ArchiWeb: A web platform for AI-driven early-stage architectural design

Abstract

As society confronts increasingly complex demands and the growing need for carbon-neutral architecture, Al-driven design methodologies are evolving rapidly. However, the lack of a unified integration platform in the design process continues to hinder AI's integration into real-world workflows. To address this challenge, we introduce ArchiWeb, a web-based platform specifically built to support AI-driven processes in early-stage architectural design. ArchiWeb transforms architectural representation and problem formulation by utilizing lightweight data protocols and a modular algorithmic network within an interactive web environment. Through its cloud-native, open-architecture framework, ArchiWeb enables deeper integration of AI technologies while accelerating the accumulation, sharing, and reuse of design knowledge across projects and disciplines. Ultimately, ArchiWeb aims to drive architectural design toward greater intelligence, efficiency, and sustainability-supporting the transition to data-informed, computationally enabled, and environmentally responsible design practices.

Headnote

KEYWORDS

Al-driven platform;

Data protocol;

Digital workflow;

Algorithmic design;

Web based interactivity

Abstract As society confronts increasingly complex demands and the growing need for carbon-neutral architecture, Al-driven design methodologies are evolving rapidly. However, the lack of a unified integration platform in the design process continues to hinder AI's integration into real-world workflows. To address this challenge, we introduce ArchiWeb, a web-based platform specifically built to support AI-driven processes in early-stage architectural design. ArchiWeb transforms architectural representation and problem formulation by utilizing lightweight data protocols and a modular algorithmic network within an interactive web environment. Through its cloud-native, open-architecture framework, ArchiWeb enables deeper integration of AI technologies while accelerating the accumulation, sharing, and reuse of design knowledge across projects and disciplines. Ultimately, ArchiWeb aims to drive architectural design toward greater intelligence, efficiency, and sustainability-supporting the transition to data-informed, computationally enabled, and environmentally responsible design practices.

1. Introduction

In the architecture, engineering, and construction (AEC) domain, digital design is undergoing a profound shift from traditional modeling tools toward algorithm-driven, intelligent systems. With the rise of computational design, parametric modeling, and generative design (Caetano t al., 2020), architects and designers increasingly rely on Programming to explore complex geometries (Hua, 2016), optimize structural performance (Wang, 2022), and achieve highly customized design solutions. The advent of generative pre-trained models (GPT) (Radford and Narasimhan, 2018) has empowered architects to codify design rules, bringing us closer to the vision of "machines collaborating with users to solve architectural problems" (Negroponte, 1976). Yet, through this evolution, a core question has emerged: how can we realize efficient data exchange and system integration across heterogeneous design platforms, diverse programming languages, and distributed algorithmic modules.

As an established practice for addressing architectural problems through programming, generative design uses a broad range of Al algorithms-including genetic algorithms and dynamic programming based optimization methods-to generate optimal design solutions by combining parametric rules with performance objectives, using an automated, iterative solving process (Guo and Li, 2017). However, the complex constraint conditions and real-time feedback mechanisms inherent to this process are difficult to implement within current BIM and CAD platforms (Steenson, 2017; Svetel et al., 2018). Moreover, most general-purpose Al and machine-learning frameworks primarily rely on text and image data formats for input and output, lacking deep expressive capabilities for geometric semantics (Hovestadt et al., 2020). Consequently, efforts to integrate Al, automated analysis, real-time collaboration, and related technologies into architectural design workflows often encounter challenges such as geometric information loss and interface incompatibilities.

To address these challenges, this paper proposes ArchiWeb, a distributed algorithmic network platform for Al-driven architectural design (Fig. 1). The platform is built on a lightweight, extensible geometric semantic data exchange protocol called ArchiJSON, establishing a unified data representation system. By utilizing WebSocket for real-time communication, ArchiWeb supports dynamic connections and collaborative computation of algorithmic modules developed in various programming languages, whether deployed on a local network or in the cloud. In ArchiWeb, architects take on the role of algorithmic module developers, enabling them to flexibly leverage Al capabilities, connect and configure existing modules, and develop their own web-based applications. The main contributions of ArchiWeb are as follows:

1) Lightweight data protocols: enables seamless, structured communication between Al modules, ensuring interoperability across platforms and tools. This facilitates flexible data exchange, which is crucial for managing the indeterminacy and iterative nature of early design.

