Split Index

Premium analytics platform for hybrid athletes. Objective fitness scoring that updates after every workout.

Split Index = 50% Endurance + 50% Strength

Tech Stack

• Next.js 16 (App Router, TypeScript)
• Tailwind CSS 4 + Framer Motion
• Supabase (Auth, PostgreSQL, RLS)
• Stripe (Subscriptions)
• OpenAI (AI Coach)
• Recharts (Analytics)

Getting Started

1. Clone and install

cd split-index
npm install

2. Configure environment

Copy .env.example to .env.local and fill in your keys:

cp .env.example .env.local

3. Set up Supabase

1. Create a project at supabase.com

2. Run the migrations in filename order — supabase db push does this for you, or paste each into the SQL editor in turn. Several later migrations ALTER tables that earlier ones create, so the order is load-bearing.

   #	File	What it adds
   001	001_initial_schema.sql	Core schema: profiles, activities, gym exercises, workout scores, the handle_new_user signup trigger
   002	002_scoring_reference_and_leaderboards.sql	Sports reference, strength scores, leaderboards
   002b	002b_apply_missing.sql	Backfill for databases created between 001 and 002
   003	003_integrations.sql	OAuth connections and import jobs
   004	004_split_weighting.sql	User-configurable Split Index endurance/strength weights
   005	005_session_templates.sql	Session templates + file activity source for GPX/TCX import
   006	006_fix_handle_new_user_trigger.sql	Signup trigger fix
   007	007_signup_trigger_bulletproof.sql	Signup trigger hardening — see the onboarding troubleshooting below
   008	008_timezone.sql	Per-user timezone
   009	009_gym_set_details.sql	Per-set weight/reps breakdown (set_details)
   010	010_avatars_storage.sql	Avatar storage bucket and policies
   011	011_predicted_benchmarks.sql	Memory-based cardio benchmark predictions
   012	012_public_read_strength_scores.sql	Public read on strength reference scores
   013	013_predicted_benchmark_quality.sql	Prediction confidence/quality
   014	014_personalized_hr.sql	resting_hr for personalised Karvonen zones
   015	015_interval_fartlek_scoring.sql	Interval and fartlek segment fields
   016	016_date_of_birth.sql	Date of birth (age-graded scoring, HR max)
   017	017_pricing_skus_and_motivation.sql	Pricing SKUs and motivation copy
   018	018_race_prediction_riegel_k.sql	Per-user Riegel exponent on predictions
   019	019_latest_strength_scores_per_exercise.sql	Latest strength score per exercise
   020	020_friend_duels.sql	Friend duels
   021	021_outdoor_cycling.sql	Outdoor cycling support
   022	022_outdoor_cycling_sports_row.sql	Outdoor cycling sports reference row
   023	023_duel_speed_strength_metrics.sql	Duel speed/strength metrics
   024	024_squads.sql	Squads
   025	025_hybrid_athlete_reports.sql	Hybrid athlete reports
   026	026_native_billing_source.sql	Native (App Store / Play) billing source
   027	027_gps_tracking.sql	Background GPS run tracking, is_partial_track
   028	028_exercise_attachment.sql	Exercise attachment/implement
   029	029_planned_races.sql	Planned races
   030	030_planned_race_elevation_source.sql	Race elevation source
   031	031_social_activity_feed.sql	Social activity feed
   032	032_share_activities_default_true.sql	Default activity sharing to on
   033	033_training_goals.sql	Training goals
   034	034_training_goal_target_date.sql	Goal target dates (tapering, feasibility)
   035	035_training_goal_custom_distance.sql	Custom goal distances
   036	036_training_goal_target_reps.sql	Rep-based strength goals
   037	037_planned_race_known_elevation_source.sql	Known race elevation profiles
   038	038_training_goal_progress.sql	Daily goal-progress snapshots for trend estimates
   039	039_hpe_athlete_profile.sql	Hybrid Plan Engine: diagnostic runs, findings, plans, sessions
   040	040_hpe_monitoring_and_rollout.sql	HPE monitoring, feature flags (kill switch), adherence and injury reporting
   041	041_admin_roles_and_fleet_review.sql	admin_users, the fleet-review gate and the rollout audit log
   042	042_hpe_intake.sql	HPE athlete intake: safety screen answers, goal, availability and preferences
   043	043_hpe_intake_plan_flexibility.sql	Plan flexibility on the intake
   044	044_hpe_training_split.sql	Preferred training split
   045	045_scoring_basis.sql	Explicit scoring basis for athletes who are not male or female
   046	046_hpe_safety_capped_outcome.sql	The health screen caps plans instead of refusing them
   047	047_hpe_intake_overrides.sql	Manual overrides for engine-proposed 1RM and heart rates
   048	048_hpe_intake_section_regroup.sql	Remaps stored intake progress onto the regrouped sections
   049	049_private_account_visibility.sql	"Private account" switch and the activity-visibility predicate behind it

