CarbonWise: An AI-Driven Framework for Carbon Footprint Estimation and Code Quality

CarbonWise presents a unified AI-driven framework for integrating software sustainability assessment into the software development lifecycle. The proposed platform combines GitHub repository analysis, software metrics, energy estimation, carbon-emission estimation, sustainability scoring, CI/CD analysis, dependency analysis, AI-assisted recommendations, dashboards, analytics, and automated reporting. The framework begins with repository-specific evidence obtained through GitHub APIs. Structural, dependency, and CI/CD characteristics are analyzed deterministically. These findings are subsequently used to estimate energy consumption and associated carbon emissions. A composite sustainability score combines carbon, CI/CD, dependency, and codebase indicators. A central feature of the architecture is the separation between deterministic analysis and generative AI. CarbonWise first verifies repository characteristics and stores structured findings before passing them to the AI recommendation layer. This design is intended to reduce generic recommendations and provide guidance that is grounded in repository-specific evidence. The system also provides developer-oriented dashboards and automated reports, allowing sustainability information to become part of normal software development activities. However, the current framework provides model-based estimates rather than direct physical measurements. The carbon and energy values therefore require appropriate interpretation, and future empirical validation is necessary to determine estimation accuracy and practical environmental impact. Future development involving real-time energy monitoring, cloud-provider data, regional carbon-intensity information, advanced static analysis, dependency verification, container analysis, Kubernetes integration, carbon-aware CI/CD, and historical sustainability tracking can further strengthen the framework. Overall, CarbonWise provides a structured approach for connecting software engineering artifacts with sustainability analysis and demonstrates how repository analytics and evidence-based artificial intelligence can support Green Software Engineering and Green DevOps practices.

[1]    O. Danushi, S. Forti, and J. Soldani, “Carbon-efficient software design and development: A systematic literature review, ” ACM Computing Surveys, 2025. [2]   S. M. K. Murthy, U. Airsang, K. Rajamani, and B. M. Schuller, “From code to carbon: Being energy efficient while coding, ” in Proc. Int. Conf. Sustainable & Resilient Futures (ICSRF), 2025. [3]     Verdecchia et al. discuss how software engineering practices can be considered from an environmental sustainability perspective, particularly in relation to the energy and resource demands of software systems. [4]  Castor presents an introductory approach for understanding and assessing the energy consumption associated with software execution. [5]     S. Podder, H. Date, and S. Murthy Green prompt engineering for sustainable generative AI, Environmental Science and Ecotechnology, vol. 30, 2026. [6]    J. Rashid, T. Mahmood, and M. W. Nisar A study on software metrics and its impact on software quality, Technical Journal, University of Engineering and Technology (UET) Taxila, vol. 24, no. 1, 2019. [7]  R. Baghdadi, M. Merouani, M. H. Leghettas, K. Benatchba, S. Amarasinghe, and T. Arbaoui, A deep learningbased cost model for automatic code optimization, in Proc. 4th MLSys Conference, 2021. [7]    R. Baghdadi, M. Merouani, M. H. Leghettas, K. Benatchba, S. Amarasinghe, and T. Arbaoui, A deep learningbased cost model for automatic code optimization, in Proc. 4th MLSys Conference, 2021. [8]   S. Ergasheva, I. Khomyakov, A. Kruglov, and G. Scucci,Metrics of energy consumption in software systems: A systematic literature review, IOP Conference Series: Earth and Environmental Science, 2020. [9]   Y. G. Kim, U. Gupta, A. McCrabb, Y. Son, V. Bertacco, D. Brooks, and C.-J. Wu, GreenScale: Carbon-aware systems for edge computing, arXiv preprint arXiv:2304.00404, 2023. [10]   A. Radovanović, R. Koningstein, I. Schneider, B. Chen, A. Duarte, B. Roy, D. Xiao, M. Haridasan, P. Hung, N. Care, S. Talukdar, E. Mullen, K. Smith, M. Cottman, and W. Cirne, Carbon-aware computing for datacenters.