cadence-boostcamp-sync

📦 This has moved into the Cadence monorepo, under services/boostcamp-sync/. This standalone repo is kept only for the existing Railway deployment — make new changes in the monorepo.

A small sync service that logs into Boostcamp, pulls your training history, classifies each workout (Push / Pull / Legs and the muscles worked), and upserts the results to Supabase. It's the companion backend for the Cadence dashboard's Training view, but it runs standalone and has no dependency on the dashboard beyond a shared Supabase project.

It is the only component that holds Boostcamp credentials. Everything else just reads the classified rows out of Supabase.

What it does

• Logs in fresh each run and fetches recent training history via boostcamp-api.
• Classifies each workout with a keyword map (EXERCISE_MAP in sync.py): majority-vote across the workout's exercises → Push / Pull / Legs / Other, plus a deduplicated list of muscles worked.
• Upserts one row per workout into boostcamp_workouts and a single summary row (streak, totals) into boostcamp_summary. Re-runs are idempotent — a workout is keyed by its Boostcamp date, so a late-logged workout is stored on its real date, never "now".
• Notifies Discord (optional) when a fresh workout lands, and on crashes.
• Auto-completes a habit (optional) if you also run the Cadence dashboard: any habit whose name contains workout / run / gym / train / exercise gets logged for that date.

Architecture

Vercel cron (3x/day) ─┐
                      ├─► POST /api/cron/boostcamp-pull ─┐
Dashboard "Pull" tap ─┴─► POST /api/boostcamp/sync ──────┴─► POST /sync  ─► sync.py
   (Cadence on Vercel)                                       (this service on Railway)
                                                                   │
                                                                   ▼
                                                               Supabase  ◄─ dashboard reads

The service is a tiny request-driven HTTP server (server.py) with no internal scheduler — every sync arrives as an HTTP POST /sync. That lets it run on Railway with App Sleeping (scale-to-zero); the inbound request wakes it. sync.py is also runnable directly for a one-off CLI pull.

Method	Path	Auth	Purpose
GET	/health	none	Liveness check → {"ok": true}
POST	/sync	header X-Sync-Secret	Start a pull. 202 {"status":"started"} or 409 {"status":"already_running"}

Quick start (local)

python -m venv .venv && . .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt

cp .env.example .env        # then fill in your values
python sync.py              # one-off CLI pull
# or run the HTTP server:
python server.py            # serves /health and /sync on :8080

Before the first run, create the tables in your Supabase project by pasting migration.sql into the Supabase SQL editor.

Configuration

Copy .env.example to .env and set:

Var	Required	Purpose
BOOSTCAMP_EMAIL	✅	Boostcamp login email
BOOSTCAMP_PASSWORD	✅	Boostcamp login password
SUPABASE_URL	✅	Supabase project URL
SUPABASE_KEY	✅	Supabase service_role key (the sync writes rows)
BOOSTCAMP_SYNC_SECRET	✅ (server)	Shared secret guarding POST /sync; match it on the caller
DISCORD_WEBHOOK_URL	optional	Webhook for "workout logged" and crash alerts
TZ_OFFSET_MINUTES	optional	UTC offset in minutes for week math; default -420 (US Pacific)
PORT	optional	HTTP port for server.py (Railway injects this)

Deploy

See DEPLOY.md for the full Railway + Vercel walkthrough (generate a public domain, enable App Sleeping, wire the shared secret, register the cron).

Customizing classification

EXERCISE_MAP in sync.py is an ordered list of (keyword_substring, classification, [muscles]). Matching is lowercase-substring, first match wins. Add your own exercises or retarget muscles by editing that list — no other code changes needed. Anything unmatched classifies as Other.

License

MIT