2) Modular algorithmic network: allows architects to flexibly combine Al-driven algorithms, automating complex tasks and expanding the exploration space in early design stages. This flexibility supports the ambiguity and iteration inherent in early design, enabling diverse algorithms to work together in real-time.

3) Interactive web interface: provides visual representations of architectural designs, offering instant feedback and fostering a collaborative and dynamic design environment. It supports continuous iteration and adapts to evolving design needs.

This paper first reviews the development of data protocols and Al algorithms in the field of Computer-Aided Architectural Design (CAAD), systematically examining the characteristics of existing web-based platforms. It then delves into the design and implementation of ArchiWeb, detailing its technical architecture, core functionalities, and user experience. Finally, a series of application cases demonstrate the effectiveness of ArchiWeb, highlighting its unique advantages in real-time, Al-enhanced collaboration and design exploration.

providing robust support for innovative design practices and the execution of complex projects. Interactive design galleries have transformed how designers engage with design alternatives, enabling comparative analysis and decision-making across multiple options (Woodbury et al., 2017). Luna Moth provides a web-based integrated development environment for architectural algorithmic design (Alfaiate, 2017), while Speckle facilitates networkinteractive data mapping to streamline information exchange throughout design and construction workflows (Poinet et al., 2020). Rhino Compute illustrates the shift of traditional CAD software toward web-based interfaces, enabling cloud-based generative design services accessible via the internet. Coenders proposes a next-generation parametric design framework (Coenders, 2021), offering an integrated solution for parametric and generative design processes.

In the commercial market, web-based design platforms already provide functionalities such as massing generation, performance optimization, and floor-plan layout. However, a systematic survey of cloud-based parametric design frameworks (Table 1) reveals two critical shortcomings: first, most platforms are purpose-built for specific building typologies; second, no universal framework currently achieves loose coupling with mainstream CAD software or enables invocation of algorithmic modules. Thus, there is an imperative need for tools that integrate extensibility and interoperability to foster deeper integration and interactive innovation in generative design during early-stage development.

3. Overview of the ArchiWeb framework

This section presents a comprehensive overview of the ArchiWeb framework, focusing on its three core components: lightweight data definition, algorithmic network construction, and interactive web platform.

3.1. Lightweight data protocols

ArchiWeb utilizes the ArchiJSON data protocol, which is characterized by its lightweight, compact, and extensible features. As a JSON-based data exchange format, ArchiJSON is specifically designed for efficiently transferring semantically rich architectural data. It serves as a common language within microservices architecture, ensuring interoperability among diverse computational design algorithmic modules. This protocol enables integration of heterogeneous modules, streamlining data exchange and fostering collaborative innovation in generative design workflows (Fig. 2).

3.1.1. Geometric based representation

ArchiJSON focuses on geometric objects, assigning semantic metadata to these objects or global entities. Common attributes include geometric type (type), property objects (properties), unique identifiers (uuid), and linear transformation matrices (matrix). Basic geometric elements (geometry) are categorized into geometric primitives and topological structures. By assembling these data components, users can construct comprehensive data packages that encapsulate architectural models and computational workflows.

Geometric primitives serves as the basic units used for processing, rendering, and storage, and function as architectural forms to represent basic volumes, spatial configurations, and other design elements. In ArchiJSON, these primitives primarily consist of standard parametric geometric elements from Constructive Solid Geometry (CSG), such as circles, rectangles, and squares, which can be parametrically defined across platforms by specifying attribute parameters. The geometric primitives implemented in the current version of ArchiJSON are illustrated in Fig. 3, with future extensions will consider complex curve and surface representations.

In ArchiJSON Mesh, Faces define the topological structure, while Vertices represent the geometric positions of points. Each face strictly adheres to a counterclockwise vertex ordering, and its normal vector is determined by applying the right-hand rule to this sequence. Faces utilize vertex sequences of arbitrary length, enabling the representation of any simple planar polygon, including concave configurations. This approach minimizes redundant data transfer by reusing shared edges. However, certain platforms require triangulation for rendering. In such cases, each face must be subdivided into triangles within its plane and aligned with the orientation defined by its normal vector. The resulting index lists are then aggregated to construct a triangular mesh.

