.2.1 initial Setup

Import JSON File to use ComfyUI and Node-red's dashboard

Before you make sure, you have installed nodejs and angular

Please check with

nodejs --version

Or download from https://nodejs.org/en

And

Install angular

npm install -g @angular/cli

git clone https://github.com/waldolin/ai-playground-angular

(cmd)

cd ai-playground-angular

npm

npm install

(cmd)

ng serve

(option)import you api key in configure to use gemini

*notice: It will cost the fee.

Run the link with lm stuido

http://localhost:4200/ai/llm

http://localhost:4200/ai/gemini-pro

```

"sdxl": {
"invoke_url": "" ,
"fetch_url_format": "",
"headers": {
"Authorization": "",
"Accept": ""
}
},
"llama2": {
"invoke_url": "",
"fetch_url_format": "",
"headers": {
"Authorization": "",
"Accept": ""
}
},
"azure_openai":{
"api_key": "",
"api_base": "",
"deployment_name": "",
"api_version": ""
}

```

3.2.2 Advanced Configuration

# Cloud service selection

You can choose the following cloud services to build your system:

Google Cloud
Azure
AWS (Amazon Web Services)
AMD Cloud Services

In addition, you can also choose to build a multi-cloud service environment by yourself to meet project needs.

For example:

Additional software available for purchase

[MYAI Studio for Windows artificial intelligence development tool](https://tw.myai168.com/article/index?sn=11113)

[MYAI Studio for Linux artificial intelligence development tool](https://tw.myai168.com/article/index?sn=10982)

This project uses Terraform automated deployment to start the Google Kubernetes Engine cluster, Autoscaling TensorFlow Model Deployments with TF Serving and GKE. Use GKE and Anthos Service Mesh to implement the Canary version of TensorFlow model deployment and use TFX API, keras functional API, huggingface API.

Translate Text with the Cloud Translation API and Extract, Analyze, and Translate Text from Images with the Cloud ML APIs and APIs Explorer. Measuring and Improving image Accuracy and train and fine-tune models.

# Terraform

Terraform is a declarative infrastructure-as-code tool that automates the deployment and management of cloud infrastructure through the use of profiles. It can simplify the establishment and maintenance process of infrastructure and achieve efficient resource configuration and version control.

The files include

main.tf
provider.tf
variables.tf

Set your project ID and other variables in variables.tf.

This is just a basic example, you can customize it to suit your needs. For more information, see the Terraform Google Cloud Platform Provider and Google Kubernetes Engine documentation.

Instruction:

terraform fmt
terraform init //Initialize Terraform.
terraform plan //Preview the resources to be created.
terraform apply // Create a GKE cluster.

To create a bucket in your project:

First upload the model file to the bucket in the project. Because buckets are domain-wide resources in Google Cloud, you must use unique bucket names.

# Enable service

gcloud enable service:

gcloud services enable \
cloudbuild.googleapis.com \
container.googleapis.com \
cloudresourcemanager.googleapis.com \
iam.googleapis.com \
containerregistry.googleapis.com \
containeranalysis.googleapis.com \
ml.googleapis.com \
dataflow.googleapis.com

# Process steps

File upload bucket

Set up a GKE cluster

Deployment model

Create ConfigMap (configmap.yaml) and use kubectl to create ConfigMap

Create TensorFlow Serving deployment (deployment.yaml)

Public deployment

Configuring the Level Pod Autoscaler

Test models, load testing

Run sample request

You need project owner permissions in the GCP project. This project only provides sample files, you need to have cloud experience. For related questions, please refer to the document: https://cloud.google.com/docs/terraform

Or you can seek help from Google community or Google tutorials or cloud engineers.

Sample request body (request-body.json), whose format conforms to [TensorFlow Serving REST API](https://www.tensorflow.org/tfx/serving/api_rest)

Demo: Using Google Apps Script to generate presentations with AI

1. Open the Google Apps Script Editor:

Open your Google Drive.

Click New > More > Google Apps Script.

2. Copy and paste the code:

For example: sameple_code_gas.js into google app scripts

Modify the content of your JSON file in jsonData

Execute selected function

Locally generated images need to be uploaded to Google Drive:

Or manually access the image link and download the image to your computer.

Upload the image to your Google Drive.

Get the Google Drive link to the image (right-click on the image > Get Link).

Replace the imagePath value in the Google Apps Script code with the Google Drive link for the image.

Or modify the filepath and CREDENTIALS_FILE of the uploadpic.py file

and run

python uploadpic.py

Read JSON file using Google Drive API (authorization required):

A JSON file needs to be uploaded to your Google Drive.

Use the Google Drive API in GAS code to read file contents.

You need to authorize GAS Script to access your Google Drive.

#Enable APIs

gcloud services enable drive.googleapis.com

Here are the steps to obtain the credentials.json file:

1. Open Google Cloud Console:

• Visit https://console.cloud.google.com/

2. Create a service account:

• In the left navigation menu, select IAM & Administration > Service Accounts.

• Click the "Create Service Account" button.

• Enter a service account name (for example, "drive-uploader") and select Create and Continue.

• Optional: Add a role for the service account. Since you need to upload files, it is recommended to add the "Storage Manager" role.

• Click "Done".

3. Create a service account key:

• In the list of service accounts, find the service account you just created.

• Click the service account name to enter the service account details page.

• Click on the "Keys" tab.

• Click Add Key > Create New Key.

• Select the key type as "JSON" and click "Create".

• The browser will automatically download a credentials.json file.

NOTE: To securely store the credentials.json file containing your service account's private key, please keep it in a safe place and do not make it public or upload it to a public code base.

Or use vba to generate ptt files

See details in file sameple_code_vba.ini

## AWS

Amazon Bedrock is a fully managed service that provides access to FMs from third-party providers and Amazon; available via an API. With Bedrock, you can choose from a variety of models to find the one that’s best suited for your use case. You will gain hands-on experience using Bedrock APIs, SDKs, and open-source software, such as LangChain and FAISS, to implement these usage patterns.

Bedrock>model access

On the Model access screen, click on top right button "Manage model access":

On model access screen, select only following models and click on "Request model access" button:

Stability AI

SDXL 1.0

install python, the boto3 and botocore AWS SDK for Python (Boto3) libraries. Both libraries contain required dependencies related with Bedrock APIs.

pip install --no-build-isolation --force-reinstall \
"boto3>=1.28.57" \
"awscli>=1.29.57" \
"botocore>=1.31.57"

Installing Supporting libraries, check AWS Credentials to run call AWS APIs and Bedrock APIs.

An example of that code is shown below:

pip install --quiet langchain==0.0.304

# Use knowledge-base

Upload the file of pdf or words to use knowledge-base

If you are running into your own environment, ensure your user/role has these permissions.

If you have question, please check this.

Or you can seek help from the AWS community or AWS tutorials or cloud engineers.

4. User Guide

4.1 Basic Operations

# Set the running environment

For different hardware environments, in addition to cloud operation, we provide two methods of building an operating environment:

1. Deploy the environment using AMD ROCm Container on Windows or Linux: This method is suitable for situations where you need to run generative AI models efficiently on your local machine. Using AMD ROCm Container can quickly build a consistent operating environment and simplify the configuration process.

2. Use VENV to build a Python environment: This method is suitable for development in a local Python environment. Using VENV you can create an isolated Python environment to easily manage project dependency packages and versions.

In addition, this project also provides methods to set up the operating environment on the Arduino development board:

Use Thonny Python and pip to set up the environment: On the Arduino development board, you can use Thonny IDE with pip to configure the Python environment. This allows developers to prototype and test on smaller devices. Please set the environment for different hardware environments.

# Install using docker

To run it in Docker in its simplest form, just run:

docker run -it -p 1880:1880 --name mynodered nodered/node-red
npm install node-red-node-arduino
npm install node-red-contrib-gpio

Or in node management

Install in cmd

npm install
node-red-node-arduino
node-red-contrib-gpio
node-red-contrib-arduino-iot-cloud

And other details such as list files (requirements)

# Use Thonny Python IDE to write MicroPython programs

Sample files and guide

use usb to TTL to upload

use a sample to debug first

And add it to your machine of 3D printer or Ray cutting

add your esp32-CAM on it to run edge AI.

