Technology Blogs by Members
Explore a vibrant mix of technical expertise, industry insights, and tech buzz in member blogs covering SAP products, technology, and events. Get in the mix!
cancel
Showing results for 
Search instead for 
Did you mean: 
patrosrinivasa
Participant

Introduction:


Hello all, Welcome back! In this blog post, we will continue our exploration of utilizing AI in SAP Build Apps. In previous Blog, we had discussed how we can automate Sales-Order query with CHATGPT. Now, we will delve into the integration of SAP Build Apps with chatGPT Turbo-3.5. This integration has the potential to revolutionize the development of SAP applications, offering a faster, smarter and more efficient process.

In the era of digital transformation, businesses are constantly seeking innovative ways to streamline their operations and deliver exceptional user experiences. SAP, a global leader in enterprise software, has always been at the forefront of providing cutting-edge solutions.

SAP Build Apps is a professional application development solution designed for anyone to quickly create apps without code regardless of role or skill level. Well, You can try with free tier option where there are some limitations too (like you can't hold project for too long, after 30days Lobby access will be deleted etc). So,pleae do check. Well, I have Paid version, So I don’t have any such issues 😊

Understanding chatGPT Turbo-3.5:


ChatGPT Turbo-3.5 is an advanced language model powered by OpenAI's GPT-3.5 architecture. It possesses a vast knowledge base and is trained on a wide range of text sources, including technical documentation, code repositories, and SAP-related content. With its natural language processing capabilities, it can assist developers in generating code snippets, providing insights, and answering complex questions related to SAP development.

Enhancing the SAP Build Apps Experience:


SAP Build Apps is an intuitive platform that enables developers to rapidly create, customize, and extend SAP applications. By integrating chatGPT Turbo-3.5 into SAP Build Apps, developers gain access to an intelligent virtual assistant that can significantly enhance their productivity and decision-making process.

a) Code Generation Assistance:


ChatGPT Turbo-3.5 can assist developers in generating code snippets for common SAP development tasks. Whether it's creating custom modules, implementing UI elements, or integrating with backend systems, the language model can provide accurate and context-aware code suggestions, reducing development time and effort.

b) Technical Documentation Exploration:


Finding the right information in vast technical documentation can be time-consuming. ChatGPT Turbo-3.5 can parse through extensive SAP documentation and provide developers with relevant answers to their queries, saving them valuable research time. From understanding API integrations to troubleshooting issues, the virtual assistant offers real-time assistance, empowering developers to make informed decisions.

c) Best Practice Recommendations:


Developers often encounter challenges when adhering to SAP development best practices. ChatGPT Turbo-3.5 can analyze the development context and recommend best practices for various scenarios. It can suggest optimal design patterns, performance optimization techniques, and security measures, enabling developers to deliver high-quality applications that align with SAP guidelines.

As you have already knew that AI evaluation is rapidly increasing so is its models. If you compare initial models, there were some immature or incorrect responses were seen but after pre-trained from human interactions/responses AI models have been improvising the prompts very effectively.

In this blog I am going to show you how can we utilize latest chatGPT openai model integration with most trending low-code no-code technology none other than “SAP Build Apps”. You can see the difference between latest model VS earlier model response in this integration’s final result 😊

Streamlining Collaboration and Support:


In addition to assisting individual developers, ChatGPT Turbo-3.5 can facilitate collaboration and support within development teams.

a) Collaborative Development:


With its conversational interface, ChatGPT Turbo-3.5 can act as a bridge between developers working on different aspects of the same SAP application. It allows them to exchange ideas, seek suggestions, and review code snippets in a seamless manner, promoting collaboration and knowledge sharing.

b) On-Demand Support:


The virtual assistant can serve as a round-the-clock support system, providing instant responses to developers' questions. Whether it's a query related to debugging, performance optimization, or platform-specific challenges, ChatGPT Turbo-3.5 offers timely guidance, minimizing downtime and fostering continuous development.

So, Let’s build AI in SAP Build Apps without any delay. Wait, I forgot 1thing.. Before going there we need to check Prerequisites. Here is check list,

  1. Please make sure you have subscribed to SAP BTP account and create project in SAP Build Apps lobby.

  2. You must create a new secret key in OPENAI. Don’t reveal to anyone 😉


Jump Into App Build


First, we need to create frontend design UI Canvas with help of predefined controls. We don’t need all the controls right now So let add a few basic controls like Title, Input field, Button, Text, and Spinner. You can change basic settings of Homepage like background color, text style, label, height/Width, cell padding etc as per your need but there are some properties that we need to consider which needs to bind with certain values/formulae/static text etc. So be careful what properties you are modifying.