Each geometric element in ArchiJSON supports customization of user-defined properties to address differentiated design requirements, while complex forms are constructed by assembling primitive geometries, thereby extending the core data schema. For instance, a spline curve can be represented by introducing an algorithm field within its Vertices attribute, executing the specified interpolation method to compute control points, and discretizing the resulting polyline into linear segments. Similarly, a nested polygon with interior voids can be encoded by annotating inner rings in clockwise order within its properties and programmatically transforming these annotations into trimmed planar surfaces or mesh topologies. Similar strategies can be applied to describe more intricate geometries and their topological relationships, aligning with ArchiWeb's modular approach to lightweight data representation and interoperability.

3.1.2. Data aggregation and open interoperability

In practical applications, data is assembled into transmission-ready packets. A single ArchiJSON packet encapsulates all necessary information, with the current version including the following components: 1) the metadata field records the ArchiJSON version; 2) a list of all the geometries contained within the packet; 3) a global properties object. Users may optionally document the transfer contents via a descriptive text field.

JSON defines a standardized format for parsing data across programming languages. ArchiJSON, built on a minimalist keyword set, employs open data structures that enable users to customize data templates or leverage predefined standards. This design enhances flexibility and addresses interoperability challenges commonly encountered in other approaches.

3.2. Modular algorithmic network

The Modular algorithmic network is the core infrastructure of ArchiWeb. Users can integrate custom algorithmic modules into ArchiWeb, enabling these modules to communicate in real time and facilitate dynamic data exchange through ArchiJSON.

3.2.1. Algorithmic modules

ArchiWeb provides standardized interfaces for data exchange in multiple programming languages (currently Java, JavaScript, Python, and C#), such as the ArchiServer and ArchiJSON classes. The ArchiServer class asynchronously creates a persistent service instance using token and identity parameters and communicates with a central server in real time. Users can customize message-handling behavior through a listener function, while the ArchiJSON class encapsulates transport data and supports both directed (point-to-point) and broadcast messaging.

Algorithmic modules can be deployed independently on any network-accessible server after registration and authentication using token and identity parameters (Fig. 4). This approach maximizes hardware resource utilization and enhances system flexibility. By enabling modules to be developed, tested, and deployed independently, the system achieves higher maintainability and scalability. These benefits are further amplified through decentralized architecture, which supports intellectual property protection while allowing architecture firms to build and manage their template libraries. This decentralization also ensures flexible access to design resources. For high-security environments, end-to-end encryption safeguards data integrity and privacy.

3.2.2. Real-time communication

ArchiWeb takes advantage of WebSocket and adopts an event-driven approach to establish a persistent, bidirectional communication between the client and the server. This setup facilitates real-time data transmission in a long-connection mode. Users must define event interfaces to invoke various algorithmic modules and transmit processing results to designated clients. In a fully developed application, a single client access typically involves real-time data interactions among multiple algorithmic modules.

In the SIMForms application, there are three primary algorithmic modules: plot division, mass generation, and Al renderer. These algorithms remain online to receive requests and return results in time. When the client is accessed via a browser, the interconnected algorithmic modules generate corresponding massing models based on user input parameters and pass them to the Al renderer for visualization (Fig. 5).

The algorithmic modules presented in this paper are currently connected to a central server. The integration of these multiple algorithmic modules is achieved by coordinating their interactions through this central server. This method supports deployment within a local area network (LAN), enhancing security.

3.2.3. Active modular network

In the ArchiWeb workspace, users generate an application token and at least one identity for each project. Each Identity instance corresponds to a specific algorithmic module, while the Client acts as a special type of Identity that identifies requests originating from the browser. The authenticated combination of an application token and identity defines an isolated namespace, enabling differentiated management across multiple users. Additionally, when an identity successfully connects to the system, it is assigned a unique Socket ID.

In this modular algorithmic network, namespaces support broadcast communication, while unique Socket IDs facilitate point-to-point data transmission. A user-defined Token often aggregates multiple algorithmic modules, with the functionality of complex applications arising from the collaborative interaction between distributed computing components. Figure 6 identifier uses real operation logs to demonstrate this hierarchical relationship, showing how active nodes can communicate simultaneously. This system supports high concurrency and low latency, enabling realtime interactions among multiple users.