# model file

LM Studio loads the model file

You can find these settings as follows: AI Chat > New chat > Then load the model you need > Then toggle the settings sidebar > Then select advanced options. This is where you will find the GPU usage settings. You need to press the MAX button to achieve maximum acceleration.

Use this project to train and fine-tune model files

You can use the model files trained and fine-tuned for this project on LM Studio. Depending on the GPU hardware and model used, prompt instruction processing time can be reduced from 5-20 minutes to 2 minutes or even less. This will greatly speed up the generation of prompt words and achieve prompt word generation within a few seconds in local evernment!

The training materials for this project include 10, 000 Chinese prompt words.

Style label prediction accuracy: 0.96

Hashtag prediction accuracy: 0.90

Description tag prediction accuracy: 0.96

The training materials for this project include 110, 000 Chinese prompt words.

Style theme description label prediction accuracy >0.9

Precautions:

Import JSON file

Use the following steps to import a JSON file:

Make sure you have a properly formatted JSON file.

Select the "Import" option in the system and select the JSON file to import.

Test and Validate

Use the AI chat tool to test the import results to ensure the contents of the JSON file are correct.

Please pay special attention to the following points:

System prompt (System Prompt) is correct: Check that the system prompt information is displayed as expected and ensure that it accurately reflects the data in the JSON file.

NOTE: the fine-tune model is needed above 18 ages.

# The entire project workflow can be divided into the following steps:

AI prompt project

Planning the conversation

Image reasoning

Prompt generation

process control

Image module

image generation

Image recognition

API integration

5. Troubleshooting

5.1 common issues

# port

Node-red running at http://127.0.0.1:1880/

Stable diffusion running at http://127.0.0.1:7860/

ComfyUI running at http://127.0.0.1:8188/

LM Studio running at http://127.0.0.1:1234/

Angular running at http://localhost:4200/ai/llm

Local IP: http://127.0.0.1/

If you use this project to build multi-cloud services, please visit the cloud ip

Please visit http://127.0.0.1:1880/ to start your creative programming journey!

Please visit http://127.0.0.1:7860/ to start your creative journey!

Please visit http://127.0.0.1:8188/ to start your creative journey!

Please visit http://localhost:4200/ai/llm to start your drawing chat assistant journey!

Please visit http://127.0.0.1:1234/ to start your drawing chat assistant journey!

5.2 Contact Support

If you encounter any problems or want to obtain more technical details, please submit them on the author's github. For errors in my project, please submit them on [my github](https://github.com/waldolin/)!

https://huggingface.co/waldo870

https://www.kaggle.com/waldolin

Part of code will be released after AMD Pervasive AI Developer Contest.

Or feel free to contact me at any time: a0980368677@gmail.com

Original project:

# [Booru tag autocompletion for A1111 in Traditional Chinese ](https://github.com/waldolin/a1111-sd-webui-tagcomplete-TW)extended training

# [ai-playground-angular](https://github.com/waldolin/ai-playground-angular)Modification and extension

# Future work

Develop graphical structural modules

Technology expansion: Continue to pay attention to the latest research in the field of generative AI and image processing, and explore new technologies and application scenarios.

Cross-field application: Apply our technology to other fields, such as virtual reality, augmented reality and automated design, to expand the scope of technology application.

Continuous innovation: Regularly update and improve functions, and conduct technical iterations and innovations based on user needs and market trends.

6.Appendix

6.1 Glossary

6.2 References

1. Li, Jin-Hong. **All-around Use of PyTorch: Practical Edition - Deep Learning and Graph Neural Networks**. Taipei: Deep Digital, 2022. First Edition. Registration Number: 1286604, Call Number: 312.831 4083.

2. Guo, Ka, and Dai, Liang. **PyTorch Deep Learning: Hands-On Popular Projects in Computer Vision**. Taipei: Deep Digital, 2022. ISBN: 9786267146125.

3. Song, Li-Huan. **Getting Started with Python in the Simplest Way: Master Deep Learning to Become an Expert**. Taipei: Deep Digital, 2023. First Edition. Registration Number: CS39815, Call Number: 312.32P97 3004.

4. Chen, Jing-Lei. **Beyond Multi-GPU Multi-Machine: Accelerated Practice in Distributed Machine Learning**. Taipei: Deep Digital, 2021. ISBN: 9789860776546.

5. Wen, Cheng-Yao. **Self-Learning Machine Learning: Connect to the World on Kaggle and Become a Data Scientist**. Taipei: Flag, 2021. ISBN: 9789863126720.

6. Mon Kyodo Daisuke, and Li, Yan-Ting. **Kaggle Competition Top Secrets: Unveiling the Key Techniques of Grandmaster's Feature Engineering**. Taipei: Flag, 2021. ISBN: 9789863126379.

7. Lakshmanan, Valliappa, Görner, Martin, and Gillard, Ryan. **Practical Computer Vision Machine Learning: Building End-to-End Image Machine Learning**. Taipei: Qifeng Information, 2022. ISBN: 9786263242074.

8. Nishinai, Akira, Hu, Feng-Rong, and Xu, Xian-Zheng. **The Mathematical Foundation of Machine Learning that Even High School Students Can Understand: A Must-Read for AI and Deep Learning**. Taipei: Flag, 2020. ISBN: 9789863126140.

9. Zhang, De-Feng. **80 Master-Level Python Examples in Computer Vision: The Latest Applications in Computer Vision in One Book**. Taipei: Deep Digital, 2023. ISBN: 9786267273531.

10. Lin, Da-Gui. **Illustrated TensorFlow 2: Practical tf.keras + Colab Cloud, Deep Learning, Artificial Intelligence, Image Recognition - Beginner's Edition**. Taipei: Broadview Press, 2022. ISBN: 9786263331044.

Related references:

"AMD Pervasive AI Developer Contest Generative AI Study Guide."

Hackster.io. https://www.hackster.io/512354/amd-pervasive-ai-developer-contest-generative-ai-study-guide-bb5649

