VitalWeave

Multi-signal health intelligence — woven from your biometrics by AI.

VitalWeave is a Streamlit app that takes your wearable data (Luna Ring / NoiseFit), mood logs, food diary, and blood reports, then uses Claude to find cross-signal patterns, forecast health risks, and answer natural-language questions about your data.

---

What it does

Tab	What you get
💍 Luna Signals	HRV, sleep score, readiness, SpO₂, skin temp — charted across every day
🧬 The Tapestry	Multi-signal timeline chart with HRV reference lines calibrated to your age/sex
🔗 Correlations	AI-detected cross-signal patterns (e.g. "low HRV → worse mood next day")
📈 Risk Forecasts	30/90/180-day risk cards with one concrete action per domain
🧬 Health Master	Personalised deficiency risks, biomarker targets, cycle guide (female), test recs
💬 Ask VitalWeave	Chat with Claude about your data — ask anything, get answers grounded in real values

---

Quick start

1. Clone and install

git clone <repo-url>
cd vitalweave
pip install -r requirements.txt

2. Set your API key

Create a .env file in the vitalweave/ directory:

ANTHROPIC_API_KEY=sk-ant-...

3. (Optional) Generate demo sessions

Pre-builds 4 realistic demo sessions (19M / 24F / 35M / 45F) with full analysis — so you can demo without waiting for Claude to run:

python create_demo_sessions.py

4. Run the app

streamlit run app.py

---

Loading data

Option A — Demo sessions (instant)

Click any session in the Sessions dropdown in the sidebar and hit ▶ Load. Full analysis loads in under a second.

Option B — Luna Ring (NoiseFit export)

Use the 💍 Load Luna Demo Data button in the sidebar, or upload your own NoiseFit .json export.

Option C — Upload your own files

Signal	Format
Sleep	.json or .csv
Mood	.csv (date + mood_score columns)
Food	.json (date + nutrient fields)
Blood report	.txt or .pdf

---

Running the analysis

1. Load at least sleep + mood data
2. Set your health profile (age + biological sex) in the sidebar — this calibrates every chart and Claude prompt to you specifically
3. Click 🧬 Weave My Health Story — takes ~15–30 seconds (uses Claude Haiku)
4. Session auto-saves after every weave run

---

Project structure

vitalweave/
├── app.py                  # Streamlit UI — all tabs, sidebar, session state
├── weave_engine.py         # Claude API — main health analysis (JSON schema)
├── chat_engine.py          # Claude API — streaming chat with full data context
├── visualiser.py           # Plotly charts — tapestry, correlation cards, risk gauges
├── profile_engine.py       # Static health knowledge — deficiency risks, HRV targets, biomarker ranges
├── data_parser.py          # File parsers — Luna Ring JSON, CSV, PDF blood reports
├── session_store.py        # Local JSON session persistence (save/load/delete)
├── ethics.py               # Ethics panel + doctor PDF export
├── create_demo_sessions.py # One-time script to pre-build 4 demo sessions
├── sample_data/            # Sample files for instant demo loading
└── sessions/               # Auto-created — saved session JSON files

---

Demo profiles

Profile	Pattern baked in
19M	Exam week stress (days 9–13, 23–27) — HRV crashes, sleep collapses
24F	28-day menstrual cycle — skin temp +0.45°C luteal, HRV dips in follicular
35M	Work-week HRV cycle — gradual burnout trend across 30 days
45F	Perimenopause — hot flash nights (8 days), HRV 14–40ms, temp spikes >0.9°C

---

Environment variables

Variable	Required	Description
ANTHROPIC_API_KEY	Yes	Your Anthropic API key

---

Tech stack

• Frontend — Streamlit 1.32+, Plotly, custom CSS (Inter + IBM Plex Mono)
• AI — Anthropic Claude (claude-haiku-4-5 for analysis, claude-sonnet-4-6 for chat)
• Storage — Local JSON files in sessions/ (zero external dependencies)
• PDF export — ReportLab + PyMuPDF