Codex VITAI is a privacy-first human-machine interface running on ATRIOM UI — a desktop-like environment designed specifically for medical literature research.
Research tool only. Not a medical device. v3.0 pre-release — All local AI outputs should be verified against primary sources.
- ATRIOM UI: A robust desktop-like environment featuring draggable windows, a persistent dock, and seamless multi-tasking.
- Multi-Model Local AI: Supports 10+ local language models, including BART, SmolLM2, Qwen2.5, and Llama 3.2.
- Multi-Source Data Aggregation: Instantly query and fetch research from ArXiv, EuropePMC, OpenAlex, and PubMed.
- Knowledge Dissimination: Ready-to-share posts for social media channels.
- Digital Networking: Look up authors over various digital and social media networks.
- Integrated Citation Metrics: Live analytical badges from Dimensions, Altmetric, PlumX, and Scite.
- Robust Export Engine: Export selected studies and AI-generated summaries to JSON, CSV, or PDF seamlessly.
- Research Diaries: Export summaries and selected outputs into various formats, for reference management or documentation formats.
- Progressive Web App (PWA): Fully offline-capable following its initial cache, utilizing internal system fonts to drastically enhance user privacy.
- Initialize Consent: Complete the initial privacy consent dialogue and select your mandatory regulatory framework.
- Configure Display: Choose your desired display options, including abstract limits, full-text preferences, and AI Summary toggles.
- Query Data: Set your article count limit and initiate the "Search Publications" workflow.
- Explore & Analyze:
- Visualize AI Summaries alongside automatically calculated trust-level indicators.
- Utilize the built-in AI Chat for direct article Q&A operations.
- Filter and aggregate literature by keywords, publication dates, and open-access availability.
- Hallucination Detection: Built-in logic that flags assumed or hallucinated output.
- Medical Claim Warnings: Triggers warnings when the AI detects absolute medical claims or unauthorized medication advice.
- Emergency Safety: Immediately fires alerts if dangerous dismissals occur in model outputs.
- PII Protection: Actively detects and filters Personally Identifiable Information (PII) including emails, phone numbers ...etc.
- Hardware-Adaptive Inference: Automatic model selection and limits dynamically adjusted based on detected GPU, RAM, and CPU telemetry.
- Edge/Local Processing: 100% of AI processing and summarization runs client-side directly within the browser ecosystem.
- Modular Regulation Engine: Hardcoded compliance parameters adapting seamlessly to GDPR, HIPAA, CCPA, or a standard Baseline mode.
- Audit Logging: Fully transparent local IndexedDB action persistence.
- Session Security: Strong HTTPS validation enforcing strict content-security integrity checks.