throwback

Bulk-imports a full Garmin Connect history into Strava. Handles the messy parts of a GDPR data export: filters the ~94% of files that are background telemetry rather than workouts, refreshes OAuth tokens on demand, treats Strava's server-side duplicate detection as success, logs every file's outcome so re-runs skip already-uploaded activities, and stays under the 200 req / 15 min API rate limit.

Built when I switched from Garmin to Strava as my primary fitness platform and discovered that the only sane way to move ~1,000 historical activities is to write the tool yourself.

Pipeline

Garmin GDPR export (a single ZIP, hundreds of MB)
        │
        ▼   unzip outer + nested UploadedFiles_*.zip
fit_staging/   ~16,000 raw FIT files (workouts + monitoring + sleep + ...)
        │
        ▼   filter_activities.py  (reads each FIT's file_id header)
activities_to_upload/   ~1,000 actual workout files
        │
        ▼   upload.py  (rate-limited, resumable, dedup-aware)
Strava

Setup

python3 -m venv .venv
.venv/bin/pip install stravalib python-dotenv fitparse
cp .env.template .env   # fill in STRAVA_CLIENT_ID and STRAVA_CLIENT_SECRET
.venv/bin/python do_oauth.py   # browser OAuth, writes tokens to .env
.venv/bin/python check_auth.py # confirm auth works
.venv/bin/python smoke_test.py # round-trip with a synthetic GPX in Central Park

Create the API app at https://www.strava.com/settings/api. Set Authorization Callback Domain to localhost. Use app_icon.png from this repo if Strava asks for an icon during setup.

Running an import

After requesting a Garmin GDPR data export, wait for the download email (typically 1–14 days), then:

# 1. Pull the FIT files out of the nested archives
unzip Garmin_Download.zip "DI_CONNECT/DI-Connect-Uploaded-Files/*.zip" -d /tmp/garmin/
mkdir -p fit_staging && cd fit_staging
unzip -oq /tmp/garmin/DI_CONNECT/DI-Connect-Uploaded-Files/UploadedFiles_0-_Part1.zip
unzip -oq /tmp/garmin/DI_CONNECT/DI-Connect-Uploaded-Files/UploadedFiles_0-_Part2.zip
cd ..

# 2. Filter to real activities only (drops sleep/HR/monitoring data)
.venv/bin/python filter_activities.py fit_staging/ activities_to_upload/

# 3. Bulk upload
.venv/bin/python upload.py activities_to_upload/

The upload script logs every result to uploads_log.jsonl. Killing the script mid-run and restarting is safe; previously-uploaded files are skipped on the next pass. Strava's daily upload limit is generous but finite — a 1,000-activity history finishes in ~85 minutes at the default 5-second pacing.

Files

File	Purpose
auth.py	Shared token refresh + Client construction
do_oauth.py	One-time OAuth: opens browser, captures token via localhost:8000
check_auth.py	Sanity check: refresh token + fetch athlete profile
smoke_test.py	Round-trip validation: uploads a synthetic GPX, verifies, then prompts manual delete
filter_activities.py	Reads FIT file headers and keeps only file_type == 'activity'
upload.py	Bulk uploader with dedup, resume, and rate-limit handling
.env.template	Credential placeholders. The real .env is gitignored.
app_icon.png	Plain orange icon for the Strava API app form

Notes on quirks

• Strava removed DELETE from the public API. Test activities have to be deleted manually in the Strava web UI.
• The GDPR export download link expires in ~3 days. Treat the email as urgent.
• Garmin's export bundles a lot more than activities — sleep, HR, weight scale, body battery, training readiness, step monitoring. The file_id message in each FIT identifies the kind of file; only those with type == 'activity' are real workouts.
• Duplicates aren't errors. Strava de-duplicates by activity start time and device, so re-uploading already-synced activities is expected and the script treats it as success.

License

MIT — see LICENSE.