   Skipping the tail of this list breaks the social feed. 031 adds profiles.share_activities_with_friends and the activities policy that lets a friend's workout be read at all; 049 hardens both. Without them the Settings privacy switch reports it cannot load and the feed stays permanently empty, because RLS returns zero rows rather than an error. Run the whole list.

   After 041: the Hybrid Plan Engine ships disabled at 0% rollout and grants no admin roles. admin_users has no INSERT policy by design, so the first operator has to be granted directly:

   insert into admin_users (user_id, role, note)
   values ((select id from auth.users where email = '[email protected]'), 'operator', 'Initial grant');
   

   That account can then open /admin/hpe-fleet and work the rollout from there. SUPABASE_SERVICE_ROLE_KEY must be set in the deployed environment or the fleet view fails closed and 404s for everyone.

3. Enable Email and Google OAuth in Authentication → Providers

4. Add your site URL to redirect allowlist: http://localhost:3000/auth/callback

Onboarding “Could not save your profile”

If onboarding fails at the final step, the database is usually missing migrations or the signup trigger:

1. In the Supabase SQL editor, run 001_initial_schema.sql then 002_scoring_reference_and_leaderboards.sql (in order).
2. 001 creates the profiles table and handle_new_user trigger (auto-creates a profile row on OAuth signup).
3. 002 adds a client-side INSERT policy on profiles so upserts work when the trigger did not run (e.g. user signed up before migrations were applied).

In development, the onboarding error message includes the Supabase error text to make this obvious (e.g. relation "profiles" does not exist or RLS violations). The app also calls /api/profile/ensure before saving, which creates a missing profile row via the service role when SUPABASE_SERVICE_ROLE_KEY is set.

Signup returns “Something went wrong on our side” (HTTP 500)

Two common causes after migrations are applied:

1. Database trigger — handle_new_user fails during INSERT INTO auth.users (rolls back signup). Run supabase/diagnostics/signup_full_diagnostic.sql in the SQL Editor; section D3 prints the exact Postgres error. Fix: re-run 007_signup_trigger_bulletproof.sql.
2. SMTP / confirmation email — user row is created but Auth returns 500 when sending the confirm email. Check Dashboard → Logs → Auth for gomail or Error sending confirmation email. Fix SMTP (host, port, TLS, sender domain) or temporarily disable Confirm email under Authentication → Providers → Email to isolate.

Before retrying with the same email, delete any stuck row from auth.users (see section F in the diagnostic SQL). Add https://splitindex.co.uk/auth/callback to Authentication → URL Configuration → Redirect URLs.

Required in .env.local for checkout to work:

• STRIPE_SECRET_KEY — secret key from Stripe Dashboard → Developers → API keys (use sk_test_… locally)
• STRIPE_PRICE_ID — recurring price ID for the £5/month Premium product (starts with price_…)
• NEXT_PUBLIC_STRIPE_PUBLISHABLE_KEY — publishable key (for future client-side Stripe.js)
• STRIPE_WEBHOOK_SECRET — signing secret from your webhook endpoint (production)
• NEXT_PUBLIC_APP_URL — e.g. http://localhost:3000 (used for checkout success/cancel redirects)

Setup steps:

1. In Stripe, create a Product with a £5/month recurring Price
2. Copy the Price ID into STRIPE_PRICE_ID in .env.local
3. Copy your test Secret key into STRIPE_SECRET_KEY
4. Restart the dev server after changing env vars
5. For production, set up webhook endpoint: https://your-domain.com/api/stripe/webhook
6. Listen for customer.subscription.created, updated, deleted

If Stripe is not configured, /settings/billing shows a friendly error instead of crashing.

