Meron

Self-hosted personal training dashboard. Brings together Strava activity history, FIT files, and strength-training logs into one place — without handing your data to a third party.

Live public preview at https://meron.gordongouger.com (Overview tab; deeper tabs are gated by SSO).

What it does

• Routes on a map — every ride and run plotted with elevation, heart rate, and pace overlays
• Training trends — CTL / ATL / TSB load curves, weekly volume, time-in-zone, distance histograms over time
• Race predictions — VDOT / Riegel-style time estimates from your recent races and tempo runs
• Strength progression — bench / squat / deadlift one-rep-max estimates pulled from FIT strength workouts
• Heatmaps — aggregate location heatmaps for runs, rides, and walks
• Chat widget — a small LLM-backed assistant that answers questions over your own ride and run data (read-only, opt-in)

The Overview tab is intentionally public-safe (counts, weekly sums, anonymized aggregates) so the project page on the main site can pull a live summary card without exposing the raw activity stream.

Stack

• Backend: Flask + Dash + pandas + scipy + plotly, served by gunicorn
• Storage: SQLite (single DB file in ~/.meron/)
• Activity parsing: fitparse for FIT, stravalib for the Strava API
• Auth: HTTP forward-auth from an upstream reverse proxy (Authelia in production; a built-in Google OAuth flow for solo dev)
• Maps: Leaflet via a thin Dash bridge
• Optional: an MCP server (mcp extra) so ChatGPT and Claude can query your training data, and an OpenAI extra for the in-app chat widget

Memory tuning for small deployments lives in the Dockerfile — gunicorn --max-requests recycling, MALLOC_ARENA_MAX=2, MALLOC_TRIM_THRESHOLD_=100000. Comfortable on a 2 GB VPS with 4 GB swap.

Quick start

# Clone + install
git clone git@github.com:gogouger/meron.git
cd meron
python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[web,api]" gunicorn

# Configure Strava OAuth — get the client id / secret from
# https://www.strava.com/settings/api, then:
cp .env.example .env
# fill in STRAVA_CLIENT_ID, STRAVA_CLIENT_SECRET, STRAVA_REDIRECT_URI

# Run
gunicorn --workers 1 --threads 4 --bind 0.0.0.0:8050 \
    strava_analytics.web.wsgi:app
# open http://localhost:8050

Or with Docker (uses the same env vars):

docker compose up -d

If you don't want to expose ports on your router, the tunnel profile starts a Cloudflare Tunnel sidecar — set TUNNEL_TOKEN in .env first, then docker compose --profile tunnel up -d.

Architecture notes

• All state lives in one SQLite database (~/.meron/meron.db by default) plus a sidecar JSON index for the FIT files. No external services required.
• Strava tokens are encrypted at rest with a key derived from MERON_SECRET_KEY so a stolen DB file doesn't immediately expose your account.
• The Dash UI is sharded into pages under strava_analytics/web/pages/ — each page- file is auto-registered. Adding a new page is one file.
• The "public Overview" pattern: Caddy's @public matcher exposes /, /login, /api/public/*, and Dash plumbing without forward_auth; everything else requires a session. See infra/Caddyfile for the production wiring.

License

MIT. Borrow it freely, ship your own variant. Personal training data is yours alone — none of it is in this repo.

Author

Gordon Gouger — built to scratch my own itch with running + lifting trends.