Temporary project documentation
This page explains what Mahita is, how the surfaces connect, what is live today, and what still needs production infrastructure or clinical review.
Start here
Boundary
Live surfaces
Content systems
Each recipe has a stable slug, cuisine/region, meal type, tags, Jain compatibility, ingredients, instructions, nutrition estimate, attribution, publication status, and visibility. Users can save, plan, shop, and confirm a log from the same record.
Exercises are classified by primary/secondary muscle, kinetic chain, force vector, difficulty, equipment, and contraindication tags. A future reviewed program builder will compose them into versioned plans and calendar sessions.
Nutrition values are estimates and should carry a source and release version. USDA FoodData Central is the planned source-backed import path; branded data needs an API key and provenance record.
Roles and access
Owns personal preferences, favorites, pantry, plans, logs, and consent decisions.
Sees assigned relationships only. Clinician actions require stronger credentials, consent, scope, and audit review in production.
Manages eligibility and aggregate programs. Named health records never enter the corporate dashboard.
Production identity still needs managed OIDC, MFA, email verification, recovery, session/device controls, and SAML/SCIM only when contracted.
Data and API
The local adapter stores synthetic records in ignored data/mahita.sqlite. The production contract is in prisma/schema.prisma; migration is intentionally waiting for managed PostgreSQL and a reviewed secret store.
Write safety
Meal plans, meal logs, nutrition logs, completed workouts, programs, calendar events, and feedback accept an Idempotency-Key. Replaying the same key and body returns the original result; changing the body is rejected. Browser writes also require the Mahita CSRF token.
This is a local-adapter correctness control. Horizontal production scale still requires Redis/PostgreSQL-backed idempotency, managed identity, encrypted storage, and durable audit.
Competitive strategy
Leaders in food logging, recovery, coaching, practice management, and corporate wellness each go deep in one room. Mahita’s flagship advantage is the context graph between them: a preference becomes a reviewed recipe or movement choice, then a plan, confirmed action, explanation, and optional consented share.
Gujarati-forward, vegetarian-first, Jain-aware content with hard exclusions and visible reasons.
Personal data stays personal; practitioner and organization views are separate grants.
Recipes, groceries, programs, calendars, logs, and embeds are connected by stable IDs.
The full inventory and competitor matrix live in the repository’s docs/DEMO_SURFACE_AUDIT.md and docs/COMPETITIVE_GAP_ANALYSIS.md.
Flagship loop
The signed-in home now exposes an explainable context graph: preferences filter what is suitable, pantry items make it practical, plans make it actionable, and confirmed entries inform the next brief. Members can see the source of each action and what is deliberately not inferred.
The context graph is a planning aid—not a diagnosis, readiness score, or employer report.
Research notes
Scale roadmap
Care delivery blueprint
Scoped roster, versioned programs, tasks, check-ins, referrals, and member-controlled sharing.
Credentialed review queue, consented clinical context, approvals, safety routing, and auditable plan versions.
Education and clinician-reviewed questions first. No autonomous diagnosis, dose changes, or prescribing from Mahita.
The implementation sequence, legal gates, cadence model, wearable architecture, and test matrix are maintained in the repository care-delivery blueprint and execution plan. See the care implementation matrix and production hardening runbook in the repository.
Feedback and measurement
Product analytics, clinical audit, reliability telemetry, and enterprise aggregate releases have separate data, access, and retention rules.
Safety-sensitive experiments run in shadow mode first; member feedback, practitioner review, and accessibility checks are release inputs.
Real health data, medication workflows, wearable imports, and employer reporting remain gated by security, privacy, clinical, and legal approvals.