Articles

Examples of applications in the architectural and civil engineering domain through extensions of generative AI (e.g. ChatGPT)

With the release of ChatGPT's API, various extensions such as libraries and plug-ins have been released. Here, we would like to focus on extensions (libraries, open source software, and plug-ins) for generative AI (especially ChatGPT) and describe a brief introduction to each and examples of applications in building and engineering domains such as real estate and construction.


Libraries

In this section, we would like to introduce LangChain and LlamaIndex as libraries that can be used to develop solutions incorporating ChatGPT API in Python, etc. Here, we would like to provide an overview of these libraries and examples of applications in the architectural and civil engineering fields.

LangChain

LangChain execution flow

https://blog.langchain.dev/langchain-chat/

LangChain is a framework for developing applications that work with language models, such as ChatGPT, that can invoke language models, connect to other data sources, and interact with each other in various situations.

As a mechanism, it is mainly composed of the following modules

  • Models: Language models such as ChatGPT
  • Prompts: management and optimization of prompts
  • Memory: storage of previous interactions
  • Indexes: interfaces for loading, querying and updating external data
  • Chain: sequence of calls, such as LLM
  • Agent: a chain in which a language model repeats an instruction, executes actions until completion, and observes the result
  • Callbacks: Log intermediate steps in the chain to observe, debug, and evaluate the application

The following is a list of common use cases from the official documentation that could be applied to the architecture and civil engineering domain.

  • Personal Assistant: Executes commands and remembers how they were executed, gaining knowledge of the user's interactions.
  • Memorize email exchanges between your own personnel and clients or partners, and gradually allow LangChain to conduct client interactions automatically.
  • Question and Answer: Answer questions about a specific document, using only the information contained in that document.
  • A question about a competitive bidding project will be answered by ChatGPT based on the solicitation for that competitive bid.
  • Chatbot: Can interact with ChatGPT as well as other language models.
  • It can interact with language models that have been trained to use terms from the architectural and civil engineering domain.
  • Code understanding: Can analyze codes using language models.
  • Can input programs used in structural calculations and refer to their explanations and errors.
  • Extraction: Extract structured information from text.
  • When building design documents are entered, the information is automatically organized in Excel.
  • Summarize: Compresses long documents.
  • Enter a competitive bid solicitation and it summarizes or extracts necessary information.


Official Documents

Welcome to LangChain - 🦜🔗 LangChain 0.0.192python.langchain.com


LlamaIndex

Flow of running LlamaIndex

https://gpt-index.readthedocs.io/en/latest/guides/primer/index_guide.html

Large-scale natural language models (LLMs) such as ChatGPT are pre-trained on large amounts of public data; therefore, in-house data that has not been pre-trained or up-to-date is not referenced LlamaIndex works with such data to complement the data referenced by the language model.

Functions include

  • Import of existing data (API, PDF, docs, SQL, etc.)
  • Ability to structure data so that it can be easily used in the language model
  • Advanced search against data
  • Integration with various application frameworks (LangChain, Flask, Docker, ChatGPT, etc.)

As for application examples, since general use cases are enumerated in the official documentation, examples that could be applied to the architectural and civil engineering domains are described from among them.

  • Referencing data: Language models respond with references to external data
  • ChatGPT answers questions about field work, referring to internal manuals on construction methods.
  • Agent: Automates reasoning and decision making, receiving user input and examining it internally in the system to return the correct results.
  • Prepare multiple documents within the company, and LlamaIndex will select the appropriate reference material to answer questions to ChatGPT.
  • Chatbots: Chatbots can be created to supplement ChatGPT and other knowledge.
  • It can ask questions and gather information about the latest zoning regulations by district.


Official Documents

LlamaIndex 🦙 0.6.21gpt-index.readthedocs.io


Open Source Software

AutoGPT and GoalGPT are representative of advanced software available to the public; here we provide an overview of their open source software and examples of applications in the architectural and civil engineering domain.

AutoGPT

Logo

https://news.agpt.co/

Auto-GPT autonomously achieves set goals through a chain of "thoughts" in a language model by means of GPT-4.

Features include

  • Internet access for searching and information gathering
  • Management of long-term and short-term memory
  • GPT-4 instances for text generation
  • Access to popular websites and platforms
  • GPT-3.5 for file storage and summary processing
  • Extensibility through plug-ins

With Auto-GPT, you can do almost any task You can use Auto-GPT to automate research, write texts, build apps, etc. Here are a few examples

  • Automatically locate construction companies close to the project site and order quotes
  • Developing applications
  • Building algorithms to predict demand forecasts for each material
  • Gathering information on regulations and surrounding environment of the project site from the Internet and organizing it into documents
  • Automate communication with clients
  • Research markets and prepare detailed reports
  • Automatically collect and distribute information related to competitive bidding.

Other similar products include BabyAGI and GoalGPT.

Official Documents

https://github.com/Significant-Gravitas/AutoGPT



Plug-ins


Since May 2023, plug-ins for ChatGPT can be utilized only with paid plans; here, among the various plug-ins, we introduce and describe examples of applications of plug-ins related to the architectural and civil engineering domain.


Scraper

Scraper allows you to extract information from the web, for example, to create a list of prices of materials of a certain manufacturer.

Lexishopper

Allows you to search for products on Amazon, for example, to compile a list of building materials products in a particular category.

Wolfram

Calculate and chart, for example, land prices in different regions and make comparisons for real estate transactions.

Show Me

Create diagrams, e.g. flowcharts of the processes of each construction project.

VoxScript

Search YouTube videos, e.g., to collect videos that may be useful for construction methods for a specific project.

Manorlead

Manorlead allows you to search for real estate in the U.S. For example, you can extract properties below a certain sales price in a certain neighborhood.

World News

You can ask questions about world news, for example, about the fluctuation of stock prices around the world.


Official Documentation

ChatGPT pluginsWe've implemented initial support for plugins in ChatGPT.

openai.com


Summary

We hope you found this article useful and informative.

Back to articles