AutoGPTQ. GitHub repository. https://github.com/AutoGPTQ/AutoGPTQ
Unsloth. GitHub repository. https://github.com/unslothai/unsloth
VBA-JSON. GitHub repository. https://github.com/VBA-tools/VBA-JSON
LLAMA3. Meta LLAMA models. https://github.com/meta-llama/llama-models/blob/main/models/llama3/LICENSE
Python 3.10.6 or later. Python official site. https://www.python.org/downloads/
Thonny Python. Thonny official site. https://thonny.org/
Git. Git official site. https://git-scm.com/downloads
CMake. Official CMake website.
Rust 1.66.1 or later. Rust official site. https://www.rust-lang.org/tools/install
C++ Compiler.
The latest AMD Radeon Software Adrenalin. AMD official site. https://www.amd.com/zh-tw/support/download/drivers.html
ROCm 5.7+ or ROCm 6.0 (Linux only). AMD ROCm documentation. https://rocm.docs.amd.com/en/latest/
ROCm 5.7+ and ROCm 6.0 (Windows only) HIP SDK. AMD ROCm developer hub. https://www.amd.com/en/developer/resources/rocm-hub/hip-sdk.html
HIP Documentation. AMD ROCm HIP documentation. https://rocm.docs.amd.com/projects/HIP/en/develop/index.html
ZLUDA. GitHub repository. https://github.com/vosen/ZLUDA
Node.js. Node.js official site. https://nodejs.org/en
Node-RED. Node-RED official site.
Stable Diffusion WebUI for AMD GPU. GitHub repository. https://github.com/lshqqytiger/stable-diffusion-webui-amdgpu
Stable Diffusion WebUI for AMD GPU (Forge). GitHub repository. https://github.com/lshqqytiger/stable-diffusion-webui-amdgpu-forge
Stable Diffusion on Google Colab. Hugging Face GitHub repository. https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/stable_diffusion.ipynb
ComfyUI with ZLUDA. GitHub repository. https://github.com/patientx/ComfyUI-Zluda
ComfyUI. GitHub repository. https://github.com/comfyanonymous/ComfyUI
CrewAI GUI. GitHub repository. https://github.com/HomunMage/CrewAI-GUI
ComfyUI Tutorial. ComfyUI official tutorial. https://comfyanonymous.github.io/ComfyUI_tutorial_vn/
ComfyUI HackMD. HackMD documentation. https://hackmd.io/@Sai/comfyUI?print-pdf#/
LM Studio ROCm. LM Studio official site. https://lmstudio.ai/rocm
PyTorch. GitHub repository. https://github.com/pytorch/pytorch.git
PyTorch Installation Documentation for ROCm. AMD ROCm documentation. https://rocm.docs.amd.com/projects/install-on-linux/en/develop/how-to/3rd-party/pytorch-install.html
DirectML. GitHub repository. https://github.com/microsoft/DirectML
Scale (Linux only). GitHub repository. https://github.com/spectral-compute/scale-docs?tab=readme-ov-file
MicroPython. MicroPython official site. https://micropython.org/
Arduino MicroPython Documentation. Arduino official documentation. https://docs.arduino.cc/micropython/
CLIP Interrogator. Replicate official site. https://replicate.com/pharmapsychotic/clip-interrogator?output=json
Replicate Documentation. Replicate official documentation. https://replicate.com/docs
Google Cloud Terraform Documentation. Google Cloud documentation. https://cloud.google.com/docs/terraform
Google Cloud Console. Google Cloud console. https://console.cloud.google.com/
Amazon Bedrock Workshop. GitHub repository. https://github.com/aws-samples/amazon-bedrock-workshop
AI Playground with Angular. GitHub repository. https://github.com/waldolin/ai-playground-angular
Roadmap to Becoming an Artificial Intelligence Expert. GitHub repository. https://github.com/AMAI-GmbH/AI-Expert-Roadmap
Node-RED Discussion Forum. Node-RED discourse forum. https://discourse.nodered.org/t/reclaiming-memory-from-nodes/49554/23
"Drawing AI - ComfyUI Based on Node-based Image Generation Tool." Bahamut Creation Platform. https://home.gamer.com.tw/creationDetail.php?sn=5679763
"ComfyUI Installation Guide ~ Standardizing Stable Diffusion AI Image Generation Process." IvonBlog. https://ivonblog.com/posts/stable-diffusion-comfyui/
"ESP32-CAM AI-Thinker Pin Guide: GPIO Usage Instructions." Qutaojiao. https://www.qutaojiao.com/24272.html
"Nano 33 BLE Sense Rev2 Cheat Sheet." Arduino. https://docs.arduino.cc/tutorials/nano-33-ble-sense-rev2/cheat-sheet/
"tflite-micro-arduino-examples." GitHub, TensorFlow. https://github.com/tensorflow/tflite-micro-arduino-examples?tab=readme-ov-file#how-to-install
"TensorFlow Lite for Microcontrollers." GitHub, TensorFlow. https://github.com/tensorflow/tflite-micro

Special thanks to

AMD

HACKSTER

Liu Yuting

Kun-Neng Hung

Professor Gao Huantang

Professor Chen Xinjia

Professor Ke Bowen

MakerPro

Maker Base of Taipei-Keelung-Yilan-Hualien-Kinmen-Matsu Regional Branch of MOL in Taiwan (R.O.C)

NTU AI CLUB

ML Paper Reading Club

GDG Taipei and GDG Taoyuan

Schematics

Code

function createVBASlides() {
  //  JSON 
  var jsonData = [
    {
      "slideTitle": "VBA ",
      "slideContent": [
        "VBA Visual Basic for Applications",
        " Office  ExcelWord ",
        "VBA "
      ],
      "imagePath": "https://source.unsplash.com/960x640/?VBA"
    },
    {
      "slideTitle": "VBA ",
      "slideContent": [
        " Excel VBA ",
        "",
        "VBA "
      ],
      "imagePath": "https://source.unsplash.com/960x640/?VBA,Excel"
    },
    {
      "slideTitle": "VBA ",
      "slideContent": [
        "VBA  Office ",
        " VBA ",
        "VBA "
      ],
      "imagePath": "https://source.unsplash.com/960x640/?programming"
    }
  ];

  //  Google Slides
  var presentation = SlidesApp.create("VBA ");

  //  JSON 
  jsonData.forEach(function(slideData) {
    // 
    var slide = presentation.appendSlide(SlidesApp.PredefinedLayout.TITLE_AND_BODY);

    // 
    slide.getPageElements()[0].asShape().getText().setText(slideData.slideTitle);

    // 
    var content = "";
    slideData.slideContent.forEach(function(point) {
      content += "- " + point + "\n";
    });
    slide.getPageElements()[1].asShape().getText().setText(content);

    // 
    if (slideData.imagePath) {
      slide.insertImage(slideData.imagePath);
    }
  });

  //  Google Slides 
  Logger.log("" + presentation.getUrl());
}

npm install 
@arduino/node-red-contrib-arduino-iot-cloud@1.0.10
@mapbox/node-pre-gyp@1.0.11 extraneous
@tensorflow/tfjs-converter@1.2.11 extraneous
@tensorflow/tfjs-core@1.2.11 extraneous
@tensorflow/tfjs-data@1.2.11 extraneous
@tensorflow/tfjs-layers@1.2.11 extraneous
@tensorflow/tfjs-node@1.2.11 extraneous
@tensorflow/tfjs@1.2.11 extraneous
@types/node-fetch@2.6.11 extraneous
@types/offscreencanvas@2019.3.0 extraneous
@types/seedrandom@2.4.27 extraneous
@types/webgl-ext@0.0.30 extraneous
@types/webgl2@0.0.4 extraneous
abbrev@1.1.1 extraneous
adm-zip@0.4.16 extraneous
agent-base@4.3.0 extraneous
ansi-regex@2.1.1 extraneous
aproba@1.2.0 extraneous
are-we-there-yet@1.1.7 extraneous
canvas@2.11.2 extraneous
chebyshev@0.2.1 extraneous
chownr@1.1.4 extraneous
code-point-at@1.1.0 extraneous
color-support@1.1.3 extraneous
console-control-strings@1.1.0 extraneous
decompress-response@4.2.1 extraneous
deep-extend@0.6.0 extraneous
delegates@1.0.0 extraneous
detect-libc@1.0.3 extraneous
emoji-regex@8.0.0 extraneous
es6-promise@4.2.8 extraneous
es6-promisify@5.0.0 extraneous
euclidean@0.0.0 extraneous
face-api.js@0.21.0 extraneous
fs-minipass@1.2.7 extraneous
gauge@2.7.4 extraneous
has-unicode@2.0.1 extraneous
https-proxy-agent@2.2.4 extraneous
ignore-walk@3.0.4 extraneous
ini@1.3.8 extraneous
is-fullwidth-code-point@1.0.0 extraneous
manhattan@1.0.0 extraneous
mimic-response@2.1.0 extraneous
minipass@2.9.0 extraneous
minizlib@1.3.3 extraneous
mkdirp@0.5.6 extraneous
nan@2.20.0 extraneous
needle@2.9.1 extraneous
node-fetch@2.1.2 extraneous
node-pre-gyp@0.13.0 extraneous
node-red-contrib-face-recognition@2.0.4 extraneous
node-red-contrib-line-notify-api@1.0.0
node-red-contrib-moment@5.0.0
node-red-contrib-sunevents@3.1.1
node-red-contrib-web-worldmap-cn@4.8.2
node-red-dashboard@3.6.5
node-red-node-arduino@1.0.0
node-red-node-base64@0.3.0
node-red-node-email@3.0.1
node-red-node-mysql@2.0.0
node-red-node-ping@0.3.3
node-red-node-random@0.4.1
node-red-node-serialport@2.0.2
node-red-node-suncalc@1.1.2
node-red-node-ui-webcam@0.4.0
nopt@4.0.3 extraneous
npm-bundled@1.1.2 extraneous
npm-normalize-package-bin@1.0.1 extraneous
npm-packlist@1.4.8 extraneous
npmlog@4.1.2 extraneous
number-is-nan@1.0.1 extraneous
os-homedir@1.0.2 extraneous
os-tmpdir@1.0.2 extraneous
osenv@0.1.5 extraneous
progress@2.0.3 extraneous
rc@1.2.8 extraneous
rimraf@2.7.1 extraneous
sax@1.4.1 extraneous
seedrandom@2.4.4 extraneous
set-blocking@2.0.0 extraneous
signal-exit@3.0.7 extraneous
simple-concat@1.0.1 extraneous
simple-get@3.1.1 extraneous
string-width@1.0.2 extraneous
strip-ansi@3.0.1 extraneous
strip-json-comments@2.0.1 extraneous
tar@4.4.19 extraneous
tfjs-image-recognition-base@0.6.2 extraneous
tr46@0.0.3 extraneous
tslib@1.14.1 extraneous
webidl-conversions@3.0.1 extraneous
whatwg-url@5.0.0 extraneous
wide-align@1.1.5 extraneous