Manual premium for testing (skip Stripe)

In the Supabase SQL editor, grant Premium to your account:

UPDATE profiles
SET subscription_tier = 'premium',
    subscription_status = 'active'
WHERE user_id = (SELECT id FROM auth.users WHERE email = '[email protected]');

Use 'trialing' instead of 'active' if you want to mimic a trial. Premium gating requires subscription_tier = 'premium' and subscription_status in ('active', 'trialing').

5. Run locally

npm run dev

Open http://localhost:3000

Project Structure

src/
├── app/                    # Next.js App Router pages & API routes
│   ├── api/activities/     # Workout logging & scoring
│   ├── api/stripe/         # Checkout & webhooks
│   ├── dashboard/          # Main dashboard
│   ├── activities/         # Workout forms
│   ├── analytics/          # Charts & insights
│   ├── social/             # Leaderboards & challenges
│   ├── profile/            # User profile & stats
│   └── settings/           # Account & billing
├── components/             # Reusable UI components
├── lib/
│   ├── scoring/            # Core scoring engine
│   ├── supabase/           # Database clients
│   ├── openai/             # AI Coach
│   └── stripe/             # Payments
└── types/                  # TypeScript definitions
supabase/migrations/        # PostgreSQL schema + RLS

Database Schema

All tables live in Postgres (Supabase) with Row Level Security enabled. Users are Supabase auth.users; every user-owned table references it with ON DELETE CASCADE and owner-based RLS policies (auth.uid()).

Core (001_initial_schema.sql)

• profiles — one per user (auto-created on signup), onboarding data, subscription state, and denormalized current index columns for fast leaderboards
• activities + gym_exercises — every logged workout; composite index on (user_id, started_at)
• workout_scores — per-activity sport index and breakdown
• split_index_history — the overall Split Index over time
• personal_records, goals, body_metrics (bodyweight history), recovery_snapshots
• ai_feedback — AI Coach output per activity
• friends (friend requests: pending/accepted/blocked), challenges + challenge_participants, achievements + user_achievements, notifications, workout_drafts

Scoring reference & leaderboards (002_scoring_reference_and_leaderboards.sql)

• sports — reference table for the 8 supported sports (name, endurance/strength category, metadata); public read
• strength_scores — per-exercise strength scoring (estimated 1RM, relative strength, volume load)
• reference_values — scoring-engine benchmark standards (e.g. 5k paces, big-3 1RM/bodyweight ratios) per sport × gender × experience; seeded, public read
• sleep_logs — nightly sleep hours, quality, bed/wake times
• leaderboard_entries — precomputed weekly/monthly/all-time rankings (refreshed by a scheduled job with the service role); public read
• Adds current_split_index (and endurance/strength) to profiles, kept in sync by trigger on split_index_history, plus missing indexes and tightened RLS for friend requests

Run the migrations in numeric order on a fresh project; 002 is purely additive on top of 001.

Re-running migrations (e.g. type "sport_category" already exists)

Migrations are not idempotent — do not paste the same file twice. If you see 42710: type "sport_category" already exists, migration 002 was already applied (at least partially). Run the verification queries below in the Supabase SQL editor, then:

• All 002 objects present → skip 002, run 003_integrations.sql only if you need integrations.
• 002 incomplete → do not re-run the full file; paste supabase/migrations/002b_apply_missing.sql in the SQL editor instead (idempotent remainder of 002).
• 003 fails similarly (integration_provider already exists) → skip 003; objects already exist.

