StravaAI

Repo →

What it does

I post a lot of runs and rides to Strava. The descriptions are usually empty. StravaAI watches my Strava account via webhooks, and whenever I upload a new activity it writes a short, contextual description — a fun fact, a stat callout, or a motivational line — and updates the activity. It shows up in the Strava app without me doing anything.

How it works

1. OAuth: I connect my Strava account once.
2. Webhook: Strava pings the backend whenever I upload a new activity.
3. AI router: The server grabs the activity data, then picks the cheapest available model — Gemini 2.0 Flash first, then GPT-4o-mini, then Claude Haiku as fallback — to generate a description in one of three tones (Playful, Motivational, Stats).
4. Update: The generated text is written back to Strava via their API.

Stack

• Backend: Node.js + Express
• Database: libSQL (SQLite, via @libsql/client) — plan is to migrate to Turso when I deploy
• Auth: Strava OAuth 2.0
• Frontend: Vanilla HTML/CSS/JS dashboard for tone + sport filters and manual enrichment
• AI: Multi-provider router with automatic fallback

Status

Works end-to-end locally. Currently fixing an API key issue and then planning to deploy to Railway with a strava.* subdomain. Next after that: submit the app to Strava for review (required to support more than one connected athlete).