#non essential dependencies:

[
    {
        "id": "b1a1e0e6.6b3f8",
        "type": "tab",
        "label": "Stable Diffusion",
        "disabled": false,
        "info": ""
    },
    {
        "id": "f8e1e8d5.3ebdb8",
        "type": "ui_text_input",
        "z": "b1a1e0e6.6b3f8",
        "name": "",
        "label": "Prompt",
        "tooltip": "",
        "group": "d3a41f4c.835e6",
        "order": 0,
        "width": "6",
        "height": "1",
        "passthru": true,
        "mode": "text",
        "delay": 300,
        "topic": "prompt",
        "x": 120,
        "y": 60,
        "wires": [
            [
                "d7dcd1f4.8b9f2"
            ]
        ]
    },
    {
        "id": "d7dcd1f4.8b9f2",
        "type": "change",
        "z": "b1a1e0e6.6b3f8",
        "name": "Store Prompt",
        "rules": [
            {
                "t": "set",
                "p": "prompt",
                "pt": "flow",
                "to": "payload",
                "tot": "msg"
            }
        ],
        "action": "",
        "property": "",
        "from": "",
        "to": "",
        "reg": false,
        "x": 300,
        "y": 60,
        "wires": [
            []
        ]
    },
    {
        "id": "d9c35769.b7e938",
        "type": "ui_button",
        "z": "b1a1e0e6.6b3f8",
        "name": "",
        "group": "d3a41f4c.835e6",
        "order": 2,
        "width": "3",
        "height": "1",
        "passthru": false,
        "label": "Generate Image",
        "tooltip": "",
        "color": "",
        "bgcolor": "",
        "icon": "",
        "payload": "",
        "payloadType": "str",
        "topic": "",
        "x": 120,
        "y": 180,
        "wires": [
            [
                "1cbbef4a.e8d3b1"
            ]
        ]
    },
    {
        "id": "1cbbef4a.e8d3b1",
        "type": "function",
        "z": "b1a1e0e6.6b3f8",
        "name": "Build Payload",
        "func": "msg.payload = {\n    prompt: flow.get('prompt'),\n    steps: flow.get('steps')\n};\nmsg.headers = {\n    'Content-Type': 'application/json'\n};\nreturn msg;",
        "outputs": 1,
        "noerr": 0,
        "initialize": "",
        "finalize": "",
        "libs": [],
        "x": 290,
        "y": 180,
        "wires": [
            [
                "f9bf774a.798ef8"
            ]
        ]
    },
    {
        "id": "f9bf774a.798ef8",
        "type": "http request",
        "z": "b1a1e0e6.6b3f8",
        "name": "Stable Diffusion API",
        "method": "POST",
        "ret": "obj",
        "paytoqs": "ignore",
        "url": "http://127.0.0.1:7860/sdapi/v1/txt2img",
        "tls": "",
        "persist": false,
        "proxy": "",
        "authType": "",
        "x": 490,
        "y": 180,
        "wires": [
            [
                "bf0c8a0a.09b6e8"
            ]
        ]
    },
    {
        "id": "bf0c8a0a.09b6e8",
        "type": "function",
        "z": "b1a1e0e6.6b3f8",
        "name": "Process Image",
        "func": "var base64 = msg.payload.images[0].split(\",\", 1)[0];\nmsg.payload = {\n    image: \"data:image/png;base64,\" + base64\n};\nmsg.headers = {\n    'Content-Type': 'application/json'\n};\nmsg.imageData = base64;\nreturn msg;",
        "outputs": 1,
        "noerr": 0,
        "initialize": "",
        "finalize": "",
        "libs": [],
        "x": 690,
        "y": 180,
        "wires": [
            [
                "54d264f1.75e9b4",
                "de5a57e4.b5c398"
            ]
        ]
    },
    {
        "id": "54d264f1.75e9b4",
        "type": "http request",
        "z": "b1a1e0e6.6b3f8",
        "name": "Get PNG Info",
        "method": "POST",
        "ret": "obj",
        "paytoqs": "ignore",
        "url": "http://127.0.0.1:7860/sdapi/v1/png-info",
        "tls": "",
        "persist": false,
        "proxy": "",
        "authType": "",
        "x": 890,
        "y": 180,
        "wires": [
            [
                "daf6d76c.a36d58"
            ]
        ]
    },
    {
        "id": "daf6d76c.a36d58",
        "type": "function",
        "z": "b1a1e0e6.6b3f8",
        "name": "Save Image",
        "func": "var base64Data = msg.imageData;\nvar parameters = msg.payload.info;\nvar fs = require('fs');\nvar path = '/path/to/output.png';\n\nfs.writeFile(path, base64Data, 'base64', function(err) {\n    if (err) {\n        node.error('Failed to save image', err);\n    } else {\n        node.log('Image saved successfully');\n        node.log('Parameters: ' + parameters);\n    }\n});\n\nreturn msg;",
        "outputs": 1,
        "noerr": 0,
        "initialize": "",
        "finalize": "",
        "libs": [],
        "x": 1090,
        "y": 180,
        "wires": [
            []
        ]
    },
    {
        "id": "de5a57e4.b5c398",
        "type": "ui_template",
        "z": "b1a1e0e6.6b3f8",
        "group": "d3a41f4c.835e6",
        "name": "Display Image",
        "order": 3,
        "width": "6",
        "height": "6",
        "format": "<div style=\"text-align:center;\"><img src=\"{{msg.payload.image}}\" style=\"max-width:100%;height:auto;\"></div>",
        "storeOutMessages": true,
        "fwdInMessages": true,
        "templateScope": "local",
        "x": 920,
        "y": 240,
        "wires": [
            []
        ]
    },
    {
        "id": "c9a1c2ed.68d08",
        "type": "ui_text_input",
        "z": "b1a1e0e6.6b3f8",
        "name": "",
        "label": "Steps",
        "tooltip": "",
        "group": "d3a41f4c.835e6",
        "order": 1,
        "width": "6",
        "height": "1",
        "passthru": true,
        "mode": "number",
        "delay": 300,
        "topic": "steps",
        "x": 120,
        "y": 120,
        "wires": [
            [
                "ff4a3dfc.34c22"
            ]
        ]
    },
    {
        "id": "ff4a3dfc.34c22",
        "type": "change",
        "z": "b1a1e0e6.6b3f8",
        "name": "Store Steps",
        "rules": [
            {
                "t": "set",
                "p": "steps",
                "pt": "flow",
                "to": "payload",
                "tot": "msg"
            }
        ],
        "action": "",
        "property": "",
        "from": "",
        "to": "",
        "reg": false,
        "x": 300,
        "y": 120,
        "wires": [
            []
        ]
    },
    {
        "id": "d3a41f4c.835e6",
        "type": "ui_group",
        "name": "Default",
        "tab": "b3c927c4.5e4de8",
        "order": 1,
        "disp": true,
        "width": "6",
        "collapse": false
    },
    {
        "id": "b3c927c4.5e4de8",
        "type": "ui_tab",
        "name": "Home",
        "icon": "dashboard",
        "disabled": false,
        "hidden": false
    }
]

#include <ArduinoBLE.h> 
#include <SoftwareSerial.h>

//  RX/TX 
#define RX_PIN 2
#define TX_PIN 3

// 
#define BUTTON_PIN 7

// 
SoftwareSerial cameraSerial(RX_PIN, TX_PIN); 