-- Enums (001: sport_type; 002: sport_category, leaderboard_period; 003: integration_*)
SELECT typname FROM pg_type
WHERE typname IN ('sport_type','sport_category','leaderboard_period','integration_provider','import_job_status','sync_status')
  AND typtype = 'e' ORDER BY typname;

-- Key tables
SELECT table_name FROM information_schema.tables
WHERE table_schema = 'public'
  AND table_name IN ('profiles','sports','goals','strength_scores','reference_values','sleep_logs','leaderboard_entries','integration_connections','import_jobs')
ORDER BY table_name;

-- 002 profile columns (leaderboard denormalization)
SELECT column_name FROM information_schema.columns
WHERE table_schema = 'public' AND table_name = 'profiles'
  AND column_name IN ('current_split_index','current_endurance_index','current_strength_index','index_updated_at');

Scoring Engine

Located in src/lib/scoring/engine.ts. Key formulas:

• Endurance Index: pace × distance × duration × elevation × HR efficiency × temperature × fatigue
• Strength Index: Epley 1RM × relative strength × volume × fatigue
• Split Index: 0.5 × endurance + 0.5 × strength
• Fatigue: Acute:Chronic Workload Ratio (ACWR)
• Recovery: Derived from fatigue, ACWR, and rest days

Data sources & background jobs

Four ways in, no partner APIs:

• Typed (/gym/log, /cardio/log, /activities/new), or logged in one tap from a Hybrid Plan prescription (/hybrid-plan, /gym/log?hpeSession=). source = 'manual'.
• Recorded by the in-app GPS run (/cardio/gps-run), with live heart rate from a Bluetooth strap or, on iOS, AirPods Pro via a HealthKit workout session. source = 'gps'.
• Imported from Apple Health on iOS (HealthImportPlugin.swift → POST /api/health/import): workouts, HRV, resting heart rate, sleep and bodyweight from the phone's health store, which every wearable writes into. source = 'apple_health'. Foreground sync on launch and resume; see docs/logging-effort/.
• Imported from a file (/cardio/import → POST /api/activities/import for the preview, then POST /api/activities): GPX, TCX or FIT, dated within the last 90 days, with the route and per-sample streams where the file carries a track. source = 'file', idempotent on the file's hash.

There is no Strava or Garmin OAuth. docs/logging-effort/revert-log.md lists every piece and how to undo it.

For production cron jobs, set CRON_SECRET and configure Vercel cron:

# Leaderboard rank refresh (weekly / monthly / all-time)
GET /api/cron/leaderboard
Authorization: Bearer $CRON_SECRET

# Hybrid Plan weekly reports
GET /api/cron/hybrid-reports
Authorization: Bearer $CRON_SECRET

Data export (Premium): GET /api/export/activities?format=csv or ?format=json

Premium feature gating

Central flags live in src/lib/premium/features.ts (canAccess, PREMIUM_FEATURES).

Feature	Free	Premium
Full logging (manual, CSV, GPX/TCX)	✓	✓
Current Split Index	✓	✓
Per-workout cardio index	✓	✓
Dashboard 7-day history	✓	✓
Rules-based coach snippet	✓	✓
Country leaderboard	✓	✓
DOTS / GL / ExRx tiers	—	✓
Cardio HR accountability (TRIMP, EF, decoupling)	—	✓
GPT AI Coach (4 sections)	—	✓
90-day trends & projections	—	✓
Period comparison analytics	—	✓
Global leaderboards & rank percentile	—	✓
Data export	—	✓
OAuth integrations & auto-sync	—	✓

Deploy to Vercel

1. Push to GitHub
2. Import in Vercel
3. Add environment variables
4. Deploy

Premium Features (£5/month)

See src/lib/premium/features.ts for the canonical gate list. Highlights:

• Full Strength Index with DOTS / IPF GL + ExRx tiers
• Cardio HR accountability (TRIMP, EF, decoupling, VO2max confidence)
• GPT AI Coach after every workout
• 90-day analytics, projections & period comparison
• Global leaderboards, rank percentile & data export
• Strava, Garmin & all OAuth auto-sync

14-day free trial included.