Avasthi: Revolutionizing the Current Mental Healthcare System

Abstract

Avasthi is an AI-powered mental health platform that offers per- sonalised stress management solutions. Users are classified as having low, medium, or high stress levels using the Perceived Stress Scale. Low and medium stress users interact with an AI chatbot based on Retrieval-Augmented Generation (RAG) for context-aware assistance and tailored recommendations, while high-stress users are connected to licensed psychologists through an integrated consultation system. Avasthi employs weighted vector fusion to combine behavioural cues explicit feedback, con- versational intent, and long-term preferences into a unified user profile. A knowledge graph–based recommendation engine uses deep contrastive learning (DCL) to create and compare activity embeddings via cosine similarity to rank optimal interventions. The platform also includes an AI fitness assistant that guides users through simple stress-bursting exercises and a diet module that provides nutri- tional guidance based on user preferences and requirements. Avasthi ensures scalable, transparent, and low latency interactions for holistic mental wellness support.

Introduction

Avasthi is an AI-driven, scalable mental wellness platform designed to provide personalized support for individuals experiencing stress, anxiety, or depression. Users first complete a Perceived Stress Scale (PSS) assessment to categorize stress levels (low, medium, high). Low- and medium-stress users interact with a Retrieval-Augmented Generation (RAG) chatbot, which delivers empathetic, context-aware guidance, asks clarifying questions, and draws from a curated mental-health knowledge base.

The system integrates a knowledge-graph–driven recommendation engine for personalized activities (meditation, journaling, mindfulness, breathing exercises), diet plans optimized for emotional wellness, and an AI-powered fitness assistant that guides stress-relief exercises using computer vision and biomechanical analysis. High-stress users are safely escalated to licensed psychologists through an online consultation module.

Technically, Avasthi uses FastAPI backend, React frontend, PostgreSQL for relational storage, and Pinecone for vector-based semantic retrieval, alongside multiple deep-learning models for stress classification, sentiment analysis, context-aware dialogue, and activity recommendation. The platform continuously adapts to user behavior and feedback, ensuring low-latency, personalized, and privacy-compliant support.

Overall, Avasthi represents a comprehensive, end-to-end mental wellness ecosystem, combining assessment, AI-driven conversation, personalized interventions, fitness and diet guidance, and professional referral to deliver accessible and contextually relevant mental-health care.

Conclusion

Avasthi is a comprehensive and scalable mental health ser- vice that combines accessible psychological care with cutting- edge AI technologies. In order to provide highly individ- ualised interventions, the platform integrates a RAG-based conversational assistant, knowledge graph-driven personalised recommendations, AI-powered fitness and mood regulation modules, and a scientifically based diet recommendation sys- tem that uses deep learning and the K-Clique algorithm. Stress reduction and emotional management are reinforced by the AI fitness assistant’s guided physical activity and posture-aware workout support. High-stress customers are given priority by Avasthi, which links them with qualified experts while continuously offering AI-driven advice to others. Avasthi, which is positioned as a revolutionary solution for improving mental well-being worldwide, successfully connects profes- sional mental healthcare with digital self-help tools thanks to its safe, low-latency infrastructure.