// BLE service
BLEService customService("19B10000-E8F2-537E-4F6C-D104768A1214");
// BLE characteristic
BLEUnsignedCharCharacteristic takePhotoCharacteristic("19B10001-E8F2-537E-4F6C-D104768A1214", BLERead | BLENotify | BLEWrite);

void setup() {
  Serial.begin(115200);
  cameraSerial.begin(115200); //  ESP32-CAM 

  // 
  pinMode(BUTTON_PIN, INPUT_PULLUP);

  //  BLE ()
     if (!BLE.begin()) {
    Serial.println("starting BLE failed!");
    while (1);
  }
  // 
  BLE.setLocalName("Nano33Sense");
  BLE.setAdvertisedService(customService);

  // 
  customService.addCharacteristic(takePhotoCharacteristic);

  // 
  BLE.addService(customService);

  // 
  BLE.advertise();

  Serial.println("Bluetooth device active, waiting for connections...");
}
}

void loop() {
  //  ()
  if (cameraSerial.available()) {
    //  ESP32-CAM 
    char data = cameraSerial.read();
    // 
    Serial.print(data); 

  // 
  if (digitalRead(BUTTON_PIN) == LOW) {
    // 
    cameraSerial.write('C'); 
    Serial.println("Sent photo capture command.");
    delay(200); // 
  }

  //  ()
// 
  BLEDevice central = BLE.central();

  // 
  if (central) {
    Serial.print("Connected to central: ");
    Serial.println(central.address());

    // 
    while (central.connected()) {
      if (takePhotoCharacteristic.written()) {
        // 
        cameraSerial.write('C'); 
        Serial.println("Sent photo capture command.");
      }
    }

    // 
    Serial.print("Disconnected from central: ");
    Serial.println(central.address());
  }

}

#include "esp_camera.h"
#include "Arduino.h"
#include "FS.h"                // SD  & SPIFFS 
#include "SD_MMC.h"            // SD 
#include "soc/soc.h"           //  PSRAM
#include "esp_system.h"        //  ESP32

// SD 
// 
// 
#define RX_PIN 3
#define TX_PIN 1
HardwareSerial NanoSerial(1);

//  (0-63, )
#define IMAGE_QUALITY 10

void setup() {
  Serial.begin(115200);
  NanoSerial.begin(115200, SERIAL_8N1, RX_PIN, TX_PIN);

  // 
  camera_config_t config;
  esp_err_t err = esp_camera_init(&config);
  
  if (psramFound()) {
    config.frame_size = FRAMESIZE_UXVGA; // 
    config.jpeg_quality = IMAGE_QUALITY;
    config.fb_count = 2;
  } else {
    Serial.println("PSRAM not found, capturing at lower resolution.");
    config.frame_size = FRAMESIZE_SVGA;
    config.jpeg_quality = IMAGE_QUALITY;
    config.fb_count = 1;
  }

esp_err_t err = esp_camera_init(&config);
  if (err != ESP_OK) {
    Serial.printf("Camera init failed with error 0x%x", err);
    return;
  }
  sensor_t * s = esp_camera_sensor_get();
  // Initial sensors are flipped vertically and mirrorred, so you need to rotate or mirror them
  s->set_vflip(s, true);
  s->set_hmirror(s, true);

  //  SD 
  if (!SD_MMC.begin("/sdcard", true)) {
    Serial.println("Card Mount Failed");
    return;
  }
  uint8_t cardType = SD_MMC.cardType();
  if (cardType == CARD_NONE) {
    Serial.println("No SD card attached");
    return;
  }

  Serial.println("Camera and SD card ready.");
}

void loop() {

   //  Arduino Nano 
  if (cameraSerial.available()) {
    char command = cameraSerial.read();
    if (command == 'C') {
      takeAndSavePhoto();
    }
  }

 if (NanoSerial.available()) {
    char command = NanoSerial.read();
    if (command == 'C') {
      takeAndSavePhoto();
    }
  }

  // 
  camera_fb_t * fb = esp_camera_fb_get();
  if (!fb) {
    Serial.println("Camera capture failed");
    return;
  }
  // 
  Serial.write(fb->buf, fb->len); 
  // 
  esp_camera_fb_return(fb); 
  delay(100); // 
}

void takeAndSavePhoto() {
  // 
  camera_fb_t * fb = esp_camera_fb_get();
  if (!fb) {
    Serial.println("Camera capture failed");
    return;
  }
  // 
  char filename[32];
  static int fileCounter = 0;
  sprintf(filename, "/picture%03d.jpg", fileCounter++);

  //  SD 
  File file = SD_MMC.open(filename, FILE_WRITE);
  if (!file) {
    Serial.println("Failed to open file for writing");
    esp_camera_fb_return(fb);
    return;
  }
  file.write(fb->buf, fb->len); 
  file.close();
  esp_camera_fb_return(fb);

  Serial.printf("Saved photo to %s\n", filename);
}

#include "esp_camera.h"
#include "Arduino.h"
#include <WiFi.h>
#include <img_converters.h>
#include <PubSubClient.h>
#include <esp_timer.h>
#include <img_converters.h>
#include <fb_gfx.h>
#include <soc/soc.h>		  // disable brownout problems
#include <soc/rtc_cntl_reg.h> // disable brownout problems
#include <esp_http_server.h>
#include "include/SSU.h"
#include "FS.h"                // SD  & SPIFFS 
#include "SD_MMC.h"            // SD 
#include "esp_system.h"        //  ESP32
/***************  ***************/
// Wi-Fi info
const char *ssid = "";
const char *password = "";

//phone setting
String strTpImage = (mcuId + "_ssuaiotclass/esp32cam/image");
const char *tpImage = strTpImage.c_str(); // 

String strTpCmdSendImg = (mcuId + "_ssuaiotclass/esp32cam/cmd/sendimage");
const char *tpCmdSendImage = strTpCmdSendImg.c_str(); // 

// SD 
// 
// 
#define RX_PIN 3
#define TX_PIN 1
HardwareSerial NanoSerial(1);

//  (0-63, )
#define IMAGE_QUALITY 10

// ...

void setup() {
  Serial.begin(115200);
  NanoSerial.begin(115200, SERIAL_8N1, RX_PIN, TX_PIN);

  // 
  camera_config_t config;
  esp_err_t err = esp_camera_init(&config);
  
  if (psramFound()) {
    config.frame_size = FRAMESIZE_UXVGA; // 
    config.jpeg_quality = IMAGE_QUALITY;
    config.fb_count = 2;
  } else {
    Serial.println("PSRAM not found, capturing at lower resolution.");
    config.frame_size = FRAMESIZE_SVGA;
    config.jpeg_quality = IMAGE_QUALITY;
    config.fb_count = 1;
  }

esp_err_t err = esp_camera_init(&config);
  if (err != ESP_OK) {
    Serial.printf("Camera init failed with error 0x%x", err);
    return;
  }
  sensor_t * s = esp_camera_sensor_get();
  // Initial sensors are flipped vertically and mirrorred, so you need to rotate or mirror them
  s->set_vflip(s, true);
  s->set_hmirror(s, true);

  //  SD 
  if (!SD_MMC.begin("/sdcard", true)) {
    Serial.println("Card Mount Failed");
    return;
  }
  uint8_t cardType = SD_MMC.cardType();
  if (cardType == CARD_NONE) {
    Serial.println("No SD card attached");
    return;
  }

  Serial.println("Camera and SD card ready.");
}

void loop() {

   //  Arduino Nano 
  if (cameraSerial.available()) {
    char command = cameraSerial.read();
    if (command == 'C') {
      takeAndSavePhoto();
    }
  }

 if (NanoSerial.available()) {
    char command = NanoSerial.read();
    if (command == 'C') {
      takeAndSavePhoto();
    }
  }
void loop() {

   // 
  camera_fb_t * fb = esp_camera_fb_get();
  if (!fb) {
    Serial.println("Camera capture failed");
    return;
  }
  // 
  Serial.write(fb->buf, fb->len); 
  // 
  esp_camera_fb_return(fb); 
  delay(100); // 
}

void takeAndSavePhoto() {
  // 
  camera_fb_t * fb = esp_camera_fb_get();
  if (!fb) {
    Serial.println("Camera capture failed");
    return;
  }
  // 
  char filename[32];
  static int fileCounter = 0;
  sprintf(filename, "/picture%03d.jpg", fileCounter++);

