Cardia: A Smartphone-Only Framework Integrating Contact Photoplethysmography and Self-Reflective LangGraph Agent for Health Informatics

Cardia successfully shows the feasibility of an integrated, AI-powered healthcare assistant designed to bridge critical information gaps in accessible, preventive health management. By                  combining             smartphone-based             contact photoplethysmography (cPPG) biometric extraction with a robust Multi-Agent System (MAS), the application gives users actionable, evidence-based insights across physiological, nutritional, and psychological domains.  The system’s architecture, highlights self-reflective safety loops and Retrieval-Augmented Generation (RAG), successfully mitigating the inherent risks of generative AI while ensuring data privacy. Empirical evaluation proved the reliability of the cPPG pipeline for the use case and the effectiveness of self reflecting agents in stopping propagation of risky hallucinated advice. Cardia establishes a scalable ecosystem that empowers users with personalized, safe, and proactive health support. Future iterations will focus on expanding the verifiable knowledge base and optimizing edge performance to further lower barriers to effective, intelligent healthcare, building the way for sustained, long-term well-being.

[1]   S. Canali, V. Schiaffonati, and A. Aliverti, "Challenges and recommendations for wearable devices in digital health: Data quality, interoperability, health equity, fairness," PLOS Digit. Health, vol. 1, no. 10, Art. no. e0000104, Oct. 2022, doi: 10.1371/journal.pdig.0000104. [2]   M. Moya-Ramon et al., "Validity and reliability of different smartphones applications to measure HRV during short and ultra-short measurements in elite athletes," Comput. Methods Programs Biomed., vol. 217, 106696, 2022, doi: 10.1016/j.cmpb.2022.106696. [3]    M. J. Bautista et al., "Clinical applications of contactless photoplethysmography for monitoring in adults: A systematic review and meta-analysis," J. Clin. Transl. Sci., vol. 7, no. 1, e129, 2023, doi: 10.1017/cts.2023.547. [4]  B. J. Singstad et al., "Estimation of Heart Rate Variability from Finger Photoplethysmography During Rest, Mild Exercise and Mild Mental Stress," J. Electr. Bioimpedance, vol. 12, no. 1, pp. 89–102, Dec. 2021, doi: 10.2478/joeb-2021-0012. [5]M. Elgendi et al., "The use of photoplethysmography for assessing hypertension," NPJ Digit. Med., vol. 2, 60, Jun. 2019, doi: 10.1038/s41746-019-0136-7. [6]     S. M. Biju, "Implementing Multi-agent Systems Using LangGraph: A Comprehensive Study," in Recent Advancements in Artificial Intelligence (ICRAAI 2025), R. Nayak, N. Mittal, A.Khunteta, and M. Kumar, Eds., Lecture Notes in Networks and Systems, vol. 1468. Singapore: Springer, 2026, doi: 10.1007/978-981-96-7760-3_19. [7] J. Jang and W.-S. Li, "AU-RAG: Agent-based Universal Retrieval Augmented Generation," in Proc. 2024 Annu. Int. ACM SIGIR Conf. Res. Develop. Inf. Retrieval Asia Pac. Region (SIGIR-AP), 2024, pp. 2–11, doi: 10.1145/3673791.3698416. [8]  J. He, C. Treude, and D. Lo, "LLM-Based Multi-Agent Systems for Software Engineering: Literature Review, Vision, and the Road Ahead," ACM Trans. Softw. Eng. Methodol., vol. 34, no. 5, 124, May 2025, doi: 10.1145/3712003. [9] Zhang Lei, "LangGraph vs LangChain: Which Framework Is Best for AI Agents?," MetaVision Journal of Multidisciplinary Studies, vol. 1, no. 4, pp. 16–26, 2024. [10]J. Spieser et al., "A Review of Multi-Agent AI Systems for Biological and Clinical Data Analysis," Methods Protoc., vol. 9, no. 2, 33, Feb. 2026, doi: 10.3390/mps9020033.