AGEWELL: Evidence-Based Longevity RAG Engine

"Age better with evidence."

Built by @lerenaminy.

Project Vision

Traditional health reports are static and siloed. Users receive blood work (ApoB, HbA1c, etc.) but lack an integrated framework to translate numbers into action. AgeWell acts as a Clinical Reasoning Engine, mimicking the logic of a senior longevity physician by cross-referencing individual biometrics against the latest peer-reviewed literature.

Technical Architecture

The project utilizes a Decoupled Architecture to ensure high scalability and modular maintenance:

• LLM Engine: Gemini 2.5 Flash (State-of-the-art 2026 Reasoning Model)
• Embedding Model: BGE-M3 (Multi-lingual, 1024-dimensional vectors for high-precision semantic retrieval)
• Vector Store: FAISS (Facebook AI Similarity Search)
• Frontend: Streamlit (Custom clinical UI with professional styling)
• Structure: Modular separation of concerns between UI (app.py) and Reasoning Engine (logic.py)

Knowledge Taxonomy (The 8 Longevity Pillars)

The AgeWell knowledge base is grounded in 45+ authoritative medical sources, categorized into eight critical domains:

1. Metabolism: Insulin sensitivity, metabolic flexibility, and HbA1c management.
2. Cardiovascular: Lipid profiles (ApoB/LDL), hypertension, and endothelial function.
3. Fitness: VO2 Max mortality correlations and sarcopenia prevention through resistance training.
4. Nutrition: Mediterranean protocols, Intermittent Fasting (TRF), and autophagy pathways.
5. Recovery: Circadian rhythm optimization, sleep hygiene, and HRV (Heart Rate Variability).
6. Biomarkers: Micronutrient density (Vitamin D/Omega-3) and chronic inflammation (CRP) monitoring.
7. Neurology: Cognitive reserve, stress management (Cortisol), and neuroplasticity.
8. Cellular Longevity: Biological age vs. chronological age, senescent cells, and mitochondrial health.

RAG Foundation Benchmarks

To ensure objective reasoning, the engine is embedded with five foundational clinical benchmarks:

• Body Composition: BMI ranges, body fat percentage standards, and Waist-to-Hip Ratio (WHR) predictors.
• Reference Ranges: Comprehensive Metabolic Panel (CMP) and lipid target values.
• Vitals: AHA/ACC Blood Pressure categories and resting heart rate (RHR) norms.
• Fitness Benchmarks: VO2 Max percentiles and grip strength standards for longevity.
• Optimal Zones: Defining optimal ranges for Vitamin D, B12, and Ferritin.

🔌 MCP Integration (Model Context Protocol)

AgeWell is now a fully functional MCP Server. You can connect this repository to Claude Desktop or Cursor to use the AgeWell clinical database as a live tool within your AI workflow.

• Tool: query_longevity_expert
• Capability: Returns structured 5-step clinical reports based on 1,100+ verified data segments.
• Transport: Stdio-based connection for local security.

Key Engineering Innovations

1) Severity Override Logic

Designed to solve AI distraction. When critical outliers (e.g., BMI over 30) are detected, the system triggers a priority-weighting override, forcing the AI to address urgent SHIFT protocols before addressing missing or secondary lab data.

2) Deterministic Output Sanitization

To maintain a clinical-grade user experience, I developed a multi-stage post-processing pipeline. It strips inconsistent AI formatting and injects sanitized HTML to ensure the professional report remains visually perfect.

3) Recursive Semantic Chunking

Optimized retrieval using 800-character chunks with a 150-character overlap. This strategy ensures that clinical context and causal logic within medical documents remain intact during vectorization.

Installation and Setup

Follow these four steps to run AgeWell locally:

1) Clone the repository and install dependencies

git clone https://github.com/yourusername/AgeWell.git
cd AgeWell
pip install -r requirements.txt

2) Configure environment variables (.env)

Create a file named .env in the root directory and add your API key:

GEMINI_API_KEY=your_actual_api_key_here

3) Initialize the clinical index (vectorization)

Place your medical documents in the /knowledge_base folder and run:

python build_index.py

4) Launch the application

Run the following command to start the web interface:

streamlit run app.py

Medical Disclaimer

This tool is for educational and informational purposes only. It does not constitute medical advice, diagnosis, or treatment. Always seek the advice of a qualified healthcare professional before making health decisions based on information provided by AgeWell.

Project Maintenance

• Author: Wei Ting Chang
• Stack: Python, FAISS, Streamlit, Google Generative AI

---