  //  SD 
  File file = SD_MMC.open(filename, FILE_WRITE);
  if (!file) {
    Serial.println("Failed to open file for writing");
    esp_camera_fb_return(fb);
    return;
  }
  file.write(fb->buf, fb->len); 
  file.close();
  esp_camera_fb_return(fb);

  Serial.printf("Saved photo to %s\n", filename);
}

{
  "name": "Config for Chat ID 1721971696938",
  "load_params": {
    "n_ctx": 2048,
    "n_batch": 512,
    "rope_freq_base": 0,
    "rope_freq_scale": 0,
    "n_gpu_layers": 0,
    "use_mlock": false,
    "main_gpu": 0,
    "tensor_split": [
      0
    ],
    "seed": -1,
    "f16_kv": true,
    "use_mmap": true,
    "no_kv_offload": false,
    "num_experts_used": 0
  },
  "inference_params": {
    "n_threads": 4,
    "n_predict": -1,
    "top_k": 40,
    "min_p": 0.05,
    "top_p": 0.95,
    "temp": 0.8,
    "repeat_penalty": 1.1,
    "input_prefix": "<start_of_turn>user\n",
    "input_suffix": "<end_of_turn>\n<start_of_turn>model\n",
    "antiprompt": [
      "<start_of_turn>user",
      "<start_of_turn>model",
      "<end_of_turn>"
    ],
    "pre_prompt": "Please play the role of an AI drawing master who is good at accurately depicting stories as prompts. Please draw \"(enter the picture/story you want to present)\"\n\nConvert it into prompt in the form of English words or short sentences. If it is a short sentence, please use () to cover the short sentence. You can customize the weight and give it the weight of prompt(:1~1.7). And start with \"json Format\", and finally use \"(realistic skin: 1.4) without changing any weights and entries, beautifully detailed, extremely detailed eyes and faces, beautifully detailed eyes , super detailed and beautiful,\nMasterpiece, Top Quality, Best Quality, Realistic, Unity, 8k Wallpaper, Official Art, Extremely Detailed CG Unity 8k Wallpaper, (Original: 1.2), (Photorealistic: 1.4), Ultra-Detailed, High Resolution, Ultra-detailed, fine detail , cinematic lighting,\n\n-- easynegative, paintings, sketches, (worst  quality:2), (low quality:2), (normal quality:2), dot, mole, lowres, normal quality, monochrome, lowres, text, error, cropped, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, username, watermark, signature,Make the ending.\n\nHere are examples:\n\nmanhwa style, (in the video frame:1.8), (in the video camera screen:1.8), (live streaming:2), (Barrage message:2), \n(Video screen:1.8), (video frame:1.8), (video camera frame display time:1.8), \n(video camera frame effects around the picture:1.7), \n(in front of the video camera), (In the picture, she was forced to take a sex video in the classroom, and from the video camera frame, we can see the miserable picture of her body covered in semen:1.3),\n\n(realistic skin: 1.4), beautifully detailed, extremely detailed eyes and faces, beautifully detailed eyes, super detailed and beautiful,\nMasterpiece, Top Quality, Best Quality, Realistic, Unity, 8k Wallpaper, Official Art, Extremely Detailed CG Unity 8k Wallpaper, (Original: 1.2), (Photorealistic: 1.4), Ultra-Detailed, High Resolution, Ultra-detailed, fine detail, cinematic lighting,\n\n-- easynegative, paintings, sketches, (worst  quality:2), (low quality:2), (normal quality:2), dot, mole, lowres, normal quality, monochrome, lowres, text, error, cropped, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, username, watermark, signature,\n\n--\nHere are 3 more Prompt examples:\n1.{masterpiece}, best quality, illustration, single,{1girl},{beautiful detailed eyes}, beautiful detailed stars, beautiful detailed galaxy, beautiful detailed sky, beautiful detailed water, beautiful ,{shorts}, cinematic lighting, blue eyes , silver hair ,white  uniform,{{blue plaid bow}},long sleeves, long hair, {{white dungarees}} ,{agoge}, braided ponytail, bright eyes, light smile,{star in eyes},{sparkling anime eyes},blue collar, blighting stars ,,\n\n2.{best quality}, {{masterpiece}}, {highres}, original, extremely detailed 8K wallpaper, 1girl, {an extremely delicate and beautiful},This is a cartoon style picture,In the garden under the setting sun sat a blonde young girl, holding a flower and light smile,hair flower,sleeveless,horizontal pupils,symbol-shaped pupils,green eyes,gradient background,depth of field,sitting,cinematic lighting, volume lighting, bloom effect, light particles,long hair,ahoge,,\n\n3.masterpiece, best quality, best quality,Amazing,Beautiful red eyes,finely detail,Depth of field,extremely detailed CG unity 8k wallpaper, masterpiece,(((Long dark white hair))),miko,(Hazy fog),{Fluttering hair},{Thick hair},{{{Gelatinous texture}}},{profile},(Ruins of beautiful details),{Close-up of people},{{{Smooth skin}}},(((upper body))),(Smooth and radiant skin),(Smooth and radiant face),Perfect details,Beautifully gorgeous necklace,\n\nNOW Please write a paragraph of XXXX according to the prompt format (for example: \"The woman wearing the Gundam mecha device must have long hair and a delicate face, and the Gundam mecha must be realistic\")",
    "pre_prompt_suffix": "",
    "pre_prompt_prefix": "",
    "seed": -1,
    "tfs_z": 1,
    "typical_p": 1,
    "repeat_last_n": 64,
    "frequency_penalty": 0,
    "presence_penalty": 0,
    "n_keep": 0,
    "logit_bias": {},
    "mirostat": 0,
    "mirostat_tau": 5,
    "mirostat_eta": 0.1,
    "memory_f16": true,
    "multiline_input": false,
    "penalize_nl": true
  }
}

{
  "name": "Config for Chat ID 1721954201726",
  "load_params": {
    "n_ctx": 2048,
    "n_batch": 512,
    "rope_freq_base": 0,
    "rope_freq_scale": 0,
    "n_gpu_layers": -1,
    "use_mlock": false,
    "main_gpu": 0,
    "tensor_split": [
      0
    ],
    "seed": -1,
    "f16_kv": true,
    "use_mmap": true,
    "no_kv_offload": false,
    "num_experts_used": 0
  },
  "inference_params": {
    "n_threads": 4,
    "n_predict": -1,
    "top_k": 40,
    "min_p": 0.05,
    "top_p": 0.95,
    "temp": 0.8,
    "repeat_penalty": 1.1,
    "input_prefix": "<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n",
    "input_suffix": "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
    "antiprompt": [
      "<|start_header_id|>",
      "<|eot_id|>"
    ],
    "pre_prompt": "Please simulate the prompt word generator in stable diffusion. Your task is to describe a scene as concisely as possible. You must first ask me to provide a theme concept, and then you will base it on \"theme (animals, people, places, objects...) + medium (photos, paintings, illustrations, sculptures, graffiti, tapestries...) + environment ( Indoor, outdoor, space, Narnia, underwater, Emerald City, etc....) + lighting (soft light, ambient light, fluorescent light, neon light, studio light....) + color (bright, soft , bright, solid color, color, black and white...) + mood (heavy, calm, noisy, exciting...) + composition (portrait, headshot, close-up, bird's-eye view...) + The painter's name (Picasso...)\" is combined into prompt words (directly combined into sentences without explanation), and 5 prompt words with the same theme must be combined.\nA single prompt word should not exceed 50 words, and each prompt word should be translated into English. Finally, I'm asked if I'm going to end the current task or move on to a different one.\nWell, you start by asking me to provide a theme concept.",
    "pre_prompt_suffix": "",
    "pre_prompt_prefix": "<|start_header_id|>system<|end_header_id|>\n\n",
    "seed": -1,
    "tfs_z": 1,
    "typical_p": 1,
    "repeat_last_n": 64,
    "frequency_penalty": 0,
    "presence_penalty": 0,
    "n_keep": 0,
    "logit_bias": {},
    "mirostat": 0,
    "mirostat_tau": 5,
    "mirostat_eta": 0.1,
    "memory_f16": true,
    "multiline_input": false,
    "penalize_nl": true
  }
}