At last, enduser will enter question in Input field and output response can be shown in TEXT control.

 


SAP Build Apps-UI Canvas


 

In Left-hand side, you can see all different types of controls that are available and Right-hand side you can see selected control’s properties window. Change accordingly.

Here we need to define some Page Variables to store a) A variable that stores question/hint provided by end user b) another variable that fetch chatGPT response and display to end-user c) a Boolean variable that defines button triggered or not (true/false)















a)      Question Variable Need to bind it to Input control.
b)     Answer Variable Need to bind it to Text control.
c)      Boolean Variable True/False

 


Execution Logic


Below portion is shows the execution logic flow. Once we click on Button control execution logic flow will trigger. On 3rd step If there is any issue occurs Alert control will show a message with 3rd node or else it will process to next stage from node 1 of HTTP request.

HTTP request control is heart of this logic execution. So here are some data you need to pre-define.

URL will be Static Text and select methos as POST.

Header section for HTTP request will be as follow, here you have to replace your Bearer <your Secret Key>


Header Section


 

So now it’s testing Time... Here is response from latest AI model.


AI Response


 

Let see the difference between earlier model VS Latest model response.








Earlier Model



New Model















 

Though old model has able to provide response on almost all question, it was unable to provide response for some questions due to its limitations and data it was trained on. But on other hand new model can answer any questions you ask because This model is the most advanced language model to date and is more cost-effective, being 10 times cheaper than previous GPT-3 models. Unlike previous models that processed unstructured text as a sequence of tokens, the new GPT-3.5-Turbo model uses a different approach that consumes a sequence of messages in a format called Chat Markup Language (ChatML).

 

Conclusion:


The integration of SAP Build Apps with chatGPT Turbo-3.5 opens new horizons for SAP application development. By leveraging the power of advanced language processing and AI capabilities, developers can expedite development cycles, improve code quality, and enhance collaboration within their teams. You can enhance this use case furthermore as per your requirement.

Hope you have enjoyed this. I am mentioning few links for more explore.

Happy Learning... Happy Explore..

See you in Next Blog.

Reference Links:


https://learning.sap.com/

https://platform.openai.com/docs/introduction

https://help.sap.com/docs/build-apps/service-guide/introduction-936ed0c9e030469ab58fff62542c76da

 
5 Comments
pr285443
Explorer
0 Kudos
Great blog patrosrinivasa .

My question is, can we train model with our own training data?

We wanted to provide some train data to this model.

 

Regards,
Prince
thomasyuan
Explorer
0 Kudos
Finetuning gpt-3.5-turbo is not available at the moment. You may want to consider grounding.

https://techcommunity.microsoft.com/t5/fasttrack-for-azure/grounding-llms/ba-p/3843857
patrosrinivasa
Participant
Thank you Prince for reading.

To Answer your Qusetion-"can we train model with our own training data?" Answer is YES and NO. Like thomasyuan mentioned currently Finetuning gpt-3.5-turbo is not available (where as you have seen this latest model basic usage in my blog post).

Well, on bright side, you can fine-tune base GPT-3 models though(As of March 1, 2023). This is because OpenAI has not yet released the fine-tuning code for gpt-3.5-turbo.

There are a few reasons why finetuning gpt-3.5-turbo is not available yet. First, gpt-3.5-turbo is a newer model and OpenAI is still working on improving it. Second, finetuning a large language model is a computationally expensive process, and OpenAI may not have the resources to finetune gpt-3.5-turbo for everyone who wants to use it.

However, there is a chance that finetuning gpt-3.5-turbo will be available in the future. OpenAI is constantly updating its models and adding new features, so it is possible that they will release the fine-tuning code for gpt-3.5-turbo in the future.

In the meantime, if you need to finetune a large language model, you can use one of the base GPT-3 models. These models are still very powerful and can be finetuned to achieve good results 🙂

good luck

 
pr285443
Explorer
Got it patrosrinivasa, thanks for the information.

Also, I tried above today and I need information on what values did you bind to below execution logic flow.
Can you share the screenshot of that also. So, I can try binding correctly as currently it ask for "Data variable" but we had only created Page variable.

What binding we have to do in these? We only did mapping for HTTP Request node only.


 

 

 

Regards,
Prince
bkershaw7
Discoverer
0 Kudos
Thanks for the blog patrosrinivasa I also had a similar question to prnndnndndnddn

I've done the mapping for the HTTP request but don't get a pull from the ChatGPT API, is there another step we're missing here?

Many thanks,
Brodie
Labels in this area