medgemma-vault 0.1.0

Local FHIR-based Personal Health Record

medgemma-vault

A local-first personal health record backed by FHIR R4 and SQLite. All data stays on your machine in a single database file at ~/.medgemma/vault.db. Records are stored as validated FHIR R4 JSON, making them portable and interoperable with any FHIR-compatible system.

The optional medgemma AI engine can be added for natural-language queries against your records, but the vault works standalone as a pure health record store.

Features

• FHIR R4 compliant — every record is a validated FHIR resource (Patient, Observation, MedicationStatement, Condition, AllergyIntolerance, Procedure, DocumentReference)
• SQLite storage — single-file database with WAL mode, no server required
• CLI interface — manage records, import data, and export FHIR Bundles from the terminal
• Apple Health import — stream-parse large export.xml files with memory-efficient iterparse
• CSV and JSON import — bring in data from spreadsheets or existing FHIR Bundles
• Document storage — attach images, PDFs, and other files as FHIR DocumentReferences with content-hash deduplication
• FHIR Bundle export — export your entire record as a standard FHIR Bundle JSON
• Optional AI queries — ask natural-language questions about your health data (requires medgemma)

Requirements

• Python 3.10+

Installation

# Clone the repository
git clone [e-mail]:chibokocl/medgemma-vault.git
cd medgemma-vault

# Install in editable mode
pip install -e .

# Or with AI features (requires medgemma)
pip install -e ".[ai]"

# Or with development dependencies
pip install -e ".[dev]"

Quick start

1. Initialise the vault

Create the database and set up your patient profile:

medgemma-vault init --given-name Jane --family-name Doe

You can also provide optional flags:

medgemma-vault init \
  --given-name Jane \
  --family-name Doe \
  --birth-date 1990-05-15 \
  --gender female

If you omit --given-name or --family-name, the CLI will prompt you interactively.

2. Add records

# Vital signs
medgemma-vault add vitals --type weight --value 68.5
medgemma-vault add vitals --type heart_rate --value 72
medgemma-vault add vitals --type blood_pressure_systolic --value 120
medgemma-vault add vitals --type blood_glucose --value 95 --date 2024-06-01

# Medications
medgemma-vault add medication --name Metformin --dosage "500mg twice daily"
medgemma-vault add medication --name Aspirin --dosage "100mg daily" --status active

# Conditions
medgemma-vault add condition --name "Type 2 Diabetes"
medgemma-vault add condition --name "Seasonal allergies" --status active

# Clinical notes
medgemma-vault add note --text "Felt dizzy after lunch, resolved after 30 minutes"

# Documents (images, PDFs, etc.)
medgemma-vault add document chest-xray.png -d "Chest X-ray 2024"
medgemma-vault add document blood-work.pdf -d "Annual blood work results"

3. View records

# List all records
medgemma-vault list

# Filter by resource type
medgemma-vault list --type Observation
medgemma-vault list --type MedicationStatement

# Filter by date
medgemma-vault list --since 2024-01-01

# Limit results
medgemma-vault list --type Observation --limit 10

# Show full FHIR JSON for a specific record
medgemma-vault show <resource-id>

# Summary counts
medgemma-vault stats

4. Export

# Print FHIR Bundle to stdout
medgemma-vault export

# Save to file
medgemma-vault export -o my-records.json

The export produces a standard FHIR Bundle of type collection containing all resources in your vault.

Importing data

Apple Health

Export your data from the Apple Health app on iPhone (Health > profile picture > Export All Health Data), then:

medgemma-vault import apple-health export.xml

The adapter uses streaming XML parsing (iterparse with elem.clear()) so it can handle large export files without loading them entirely into memory. Units are automatically converted (e.g. lbs to kg, Fahrenheit to Celsius).

Supported Apple Health types:

Apple Health Identifier	Mapped vital type
HKQuantityTypeIdentifierBodyMass	weight
HKQuantityTypeIdentifierHeight	height
HKQuantityTypeIdentifierHeartRate	heart_rate
HKQuantityTypeIdentifierBloodPressureSystolic	blood_pressure_systolic
HKQuantityTypeIdentifierBloodPressureDiastolic	blood_pressure_diastolic
HKQuantityTypeIdentifierBodyTemperature	temperature
HKQuantityTypeIdentifierRespiratoryRate	respiratory_rate
HKQuantityTypeIdentifierOxygenSaturation	oxygen_saturation
HKQuantityTypeIdentifierBodyMassIndex	bmi
HKQuantityTypeIdentifierBloodGlucose	blood_glucose

CSV

medgemma-vault import csv vitals.csv

Expected CSV format:

type,value,date
weight,72.5,2024-06-01
heart_rate,68,2024-06-02
blood_glucose,95,2024-06-03

The type column must be one of the supported vital types (see table below). Rows with unknown types or invalid values are skipped.

