OSYRIO as a Traceable System for Personalized Wellness Sessions
Working paper · Released online November 4, 2026 · Not peer reviewed
Abstract
Background. OSYRIO is described as a digital setting for guided sessions, personalized text, audio and reports, with Connie as a guide.
Objective. To define a record that makes personalization, output provenance and user interpretation examinable.
Approach. Synthesis of the founding dossier and related product-design material with a proposed technical audit and usability evaluation.
Evidence status. Recovered descriptions and development history are not an audit of a specific software release. No efficacy or biological inference is established.
Conclusion. Versioned inputs, generation rules and matching outputs are a testable foundation for evaluating personalization without treating generated content as measurement.
Keywords: artificial intelligence; personalization; digital wellness; traceability; usability; generated audio
Introduction
OSYRIO seeks continuity between a person's intention, a guided experience and later reflection. The recovered dossier describes Connie guiding intention, personalization, duration and review, with generated text, audio and a report as intended outputs [1]. This description supplies a candidate workflow rather than certification that every function is operational in a current release. The research question concerns whether those outputs are coherent, traceable and correctly understood.
System boundaries and source basis
OSYRIO and Neuroakashic Connect remain distinct applications even when parts of their infrastructure or conceptual vocabulary are shared. The related Expandia Intelligence product requirements document describes authorized data, modular workspaces and permissions [2]. Those are useful design antecedents; its business-creation functions are not automatically OSYRIO features. Similarly, Expandia as a conceptual organizer differs from an implemented algorithm. A manuscript about the software must identify the exact release, tested workflow and available configuration instead of combining every historical proposal into a deployed system.
Proposed session provenance
For each test session, retain an authorized input record, application version, generation configuration, model identifier when available, prompt-template version, language, timestamps, selected duration and output identifiers. Link the final audio and report to the text from which they were produced. Record retries, failures and revisions. A session marked complete should meet declared output criteria rather than merely confirm that generation was requested. Source labels should persist: an input reported by the person stays reported, a computed value stays inferred, and an untested explanatory statement stays hypothetical.
Technical evaluation
Use predefined test cases to check whether generated text, audio and PDF agree; whether the requested language is maintained; whether unavailable content is distinguished from complete content; and whether authorized users can access only their own records. Include incomplete generation, repeated requests and version changes. Review outputs for unsupported diagnostic language and for assertions about biology that were not supplied by an instrument. WHO guidance offers governance context on generative AI, transparency and oversight [3]; it does not validate this application or make it a clinical system.
User evaluation and outcomes
Usability work should ask whether a person can finish a session, retrieve the report and explain the difference between a reflective suggestion and a measurement. Candidate outcomes include task completion, errors, time burden and comprehension. Continued use or satisfaction is a different question from a change in wellbeing. A later outcome study would need a defined population, comparator and prespecified measures. Repeated AI generation can vary, so any evaluation must retain enough version information to identify what participants actually received. No accuracy rate, retention statistic or benefit estimate is supplied here.
Discussion and conclusion
Personalization can make content relevant without establishing that the system reads unsupplied bodily states. Terms such as scan, map and triad require clear descriptions of their inputs and computation. A generated narrative is not neural imaging, an anatomical assessment or independent evidence of its own explanation. The immediate contribution is a traceability specification that can be audited. Its limitations are the absence of a release-specific audit, no reviewed user dataset and reliance on recovered design descriptions.
Authorship and disclosures
Ana Silvia Lara develops the concepts, methods and products discussed and is associated with the publisher. No participant dataset was analyzed for this draft. Proposed studies have not been conducted through this work. The author is responsible for the final work. This working paper has not undergone independent peer review.
The inaugural journal issue is planned for March 2027. This working paper was released separately in November 2026. No ISSN, DOI, indexing status or institutional endorsement is claimed.
References
- Lara AS. Neuroconsciousness Journal Editorial March 2027. English design and manuscript review proof. Version 2; November 2026. Unpublished development dossier; relevant article identified in text.
- Lara AS. Expandia Intelligence PRD v1.0. July 2026. Confidential product requirements document. Internal design source; proposed capabilities are not an implementation audit.
- World Health Organization. Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models. Geneva: WHO; 2024. https://www.who.int/publications/i/item/9789240084759