3.3. Interactive design interface

The ArchiWeb platform features a user-friendly frontend interface paired with a robust Node.js backend. The frontend leverages technologies such as Vue, Vuetify, and Three.js to deliver an intuitive and interactive user experience. The backend is cloud-based and utilizes Koa for server side operations, while MongoDB handles real-time messaging via ArchiJSON and stores user data securely. This architecture ensures seamless communication and efficient management of design data across the entire platform.

ArchiWeb's architecture involves three primary roles: developers, users, and clients (Fig. 7).

* Developers are responsible for building and maintaining the core framework and platform. They collaborate on GitHub to define the overall architecture and provide technical support.

* Users, typically architects with programming experience, install the core libraries via npm and write custom code to develop web applications. They utilize ArchiWeb's tools and interfaces to tailor the platform to specific design requirements.

* Clients are the end-users of these applications. They focus on addressing real-world design challenges and transforming digital models into tangible, buildable structures.

It's important to note that an individual developer working on a generative design application may assume more than one of these roles simultaneously, depending on their involvement in both development and application use.

3.3.1. Web interface components

Open frontend scaffolding makes application development easy and efficient. Anyone can create their own ArchiWeb project by installing archiweb-core via npm to access ArchiWeb's core functionalities (Fig. 8). This allows users to develop their own interactive web applications that render 3D models directly in the browser and interact with algorithms within the algorithmic network. Using customizable settings, clients can manipulate and modify 3D architectural models, exploring and adjusting design options as needed. This enhanced interaction significantly increases the flexibility and adaptability of the design process.

ArchiWeb Core is composed of three main categories: Editors, Creators, and Viewers, each offering specialized tools to enhance architectural design workflows. Within the Editors, the Gumball tool enables users to translate, rotate, and scale objects while supporting switching between local and world coordinate systems, and the Box Selection tool allows for multi-object selection with capabilities for combining and splitting elements to manage complex structures. The Creators section includes the Geometry and Material Factory tool, which empowers users to create, modify, and delete basic geometric shapes. İn the Viewers category, the MultiCamera component provides flexible switching between perspective and orthographic views, along with viewpoint saving and smooth animation transitions. ArchiWeb also offers Environment settings that allow users to adjust ambient lighting and shadow mapping for improved visualization guality. Additionally, the Storage component maintains an operation history and supports automatic scene saving, while the Viewport component enables switching between 3D and 2D views to suit different design tasks and preferences.

ArchiWeb supports a variety of file formats for input and output, including DXF, IFC, OBJ, DAE, and FBX, enabling users to import and export architectural models and facilitate data exchange with other CAD software. Additionally, ArchiWeb recommends using a JSON-based database to store design data, allowing for persistent storage and efficient data management.

3.3.2. Integration with CAD/BIM and design tools

Given that ArchiWeb can exchange data with various AEC software and programming languages, users can create algorithmic modules and make full use of a wide range of algorithms and tools for enhanced design functionalities. For instance, users can utilize GIS functions in Java for urban planning and site analysis, perform discrete geometry processing of polygon meshes in C++, or employ machine learning and deep learning algorithms in Python to generate architectural designs. Additionally, ArchiWeb supports the use of Rhino Grasshopper for performancebased architectural designs.

We have developed a series of components based on the C# version of ArchiJSON. By leveraging multi-threaded components ArchiServer, grasshopper components can integrate into the broader algorithmic network. Figure 9 illustrates the process of messaging from the code editor on ArchiWeb Playground to Grasshopper. Playground represents an initial attempt at creating a configuration-free development ecosystem, offering instant browser-based accessibility while aiming to reduce complexity in development environments and coding prerequisites.

4. Applications and use cases

ArchiWeb integrates Al with generative design to offer a platform for early-stage design. It includes algorithmic modules such as field optimization, site division, building indicators, and mass generation, which can be reused across various design processes. This setup facilitates exploring solutions to a variety of design problems (Fig. 10). In this section, we introduce several applications developed and tested within ArchiWeb, including Archindex, ANYPlace, FLEXUrban and SnapRender. These applications have been utilized in undergraduate design instruction and workshops for over two years, showcasing different aspects of ArchiWeb's capabilities in early design stages and its effectiveness in addressing diverse design challenges.