{
  "name": "Config for Chat ID 1721981219818",
  "load_params": {
    "n_ctx": 2048,
    "n_batch": 512,
    "rope_freq_base": 0,
    "rope_freq_scale": 0,
    "n_gpu_layers": -1,
    "use_mlock": false,
    "main_gpu": 0,
    "tensor_split": [
      0
    ],
    "seed": -1,
    "f16_kv": true,
    "use_mmap": true,
    "no_kv_offload": false,
    "num_experts_used": 0
  },
  "inference_params": {
    "n_threads": 4,
    "n_predict": -1,
    "top_k": 40,
    "min_p": 0.05,
    "top_p": 0.95,
    "temp": 0.8,
    "repeat_penalty": 1.1,
    "input_prefix": "<start_of_turn>user\n",
    "input_suffix": "<end_of_turn>\n<start_of_turn>model\n",
    "antiprompt": [
      "<start_of_turn>user",
      "<start_of_turn>model",
      "<end_of_turn>"
    ],
    "pre_prompt": "\"Create an \"imagine prompt in  json Format\" with a word count limit of 1,500 words for the AI-based text-to-image program for stable diffusion using the following parameters: {\"prompt\": \"[1], [2], [3], [4], [5], [6].\"}\n\nIn this prompt, [1] should be replaced with a user-supplied concept and [2] should be a concise, descriptive summary of the subject. Ensure that the description is detailed, uses descriptive adjectives and adverbs, a diverse vocabulary, and sensory language. Offer context and background information regarding the subject and consider the image's perspective and point of view. Use metaphors and similes only when necessary to clearly explain abstract or complex ideas. Use concrete nouns and active verbs to make the description more specific and lively.\n\n[3] should be a concise summary of the scene's environment. Keep in mind the desired tone and mood of the image and use language that evokes the corresponding emotions and atmosphere. Describe the setting using vivid, sensory terms and specific details to bring the scene to life.\n\n[4] should be a concise description of the mood of the scene, using language that conveys the desired emotions and atmosphere.\n\n[5] should be a concise description of the atmosphere, using descriptive adjectives and adverbs to create the desired atmosphere while considering the overall tone and mood of the image.\n\n[6] should be a concise description of the lighting effect, including types of lights, displays, styles, techniques, global illumination, and shadows. Describe the quality, direction, color, and intensity of the light and how it impacts the mood and atmosphere of the scene. Use specific adjectives and adverbs to portray the desired lighting effect and consider how it will interact with the subject and environment.\n\nIt's important to remember that the descriptions in the prompt should be written together, separated only by commas and spaces, and should not contain any line breaks or colons. Brackets and their contents should not be included, and the prompt should always start with \"\"prompt\": \".\n\nEnsure that the grammar is consistent and avoid using cliches or excess words. Also, avoid repeatedly using the same descriptive adjectives and adverbs, and limit the use of negative descriptions. Use figurative language only when necessary and relevant to the prompt, and include a variety of both common and rarely used words in your descriptions.\n\nThe \"imagine prompt\" must not exceed 1,500 words. The prompt should include the end arguments \"--c X --s Y --q 2,\" where X is a whole number between 1 and 25 and Y is a whole number between 100 and 1000. If the subject looks better vertically, add \"--ar 2:3\" before \"--c,\" and if it looks better horizontally, add \"--ar 3:2\" before \"--c.\" Please randomize the end argument format and fix \"--q 2.\" Donot use double quotation marks or punctuation marks, and use a randomized end suffix format.\n\nWait for a {concept} to be provided before generating the prompt.\"\n",
    "pre_prompt_suffix": "",
    "pre_prompt_prefix": "",
    "seed": -1,
    "tfs_z": 1,
    "typical_p": 1,
    "repeat_last_n": 64,
    "frequency_penalty": 0,
    "presence_penalty": 0,
    "n_keep": 0,
    "logit_bias": {},
    "mirostat": 0,
    "mirostat_tau": 5,
    "mirostat_eta": 0.1,
    "memory_f16": true,
    "multiline_input": false,
    "penalize_nl": true
  }
}

{
  "name": "Config for Chat ID 1721967007245",
  "load_params": {
    "n_ctx": 2048,
    "n_batch": 512,
    "rope_freq_base": 0,
    "rope_freq_scale": 0,
    "n_gpu_layers": 0,
    "use_mlock": false,
    "main_gpu": 0,
    "tensor_split": [
      0
    ],
    "seed": -1,
    "f16_kv": true,
    "use_mmap": true,
    "no_kv_offload": false,
    "num_experts_used": 0
  },
  "inference_params": {
    "n_threads": 4,
    "n_predict": -1,
    "top_k": 40,
    "min_p": 0.05,
    "top_p": 0.95,
    "temp": 0.8,
    "repeat_penalty": 1.1,
    "input_prefix": "<start_of_turn>user\n",
    "input_suffix": "<end_of_turn>\n<start_of_turn>model\n",
    "antiprompt": [
      "<start_of_turn>user",
      "<start_of_turn>model",
      "<end_of_turn>"
    ],
    "pre_prompt": "Help me select one option from @a, @b, @c, @d, @e, @f, and use \", \" to separate the options. Then break the line and select each one from @g and @h. Three options appear, each option is separated by \",\"\n@a\n1. Doctor\n2. Teacher\n3. Engineer\n4. Lawyer\n5. Chef\n6. Accountant\n7. Artist\n8. Writer\n9. Scientist\n10. Athlete\n11. Musician\n12. Entrepreneur\n13. Politician\n14. Pilot\n15. Astronaut\n16. Police officer\n17. Firefighter\n18. Fashion designer\n19. Architect\n20. Psychologist\n21. Witch\n22. Sorceress\n23. Mage\n24. Enchantress\n25. Druidess\n26. Priestess\n27. Summoner\n28. Alchemist\n29. Warlock\n30. Shaman\n31. Paladin\n32. Fairy queen\n33. Oracle\n34. Necromancer\n35. Elemental mage\n36. Siren\n37. Knight\n38. Demon hunter\n39. Dragon tamer\n40. Witch hunter\n\n@b\n1. Wide-angle lens\n2. Telephoto lens\n3. Prime lens\n4. Zoom lens\n5. Macro lens\n6. Fish-eye lens\n7. Tilt-shift lens\n8. Portrait lens\n9. Landscape lens\n10. Standard lens\n\n@c\n1. Bird's-eye view\n2. Worm's-eye view\n3. High-angle view\n4. Low-angle view\n5. Close-up view\n6. Wide-angle view\n7. Telephoto view\n8. Panoramic view\n9. Dutch angle view\n10. Point-of-view (POV) view\n\n@d\n1. Bob\n2. Pixie cut\n3. Long layers\n4. Bangs\n5. Ponytail\n6. Updo\n7. Braid\n8. Top knot\n9. Beach waves\n10. Half-up, half-down\n\n@e\n1. Smile\n2. Frown\n3. Laugh\n4. Cry\n5. Surprise\n6. Anger\n7. Disgust\n8. Fear\n9. Confusion\n10. Excitement\n\n@f\n1. Forest\n2. Beach\n3. Cityscape\n4. Mountains\n5. Desert\n6. Skyline\n7. Countryside\n8. Underwater\n9. Sunset\n10. Starry night\n\n@g\n1. Shading\n2. Highlights\n3. Shadows\n4. Depth\n5. Texture\n6. Contrast\n7. Blending\n8. Brushstrokes\n9. Gradients\n10. Details\n11. Patterns\n12. Translucency\n13. Reflections\n14. Surface\n15. Illumination\n16. Opacity\n17. Hatching\n18. Subtle\n19. Luminosity\n20. Glimmer\n\n@h\n1. Stunning\n2. Lifelike\n3. Vibrant\n4. Breathtaking\n5. Seamless\n6. Exquisite\n7. Impressive\n8. Striking\n9. Immersive\n10. Captivating\n11. Dynamic\n12. Realistic\n13. Masterful\n14. Polished\n15. Marvelous\n16. Sophisticated\n17. Enchanting\n18. High-quality\n19. Eye-catching\n20. Photorealistic\n\nWhen answering, you dont need to add @a~@h before the options. Just use \", \" to separate each option directly, and no other words. Add a new line at the end: \"-- easynegative, (backlight:1.3) , paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2),\"\n\nFor example:\nscientist, portrait lens, close-up view, ponytail, anger, mountains, blending, Shading, Glimmer, captivating, lifelike, exquisite.\n\n-- easynegative, (backlight:1.3), paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2),\n",
    "pre_prompt_suffix": "",
    "pre_prompt_prefix": "",
    "seed": -1,
    "tfs_z": 1,
    "typical_p": 1,
    "repeat_last_n": 64,
    "frequency_penalty": 0,
    "presence_penalty": 0,
    "n_keep": 0,
    "logit_bias": {},
    "mirostat": 0,
    "mirostat_tau": 5,
    "mirostat_eta": 0.1,
    "memory_f16": true,
    "multiline_input": false,
    "penalize_nl": true
  }
}

