Confidente App — Overview
Food sensitivity tracking powered by structured meal plans and statistical analysis.
The Problem
Traditional elimination diets fail for three reasons:
- Compliance is the goal — one slip “ruins” the experiment
- No statistical rigor — pattern recognition is informal and unreliable
- Confounders are ignored — a bad night’s sleep looks like a food reaction
The Reframe
Confidente treats food sensitivity testing as a controlled experiment. Key insight: compliance is not the goal — logging is. Off-plan meals are additional data points, not failures.
Core Loop
Onboarding → suspect foods identified
↓
Meal plan generated (Latin square-inspired scheduling)
↓
User logs meals (planned or not) + daily symptoms + controls
↓
Statistical model correlates ingredient exposure to symptom outcomes
↓
Plain-language report: "You seem to react to X"
What Gets Built
See Tech Stack for implementation details.
Two Deliverables
1. TUI (Terminal UI) — POC Built with TUI for immediate use while the web app is in development. Rails models + PostgreSQL backend. Designed for two specific users (Chris + wife) to validate the concept with real data.
2. Rails PWA — Production App Rails 8.1 + Hotwire + Tailwind. Progressive web app → Hotwire Native (planned). The science is hidden behind a clean consumer UX.
The Science (Hidden from Users)
See ANOVA and Experimental Design — the full statistical methodology. See Sensitivity Categories — the food knowledge graph. See Confounder Control — how sleep, stress, etc. are handled. See Hypothesis Engine — how the app suggests new foods to test.
Design Specifications
The following specs define the behavioral contracts for the four core services. They describe what each service must do and why (with biochemical rationale), not how it’s implemented.
- Symptom Correlator — 24h trailing attribution, time-decay, weighted correlation, FDR correction
- Daily Control Quality Score — per-category quality scoring with flag-specific biochemical logic
- Hypothesis Engine — evidence-tiered pattern detection, negative hypotheses, washout awareness
- Meal Plan Generator — constraint satisfaction for exposure scheduling, washout enforcement, decorrelation
Roadmap
| Phase | Scope |
|---|---|
| v0.1 MVP | Onboarding, meal plans, logging, basic correlation display |
| v0.2 | Hypothesis engine, confounder weighting, stats layer |
| v0.3 | HealthKit/Google Health Connect, Hotwire Native |
| v0.4 | Aggregate data, population-level validation |
| Future | Lab integration (Food Science Kit + Mast Cell Test data) |