Import Apple Health data into PostgreSQL

The Health app on the iPhone can export everything it holds, but the export is one large XML file that a spreadsheet or a database cannot read as it is. This guide loads it into PostgreSQL as clean tables (samples, sleep sessions, workouts, daily summaries) with Open Health Sync, then queries them with SQL. It takes about 15 minutes, and everything runs on your own computer.

1. Start Postgres and Open Health SyncCopy Anchor Link

You need Docker with Compose v2.24 or later, plus curl, jq and openssl. From a clone of the repository:

git clone https://github.com/soma-intelligence/open-health-sync.git && cd open-health-sync
cp .env.example .env
sed -i.bak "s|^OHS_ENCRYPTION_KEY=.*|OHS_ENCRYPTION_KEY=$(openssl rand -base64 32)|" .env
docker compose --env-file .env -f deploy/docker-compose.yml up -d --wait

This starts PostgreSQL (with TimescaleDB) and the ohs server next to it. The quickstart explains each line.

2. Create your userCopy Anchor Link

This creates a tenant called personal, a user called me and an API key, and writes the three to a file the next steps read:

docker compose --env-file .env -f deploy/docker-compose.yml exec -T ohs \
  /ohs admin personal setup > ~/.ohs.env
chmod 600 ~/.ohs.env
set -a; . ~/.ohs.env; set +a

The key in that file reads everything on your instance: treat it like a password.

3. Export from the iPhoneCopy Anchor Link

Open the Health app, tap your picture, then Export All Health Data. Send the export.zip to your computer (AirDrop works). A few years of data make a file of a few hundred megabytes.

4. Import the exportCopy Anchor Link

Install the ohs command on your computer: take the archive for your system from the latest release (darwin is macOS), unpack it and put ohs on your PATH. Then import the export through a token that can only upload:

export OHS_TOKEN=$(curl -s -X POST "$OHS_URL/v1/users/$OHS_USER/tokens" \
  -H "Authorization: Bearer $OHS_KEY" -d '{"scopes":["upload"]}' | jq -r .token)
ohs import apple-health export.zip --dry-run    # counts what it would send
ohs import apple-health export.zip

The import reads the zip without extracting it, in about 150 MB of memory, and prints counts, never a health value. Running it again sends nothing twice, so an interrupted import is resumed by running it again. Many samples carry no time zone: add --tz Europe/Paris (your own zone) so they count on the right local day. All the options are in importing a Health app export.

The server then converts units, rebuilds each night from its sleep stages and removes duplicates. Wait until pending_jobs is 0:

curl -s "$OHS_URL/v1/users/$OHS_USER/coverage" -H "Authorization: Bearer $OHS_KEY" | jq .pending_jobs

5. Query it with SQLCopy Anchor Link

Open psql in the database container, with your user id in a variable:

docker compose --env-file .env -f deploy/docker-compose.yml exec postgres \
  psql -U ohs -d ohs -v user="$OHS_USER"

What came in, by type:

SELECT quantity_type, unit, count(*) AS samples,
       min(start_date)::date AS first_day, max(start_date)::date AS last_day
FROM ohs_data.quantity_samples
WHERE user_id = :'user'
GROUP BY quantity_type, unit
ORDER BY samples DESC
LIMIT 15;

Resting heart rate, week by week:

SELECT date_trunc('week', start_date)::date AS week,
       round(avg(value)::numeric, 1) AS resting_bpm
FROM ohs_data.quantity_samples
WHERE user_id = :'user' AND quantity_type = 'resting_heart_rate'
GROUP BY week
ORDER BY week DESC
LIMIT 12;

Your last nights:

SELECT bedtime_start, round(total_sleep_seconds / 3600.0, 1) AS hours,
       deep_sleep_seconds / 60 AS deep_min, rem_sleep_seconds / 60 AS rem_min
FROM ohs_data.sleep_sessions
WHERE user_id = :'user'
ORDER BY bedtime_start DESC
LIMIT 7;

Your last workouts:

SELECT start_date, activity_type, duration_s / 60 AS minutes,
       round((distance_m / 1000)::numeric, 2) AS km, avg_heart_rate
FROM ohs_data.workouts
WHERE user_id = :'user'
ORDER BY start_date DESC
LIMIT 10;

Steps need care: an iPhone and an Apple Watch both count them, so adding every sample counts a walk twice. Look at each source apart:

SELECT (start_date AT TIME ZONE 'Europe/Paris')::date AS day, source_name,
       sum(value)::int AS steps
FROM ohs_data.quantity_samples
WHERE user_id = :'user' AND quantity_type = 'step_count'
  AND start_date > now() - interval '7 days'
GROUP BY day, source_name
ORDER BY day DESC, steps DESC;

The REST API does this choice for you: GET /v1/users/{id}/daily keeps one source per figure and day (the merged view).

About the tablesCopy Anchor Link

Every table is described in the data model. Times are stored in UTC, every value in one unit per type (bpm, kg, count), and each row says where it came from (provider, source_name, device_model). The tables follow the migrations of each release; the REST API under /v1 is the interface that stays stable, so prefer it for code you ship to others.

Keep it up to dateCopy Anchor Link

Export again and run the same import: only the new records are stored. For data that arrives on its own, an iOS app built with the Swift SDK uploads new HealthKit data in the background.

NextCopy Anchor Link

• Connect Claude to your Apple Health data and ask it questions.
• Export everything as CSV or Parquet for a spreadsheet or pandas.
• Add a ring or a band: every connector lands in the same tables.