4.1. Archindex: intelligent indexing of city blocks

Archindex is a city block search engine built on ArchiWeb. It provides urban cognitive intelligence to simplify the understanding and research of urban design in early stages by extracting multiple sources of urban data and establishing an urban spatial retrieval system. Based on a regularly updated spatial relationship database, Archindex collects block instance data and establishes vectorized indicators for form, function, and activity, which are used for retrieving and visualizing city blocks.

The morphology indicator was created by training an auto-encoder on a dataset of 39,793 city blocks, enabling the extraction of latent features and identification of urban typologies through vector analysis (Fig. 11). Users can interactively modify and search for similar city blocks via a browser interface, exploring cities based on indicators such as morphology, activity, functionality, floor area, and building densities (Fig. 12).

Archindex uses ArchiWeb as its frontend interface, connecting with Python and Java backends using ArchiJSON. User queries are first processed by a pre-trained neural network model on the Python side to extract image features, then integrated query data is passed to the Java side for database querying and similarity comparisons. Designed for distributed operation deployment, Archindex can establish indexes for any city, provide easy-to-use tools and innovative methods for urban morphology research.

4.2. ANYPlace: adaptive design for different contexts

The initial phase of architectural design often involves the systematic acquisition of site-specific data. ANYPlace, a web-based application built on the ArchiWeb platform and utilizing opensource geolocation datasets (Fig. 13), streamlines this process by retrieving geospatial data from OpenStreetMap (Haklay, 2010) and dynamically generating real-time 3D site models.

Designed to provide an intuitive 3D environment for urban design, ANYPlace allows users to input geographic coordinates (latitude and longitude) and parametrically define contextual boundaries. The platform then constructs 3D volumetric representations in real time, generates road networks based on topological tags from OpenStreetMap, and algorithmically subdivides zones into buildable plots.

The resulting outputs-including the 3D model and associated geospatial data-can be exported in ArchiJSON format for integration into downstream web applications that support advanced analytical workflows. These include pedestrian flow simulations and public space evaluations using algorithms such as A pathfinding and the Wave Function Collapse (WFC) algorithm, which synthesizes contextually responsive building configurations (Fig. 14).

Based on the data generated by ANYPlace, these computational methodologies enhance evidence-based design by simulating urban systems, iteratively testing spatial hypotheses, and generating visual analytics that inform decision-making processes.

4.3. FLEXUrban: generative urban design

FLEXUrban is a multi-functional web application for generative urban design that supports the entire work-flow-from site division and typology-based architectural generation to façade detailing (Fig. 15). Highly adaptable and flexible, FLEXUrban can intelligently generate site divisions based on the geographic and environmental context of a given area by utilizing tensor field analysis.

Building upon the generated site divisions, FLEXUrban offers a wide range of configuration options for defining functional urban zones, such as residential, commercial, and industrial areas. This ensures that the resulting designs meet practical needs while reflecting the diversity and complexity of modern urban planning. For each functional block, users can choose from several pre-defined building prototypes that automatically adapt to the specific conditions of the selected site, enabling optimal building layouts, forms, and detailed façade generation.

In response to complex urban design challenges, FLEXUrban accelerates the transition from conceptual ideation to actionable solutions, helping to shape urban environments that are more human-centered, responsive, and sustainable.

4.4. SnapRender: rendering with diffusers

SnapRender is a web-based architectural rendering tool that enables users to upload sketches or line drawings and instantly convert them into realistic architectural renderings with a single click within 10s (Fig. 16). Users can choose froma variety of rendering styles (such as Spring, Summer, Minimalist) and further refine the output by entering custom descriptors related to scene angles, styles, colors, and materials. A built-in library of pre-defined options and descriptors serves as a reference guide, helping ensure the quality, coherence, and consistency of the generated images (Fig. 17).

At its core, SnapRender leverages a Stable Diffusion model (Rombach et al., 2022) enhanced with ControlNet (Zhang et al., 2023), both running on a local server hosted in a hardware environment equipped with NVIDIA 2080 Ti GPUs. This high-performance configuration ensures efficient computational processing and fast rendering speeds. As a result, SnapRender is capable of handling complex rendering tasks swiftly, meeting users' immediate demands for high-quality visual outputs.

5. Conclusion and discussion

This paper proposes and implements ArchiWeb, an open algorithmic network platform designed to support earlystage application development and deep integration with AI. ArchiWeb achieves comprehensive integration of generative design algorithms throughout the entire design workflow by employing a modular algorithmic network and standard data protocol. This enables the construction of a generative design framework that facilitates interdisciplinary collaboration and supports the continuous evolution of design knowledge.