References

[1] L. Zhang, J. Zhan, V. K. W. Wan, and Y. Wang, “Designing and evalu- ating online health consultation interfaces: A perspective of physician- patient power asymmetry,” IEEE Access, vol. 12, pp. 124 111–124 127, 2024. [2] S. Sai, A. Gaur, R. Sai, V. Chamola, M. Guizani, and J. J. P. C. Ro- drigues, “Generative ai for transformative healthcare: A comprehensive study of emerging models, applications, case studies, and limitations,” IEEE Access, vol. 12, pp. 31 078–31 106, 2024. [3] B. Zhang, Z. Yan, X. Song, Y. Liao, and P. Li, “Influence of free consultation services on patients’ willingness to pay in online medical platforms,” International Journal of Crowd Science, vol. 8, no. 1, pp. 28–37, 2024. [4] R. Ren, M. Zapata, J. W. Castro, O. Dieste, and S. T. Acun˜a, “Exper- imentation for chatbot usability evaluation: A secondary study,” IEEE Access, vol. 10, pp. 12 430–12 464, 2022. [5] S. Elmajali and I. Ahmad, “Toward early detection of depression: Detect- ing depression symptoms in arabic tweets using pretrained transformers,” IEEE Access, vol. 12, pp. 88 134–88 145, 2024. [6] Y. B. Sree, A. Sathvik, D. S. Hema Akshit, O. Kumar, and B. S. P. Rao, “Retrieval-augmented generation based large language model chatbot for improving diagnosis for physical and mental health,” in Proceedings of the 2024 6th International Conference on Electrical, Control and Instrumentation Engineering (ICECIE), Pattaya, Thailand, 2024, pp. 1– 8. [7] R. Patil, S. Boit, V. Gudivada, and J. Nandigam, “A survey of text representation and embedding techniques in nlp,” IEEE Access, vol. 11, pp. 36 120–36 146, 2023. [8] M. Amin Kuhail, M. Bahja, O. Al-Shamaileh, J. Thomas, A. Alkazemi, and J. Negreiros, “Assessing the impact of chatbot-human personality congruence on user behavior: A chatbot-based advising system case,” IEEE Access, vol. 12, pp. 71 761–71 782, 2024. [9] A. N. Ngaffo, W. E. Ayeb, and Z. Choukair, “A bayesian inference based hybrid recommender system,” IEEE Access, vol. 8, pp. 101 682–101 701, 2020. [10] C. Arihta, D. Fujiono, Veronica, J. V. Moniaga, and I. D. Rianto, “A re- view of gamification related to mental health treatment,” in Proceedings of the 2022 3rd International Conference on Electrical Engineering and Informatics (ICon EEI), Pekanbaru, Indonesia, 2022, pp. 172–176. [11] F. Wu, G. Zhao, T. Li, J. Shen, and X. Qian, “Improving conversational recommendation system through personalized preference modeling and knowledge graph,” IEEE Transactions on Knowledge and Data Engi- neering, 2024. [12] D. Su, B. Fan, Z. Zhang, H. Fu, and Z. Qin, “Dcl: Diversified graph recommendation with contrastive learning,” IEEE Transactions on Com- putational Social Systems, vol. 11, no. 3, pp. 4114–4126, 2024. [13] L. Guo et al., “Knowledge graph-based personalized multitask enhanced recommendation,” IEEE Transactions on Computational Social Systems, 2024. [14] C. Fu, X. Pan, J. Wu, J. Cai, Z. Huang, F. van Harmelen, W. Zhao, X. Jiang, and T. He, “Kg4nh: A comprehensive knowledge graph for question answering in dietary nutrition and human health,” IEEE Journal of Biomedical and Health Informatics, 2023. [15] X. Li, “Design of a web content personalized recommendation system based on collaborative filtering improved by combining k-means and lightgbm,” Journal of Web Engineering, vol. 24, no. 2, pp. 267–290, 2025. [16] W. Lei, X. He, M. de Rijke, and T.-S. Chua, “Conversational rec- ommendation: Formulation, methods, and evaluation,” in Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, 2020, pp. 2425–2428. [17] D. Jannach, A. Manzoor, W. Cai, and L. Chen, “A survey on conversa- tional recommender systems,” ACM Computing Surveys, vol. 54, no. 5, pp. 1–36, 2021. [18] W. Lei et al., “Estimation-action-reflection: Towards deep interaction between conversational and recommender systems,” in Proceedings of the 13th International Conference on Web Search and Data Mining, 2020, pp. 304–312. [19] Q. Chen et al., “Towards knowledge-based recommender dialog system,” in Proceedings of the Conference on Empirical Methods in Natural Language Processing and the International Joint Conference on Natural Language Processing, 2019, pp. 1803–1813. [20] K. Zhou, W. X. Zhao, S. Bian, Y. Zhou, J.-R. Wen, and J. Yu, “Improving conversational recommender systems via knowledge graph- based semantic fusion,” in Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2020, pp. 1006–1014. [21] Y. Lu et al., “Revcore: Review-augmented conversational recommen- dation,” in Proceedings of the Annual Meeting of the Association for Computational Linguistics, 2021, pp. 1161–1173. [22] C.-M. Wong et al., “Improving conversational recommender system by pretraining billion-scale knowledge graph,” in Proceedings of the IEEE 37th International Conference on Data Engineering, 2021, pp. 2607– 2612. [23] G. Penha and C. Hauff, “What does bert know about books, movies and music? probing bert for conversational recommendation,” in Proceedings of the 14th ACM Conference on Recommender Systems, 2020, pp. 388– 397. [24] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” in Pro- ceedings of the Annual Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2019, pp. 4171–4186. [25] Y. Zhang et al., “Dialogpt: Large-scale generative pre-training for con- versational response generation,” in Proceedings of the Annual Meeting of the Association for Computational Linguistics, 2020, pp. 270–278.

Copyright

Copyright © 2026 Dr. Jisha P Abraham, Prof. Smitha Paulose, Acqulene Sara Thomas, Shraddha Sasikumar, Amritha M. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.