JSON / FHIR Bundle

medgemma-vault import json records.json

Accepts either a single FHIR resource or a FHIR Bundle. Resources are validated against the FHIR R4B schema before being stored.

Supported vital types

These can be used with medgemma-vault add vitals --type <type>:

Type	LOINC Code	Display	Unit
weight	29463-7	Body weight	kg
height	8302-2	Body height	cm
heart_rate	8867-4	Heart rate	beats/minute
blood_pressure_systolic	8480-6	Systolic blood pressure	mmHg
blood_pressure_diastolic	8462-4	Diastolic blood pressure	mmHg
temperature	8310-5	Body temperature	C
respiratory_rate	9279-1	Respiratory rate	breaths/minute
oxygen_saturation	2708-6	Oxygen saturation	%
bmi	39156-5	Body mass index	kg/m2
blood_glucose	2339-0	Glucose [Mass/volume] in Blood	mg/dL

CLI reference

Command	Description
medgemma-vault init	Create DB and patient profile (prompts for name)
medgemma-vault add vitals --type <type> --value <n>	Add a vital sign observation
medgemma-vault add medication --name <name> [--dosage <text>]	Add a medication statement
medgemma-vault add condition --name <name>	Add a condition
medgemma-vault add note --text <text>	Add a clinical note
medgemma-vault add document <file> [-d <description>]	Upload an image or PDF
medgemma-vault import apple-health <file>	Import Apple Health export.xml
medgemma-vault import csv <file>	Import observations from CSV
medgemma-vault import json <file>	Import FHIR JSON or Bundle
medgemma-vault list [--type <type>] [--since <date>]	List records
medgemma-vault show <id>	Show full FHIR JSON for a record
medgemma-vault export [-o <file>]	Export all records as a FHIR Bundle
medgemma-vault stats	Show summary counts by resource type
medgemma-vault ask "<question>"	AI query (requires medgemma)

Architecture

medgemma-vault/
├── pyproject.toml
├── src/medgemma_vault/
│   ├── __init__.py
│   ├── __main__.py          # python -m medgemma_vault
│   ├── _version.py           # 0.1.0
│   ├── config.py              # paths, LOINC codes, Apple Health mapping
│   ├── db.py                  # SQLite connection, schema, low-level CRUD
│   ├── store.py               # VaultStore facade (high-level CRUD)
│   ├── models.py              # FHIR R4B resource builders + validation
│   ├── cli.py                 # Click CLI commands
│   ├── agent.py               # optional medgemma AI integration
│   └── adapters/
│       ├── base.py            # BaseAdapter ABC
│       ├── apple_health.py    # Apple Health XML adapter
│       ├── csv_import.py      # CSV adapter
│       ├── json_import.py     # JSON/FHIR Bundle adapter
│       └── document.py        # image/PDF adapter
└── tests/
    ├── conftest.py
    ├── test_config.py
    ├── test_db.py
    ├── test_models.py
    ├── test_store.py
    ├── test_cli.py
    └── test_adapters/
        ├── test_apple_health.py
        ├── test_csv_import.py
        └── test_document.py

Data flow

CLI / Adapter
      │
      ▼
  VaultStore          (store.py — high-level facade)
      │
      ├──► models.py  (build + validate FHIR resources)
      │
      ▼
    db.py             (SQLite CRUD)
      │
      ▼
  ~/.medgemma/vault.db

Database schema

The SQLite database has three tables:

resources — one row per FHIR resource, data stored as JSON:

• id (TEXT PK) — FHIR resource ID (UUID)
• resource_type (TEXT) — e.g. "Patient", "Observation"
• patient_id (TEXT) — reference to the owning Patient
• recorded_date (TEXT) — effective/recorded date for filtering
• source (TEXT) — origin of the record ("manual", "apple_health", "csv", etc.)
• data (TEXT) — full FHIR resource as JSON
• created_at, updated_at (TEXT) — ISO timestamps

documents — file metadata linked to DocumentReference resources:

• id (TEXT PK)
• resource_id (TEXT FK) — references resources.id
• file_path, mime_type, sha256, file_size

meta — key-value store for schema version tracking.

Storage locations

Path	Description
~/.medgemma/vault.db	SQLite database
~/.medgemma/documents/	Uploaded document files (content-hash prefixed)

Override the base directory by setting the MEDGEMMA_VAULT_DIR environment variable:

export MEDGEMMA_VAULT_DIR=/path/to/custom/vault

Development

# Install with dev dependencies
pip install -e ".[dev]"

# Run tests
pytest tests/ -v

# Run a specific test file
pytest tests/test_store.py -v

# Run with output
pytest tests/ -v -s

Dependencies

Package	Version	Purpose
fhir.resources	>=7.0.0	FHIR R4B data models and validation
click	>=8.0	CLI framework
medgemma	optional	AI-powered health queries ([ai] extra)
pytest	>=7.0	Testing ([dev] extra)
pytest-mock	any	Test mocking ([dev] extra)

License

MIT