The contributions of ArchiWeb are reflected in three key aspects:

* As a universal platform for generative design: ArchiWeb tightly integrates various architectural design algorithms into the generative design process, offering an interactive computational design environment. This enables designers to easily access complex data flows and intelligent tools, lowering the technical barriers to advanced design exploration.

* Reshaping the role of architects: Through ArchiWeb, architects transition from being traditional design executors to becoming coordinators of algorithms and creativity. They can now embed original design concepts into parametric modeling and logical workflows, enabling more effective collaboration with technical experts and expanding the scope of architectural design.

* Building a design knowledge base: ArchiWeb systematically accumulates cross project practical experience through its modular node structure and collaborative work mechanisms. This results in a reusable and continuously evolving repository of design knowledge, enhancing both design efficiency and technological innovation capabilities.

While preliminary demos have validated the effectiveness of ArchiWeb, further efforts are needed to strengthen its integration into real-world design workflows-particularly through closer industry collaboration. Additionally, although ArchiWeb supports graphical programming and zero-code development paradigms, fully leveraging its potential still requires users at a certain level of programming proficiency. Future research directions include:

* Optimizing user experience, particularly in the early stages of the design process, to make the platform more intuitive and accessible;

* Developing Al-assisted modules, such as code generation powered by large language models, to enhance architects' ability to interact with and customize computational workflows;

* Strengthening data privacy and security protocols, especially for centralized server deployments, to ensure secure storage and transmission of sensitive design data.

In summary, ArchiWeb-through its open and scalable architecture-achieves deep technical integration of generative design and Al while also providing structured solutions for interdisciplinary collaboration and knowledge management at the practical level. With its combination of openness, modularity, and practical applicability, ArchiWeb has the potential to become foundational infrastructure in the architectural industry, facilitating a paradigm shift from experience-driven to rule- and data-driven architectural design practices.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This research was funded by the National Natural Science Foundation of China (No. 52378008) and Postgraduate Research Innovation Program of Jiangsu Province (No. KYCX22_0189).

References

References

Alexander, C., 1977. A Pattern Language. Oxford University Press.

Alexander, C., 1964. Notes on the Synthesis of Form, Later Pr. Edition. Harvard University Press, Cambridge, Mass.

Alfaiate, P.C., 2017. Luna moth: supporting creativity in the cloud. In: ACADIA 2017: DISCIPLINES & DISRUPTION [Proceedings of the 37th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), Cambridge, MA 2-4 November, 2017], Pp. 72-81.

Bhat, R.R., 1993. Communication in cooperative building design. In: CAAD Futures '93 [Conference Proceeding] (Pittsburgh/USA), 1993, Pp. 481-493.

buildingSMART International, 2020. Industry Foundation Classes (IFC) Schema Specifications: IFC Version 4.3.

Búhlmann, V., Hovestadt, L., Moosavi, У. (Eds.), 2015. Coding as Literacy: Metalithikum IV, first ed. Birkhauser, Basel.

Caetano, I., Santos, L., Leitão, A., 2020. Computational design in architecture: defining parametric, generative, and algorithmic design. Front. Architect. Res. 9, 287-300.

Campbell, M.I., Shea, K., 2014. Computational design synthesis. Al EDAM 28, 207-208.

Coenders, J., 2012. NetworkedDesign, next generation infrastructure for design modelling. In: Gengnagel, C., Kilian, A., Palz, N., Scheurer, F. (Eds.), Computational Design Modelling. Springer, Berlin, Heidelberg, pp. 39-46.

Coenders, J.L., 2021. Next generation parametric design. J. Int. Assoc. Shell Spatial Struct. 62, 153-166.

Eastman, C., 1968. Explorations of the Cognitive Processes in Design 103.

Guo, Z., Li, B., 2017. Evolutionary approach for spatial architecture layout design enhanced by an agent-based topology finding system. Front. Architect. Res. 6, 53-62.

Haklay, M., 2010. How good is volunteered geographical information? A comparative study of OpenStreetMap and ordnance survey datasets. Environ. Plann. Plann. Des. 37, 682-703.

