Track what you eat, drink, and do — Buddie connects the dots to your symptoms Multi-window temporal correlation analysis with LLM validation, HIPAA compliance, and 590+ tests
Just tell Buddie what you ate, how you feel, what medicines you took. Buddie remembers everything, spots patterns you'd never notice, and helps you understand your body.
Powered by Claude Haiku 4.5 for real-time conversational AI with tool calling. Every message is processed through structured tool calls (log_intake, log_symptom, ask_clarification, save_memory) — no regex parsing. Dynamic context injection includes recent intakes, active correlations, known side effects, and persistent patient memories capped at ~500 tokens.
Tell Buddie what you ate in plain English. It automatically looks up ingredients.
OpenFoodFacts + USDA FoodData Central APIs. 'Coca-Cola' → caffeine, phosphoric acid, high fructose corn syrup. Automatic dietary tagging: histamine, FODMAP, migraine triggers, allergens.
Log your medications and Buddie checks for known side effects.
OpenFDA + NIH RxNorm + SIDER database. 38 food-drug interactions checked per intake. Side effect frequency enrichment (common/uncommon/rare) for 24+ drugs.
Buddie spots connections between what you consume and how you feel.
5 time windows: acute (0-4h), standard (4-12h), extended (12-48h), cumulative 3-day, withdrawal (24-72h). Two-pass: deterministic statistics → Claude Sonnet 4.6 LLM validation. Confounder detection, ingredient deduplication, 85 mechanism pathways.
See your health patterns in easy-to-read charts and summaries.
Quantity parsing (2 cups → 474ml), severity parsing (7/10 → 7.0), tercile binning, gradient detection. Results stored in correlation metadata for pattern strength visualization.
Test your triggers with guided challenges. 'Try cutting coffee for 2 weeks.'
Claude Sonnet 4.6 generates 2-4 actionable hypotheses from high-confidence correlations. 14-day elimination challenges with 7-day baseline, auto-outcome determination (>30% improvement = success).
A living document about YOUR body — confirmed triggers, protective factors, questions for your doctor.
4 sections: confirmed triggers, under investigation, protective factors, doctor questions. Priority-sorted questions generated from patterns, dose-response, medications.
Connect Apple Health or Google Fit for automatic sleep, heart rate, and activity tracking.
HealthMetric model with batch import, summary stats, trend detection. Correlation engine enrichment via _enrich_correlation_with_health_context(). Unified SDK abstraction with graceful Expo Go fallback.
Your health data is encrypted and never shared without your permission.
Fernet field-level encryption, append-only audit trail, RBAC (patient/provider/admin/researcher), 6-year retention policy, GDPR deletion with cascade + anonymization, consent framework with explicit opt-in.
Share what you consumed in natural language.
Claude Haiku 4.5 extracts structured data via tool calls → ingredient lookup via OpenFoodFacts/OpenFDA → dietary tagging → interaction checking (38 critical interactions).
Report your symptoms with severity and timing.
Symptom logged with severity, duration, body area → triggers correlation refresh → temporal analysis begins.
AI analyzes correlations and builds your unique profile.
5-window temporal analysis → confounder detection → mechanism matching → LLM validation pass (Claude Sonnet 4.6) → spurious pattern filtering.
Receive personalized recommendations and track experiments.
Hypothesis generation → elimination challenges → dose-response analysis → body manual compilation → doctor-ready reports.
BodyBuddie generates a personalized health report with your confirmed triggers, active experiments, and curated questions for your next appointment.
Web-based clinical interface with generative UI. Claude Sonnet 4.6 generates interactive Chart.js visualizations inline via SSE-streamed tool calls. Providers can ask exploratory questions and receive bespoke data visualizations on demand. Architecture: FastAPI SSE endpoint → Claude streaming with show_widget tool → morphdom progressive rendering → bidirectional communication.
Your health data is yours. Everything is encrypted. We never sell your data. You can delete everything at any time.
Enterprise-grade security infrastructure designed for HIPAA and SOC 2 compliance. Every feature is built with privacy-first principles.
All sensitive health data is encrypted at rest and in transit.
Symmetric encryption (Fernet) on all PHI columns. Queryable fields (patient_id, dates, categories) remain unencrypted for efficient filtering.
Every action is logged and cannot be altered.
Immutable audit_logs table with timestamped entries. Action types: login, data_access, data_modify, data_delete, consent_change.
You control who sees your data.
4 role tiers: patient, provider, admin, researcher. Consent-gated data access. De-identification pipeline for research queries.
Your data is kept as long as required, then securely deleted.
Configurable retention windows. GDPR right-to-deletion with cascade deletion and full anonymization option for research datasets.
You explicitly opt-in to data usage.
consent_records table with policy versioning, grant/revoke timestamps, IP/user-agent tracking. No pre-checked boxes, no implied consent.
Protection against common attack vectors.
CSP, HSTS, X-Frame-Options headers. Pydantic schema validation on all inputs. SQL injection and XSS prevention throughout.
Contribute to health research safely.
HMAC-SHA256 patient ID hashing, date rounding, PII stripping. Cohort analytics with explicit opt-in consent. No re-identification possible.
Built to healthcare and data protection standards.
19 compliance tests. Audit logging, retention policy enforcement, deletion verification, consent workflows, security header validation.
Free to get started. No credit card required.