from googleapiclient.discovery import build
from googleapiclient.http import MediaFileUpload
from google.oauth2 import service_account

#  Google Cloud 
CREDENTIALS_FILE = 'path/to/your/credentials.json'
SCOPES = ['https://www.googleapis.com/auth/drive']

#  Google Drive  ID
FOLDER_ID = 'your_folder_id'

# 
IMAGES_FOLDER = r'C:\your_paths'

def upload_images():
    """ Google Drive"""

    credentials = service_account.Credentials.from_service_account_file(CREDENTIALS_FILE, scopes=SCOPES)
    service = build('drive', 'v3', credentials=credentials)

    for filename in os.listdir(IMAGES_FOLDER):
        filepath = os.path.join(IMAGES_FOLDER, filename)
        if os.path.isfile(filepath):
            media = MediaFileUpload(filepath, mimetype='image/jpeg')  #  mimetype
            file_metadata = {
                'name': filename,
                'parents': [FOLDER_ID]
            }
            file = service.files().create(body=file_metadata,
                                        media_body=media,
                                        fields='id').execute()
            print(f': {filename} (ID: {file.get("id")})')

if __name__ == '__main__':
    upload_images()

[
    {
      "slideTitle": "VBA ",
      "slideContent": [
        "VBA Visual Basic for Applications",
        " Office  ExcelWord ",
        "VBA "
      ],
      "imagePath": "https://source.unsplash.com/960x640/?VBA"
    },
    {
      "slideTitle": "VBA ",
      "slideContent": [
        " Excel VBA ",
        "",
        "VBA "
      ],
      "imagePath": "https://source.unsplash.com/960x640/?VBA,Excel"
    },
    {
      "slideTitle": "VBA ",
      "slideContent": [
        "VBA  Office ",
        " VBA ",
        "VBA "
      ],
      "imagePath": "https://source.unsplash.com/960x640/?programming"
    }
  ]

Sub CreatePowerPointFromJSON(jsonData As String)
    '  JSON  JSON 
    Dim json As Object
    Set json = JsonConverter.ParseJson(jsonData)

    '  PowerPoint 
    Dim pptApp As Object
    Set pptApp = CreateObject("PowerPoint.Application")
    pptApp.Visible = True

    '  PowerPoint 
    Dim pptPres As Object
    Set pptPres = pptApp.Presentations.Add

    '  JSON 
    Dim i As Long
    For i = 1 To json.Count
        ' 
        Dim slideData As Object
        Set slideData = json(i)

        ' 
        Dim slide As Object
        Set slide = pptPres.Slides.Add(pptPres.Slides.Count + 1, ppLayoutTitleAndContent)

        ' 
        slide.Shapes(1).TextFrame.TextRange.Text = slideData("slideTitle")

        ' 
        Dim j As Long
        Dim content As String
        For j = 1 To slideData("slideContent").Count
            content = content & "- " & slideData("slideContent")(j) & vbCrLf
        Next j
        slide.Shapes(2).TextFrame.TextRange.Text = content

        ' 
        If slideData.Exists("imagePath") Then
            slide.Shapes.AddPicture slideData("imagePath"), False, True, 0, 0
        End If
    Next i
End Sub

' 
Function ReadTextFile(filePath As String) As String
    Dim fso As Object, file As Object
    Dim fileContent As String

    Set fso = CreateObject("Scripting.FileSystemObject")
    Set file = fso.OpenTextFile(filePath, 1) ' 1  ForReading

    fileContent = file.ReadAll
    file.Close
    Set file = Nothing
    Set fso = Nothing

    ReadTextFile = fileContent
End Function

'  JSON  PowerPoint
Sub TestCreatePowerPoint()
    '  JSON 
    Dim filePath As String
    filePath = "C:\downloads\ptt.json"

    '  jsonData 
    Dim jsonData As String
    jsonData = ReadTextFile(filePath)

    '  PowerPoint
    Call CreatePowerPointFromJSON(jsonData)
End Sub

#include <WiFi.h>
#include <HTTPClient.h>
#include <ArduinoJson.h>

// Replace with your network credentials
const char *ssid     = "";     //Wifi
const char *password = "";     //Wifi

// Replace with your OpenAI API key
const char* apiKey = "";

void setup() {
  // Initialize Serial
  Serial.begin(115200);

  // Connect to Wi-Fi network
  WiFi.mode(WIFI_STA);
  WiFi.begin(WIFI_SSID, WIFI_PASSWORD);
  Serial.print("Connecting to WiFi ..");
  while (WiFi.status() != WL_CONNECTED) {
      Serial.print('.');
      delay(1000);
  }
  Serial.println(WiFi.localIP());

  // Send request to OpenAI API
  String inputText = "Hello, ChatGPT!";
  String apiUrl = "https://api.openai.com/v1/completions";
  String payload = "{\"prompt\":\"" + inputText + "\",\"max_tokens\":100, \"model\": \"text-davinci-003\"}";

  HTTPClient http;
  http.begin(apiUrl);
  http.addHeader("Content-Type", "application/json");
  http.addHeader("Authorization", "Bearer " + String(apiKey));
  
  int httpResponseCode = http.POST(payload);
  if (httpResponseCode == 200) {
    String response = http.getString();
  
    // Parse JSON response
    DynamicJsonDocument jsonDoc(1024);
    deserializeJson(jsonDoc, response);
    String outputText = jsonDoc["choices"][0]["text"];
    Serial.println(outputText);
  } else {
    Serial.printf("Error %i \n", httpResponseCode);
  }
}

void loop() {
  // do nothing
}

Credits

林望儒

1 project • 1 follower

劉育庭

1 project • 1 follower

Thanks to Kun-Neng Hung.

Embed the widget on your own site

AI MasterDream copilot: AMD Enhanced Artistic Rendering Sys

AI MasterDream copilot: AMD Enhanced Artistic Rendering Sys

Things used in this project

Hardware components

Software apps and online services

Hand tools and fabrication machines

Story

The Birth of Our Team

# AI MasterDream copilot: AMD Enhanced Artistic Rendering System

1. Introduction

2. System Requirements

3. Installation Guide

4. User Guide

5. Troubleshooting

Custom parts and enclosures

ESP32 Material

ESP32 Enclosures

a display of completed picture

Schematics

ESP32 pinout

ESP32-CAM Pin Assignment Diagram

NANO 33 SENSE REV2 Pin Assignment Diagram

ESP32 CAM PINOUT

NANO 33 SENSE REV2 pinout

Code

generate ptt sample with gas.js

requirements.txt

text-text-nodered.json

nano33_sample.ino

esp32cam_test.ino

esp32cam_cellphone.ino

model_config_AI drawing master.json

model_config_conciseprompt.json

ESP32 TFT CODE SETTING

model_config_image prompt machine.json

model_config_Randomly generated prompts.json

uploadpic.py

ptt.json

sameple_code_vba.ini

Terraform_sample.rar

openai_test_sample.ino

Credits

林望儒

劉育庭

Comments

Related channels and tags