Hovestadt, L., 1993. A4 digital building: extensive computer support for building design, construction, and management. In: CAAD Futures '93 [Conference Proceedings] (Pittsburgh/USA), 1993, Pp. 405-421.

Hovestadt, L., Hirschberg, U., Fritz, O. (Eds.), 2020. Atlas of Digital Architecture: Terminology, Concepts, Methods, Tools, Examples, Phenomena, first ed. Birkhauser, Boston.

Hua, H., 2016. Irregular architectural layout synthesis with graphical inputs. Autom. Construct. 72, 388-396.

Jabi, W., 1999. Integrating databases, objects and the world-wide web for collaboration in architectural design. https://orca. cardiff.ac.uk/27298/. (Accessed 1 July 2022).

Kalay, Y.E., 1999. Performance-based design. Autom. Construct. 8, 395-409.

Ledoux, H., Arroyo Ohori, K., Kumar, K., Dukai, B., Labetski, A., Vitalis, S., 2019. CityJSON: a compact and easy-to-use encoding of the CityGML data model. Open Geo-spatial Data Software Stand. 4, 4.

Mele, T.V., others, m., 2017. COMPAS: a Framework for Computational Research in Architecture and Structures.

Mitchell, W.J., 1977. Computer-Aided Architectural Design. Petrocelli/Charter.

Negroponte, N., 1976. Soft Architecture Machines, first ed. The MIT Press, Cambridge, Mass.

Poinet, P., Stefanescu, D., Papadonikolaki, E., 2021. Collaborative workflows and version control through open-source and distributed common data environment. Proceedings of the 18th International Conference on Computing in Civil and Building Engineering.

Poinet, P., Stefanescu, D., Papadonikolaki, E., 2020. SpeckleViz: a web-based interactive activity network diagram for AEC. In: SimAUD 2020: Proceedings of the 11th Annual Symposium on Simulation for Architecture and Urban Design (SimAUD). The Society for Modeling and Simulation International (SCS), pp. 419-428.

Radford, A., Narasimhan, K., 2018. Improving Language Understanding by Generative Pretraining, In:.

Rittel, H.W.J., Webber, M.M., 1973. Dilemmas in a general theory of planning. Policy Sci. 4, 155-169.

Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B., 2022. High-resolution image synthesis with latent diffusion models. In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, New Orleans, LA, USA, pp. 10674-10685.

Salingaros, N.A., 2010. Twelve Lectures on Architecture: Algorithmic Sustainable Design, first ed. Sustasis Press.

Schneider, S.N.R., 2010. Open architectural design: an approach to managing complexity and uncertainty in an open design process. In: Proceedings of the 15th International Conference on Computer Aided Architectural Design Research in Asia/Hong Kong 7-10 April 2010, pp. 629-638.

Steenson, M.W., 2017. Architectural Intelligence. The MIT Press.

Svetel, I., 1994. DDS-distributed design system: a paradigm for modeling conceptual architectural design. Advances in Computer-based Design Systems, (The International Institute for Advanced Studies in Systems Research and Cybernetics, Baden-Baden 1994, pp.79-88.

Svetel, I., Kosic, T., Pejanovic, M., 2018. Digital Vs. Traditional Design Process.

Wang, L., 2022. Workflow for applying optimization-based design exploration to early-stage architectural design - case study based on EvoMass. Int. J. Architect. Comput. 20, 41-60.

Wing, J.M., 2006. Computational thinking. Commun. ACM 49, 33-35.

Woodbury, R., Mohiuddin, A., Cichy, M., Mueller, V., 2017. Interactive design galleries: a general approach to interacting with design alternatives. Des. Stud. Parametric Des. Think. 52, 40-72.

Zhang, B., Mo, Y., Li, B., Wang, Y., Zhang, C., Shi, J., 2024. SIM-Forms: a web-based generative application fusing forms, metrics, and visuals for early-stage design. In: Gardner, Nicole, Herr, Christiane M., Wang, Likai, Toshiki, Hirano, Ahmad Khan, Sumbul (Eds.), ACCELERATED DESIGN-Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, Pp. 373-382.

Zhang, L., Rao, A., Agrawala, M., 2023. Adding Conditional Control to text-to-image Diffusion Models.

Footnote

* Corresponding author.

E-mail address: [email protected] (B. Li).

Peer review under the responsibility of Southeast University.