HACS Persistence

PostgreSQL + pgvector persistence for healthcare data storage

Database and vector storage adapters optimized for healthcare AI applications.

🗄️ Database Support

PostgreSQL with pgvector

Primary storage solution for healthcare data:

• Relational Data - Patient records, observations, clinical data
• Vector Storage - Clinical embeddings via pgvector extension
• Schema Management - Automated migrations and versioning
• Healthcare Compliance - HIPAA-aware design patterns

🏥 Healthcare Schema

Optimized database tables for clinical workflows:

-- Core healthcare tables
patients              -- Patient demographics and clinical context
observations         -- Clinical measurements and findings
actors              -- Healthcare providers with role-based permissions
memory_blocks       -- AI agent episodic/procedural memory
evidence_items     -- Clinical guidelines and research
knowledge_base     -- Structured clinical knowledge

-- Vector storage for AI operations
patient_vectors     -- Patient data embeddings
clinical_vectors    -- Clinical note embeddings
memory_vectors      -- Memory content embeddings

📦 Installation

pip install hacs-persistence

🚀 Quick Start

Setup via HACS

# Automatic setup with migrations
python setup.py --mode local

# Database runs on localhost:5432
# Automatic pgvector extension installation

Direct Usage

from hacs_persistence import Adapter
from hacs_core import Patient

# Connect to healthcare database
adapter = Adapter(
    database_url="postgresql://hacs:password@localhost:5432/hacs"
)

# Store patient record
patient = Patient(
    full_name="Maria Rodriguez",
    birth_date="1985-03-15",
    gender="female"
)

# Save with automatic validation
saved_patient = adapter.save_resource(patient)
print(f"Saved patient: {saved_patient.id}")

🔧 Configuration

Environment Variables

# Primary database connection
DATABASE_URL=postgresql://hacs:password@localhost:5432/hacs

# Vector store configuration (uses pgvector by default)
VECTOR_STORE=pgvector

# Optional: External PostgreSQL for production
DATABASE_URL=postgresql://hacs:secure_password@prod-db:5432/hacs_production

Migration Management

# Run database migrations
python -m hacs_persistence.migrations $DATABASE_URL

# Check migration status
python -c "from hacs_persistence import get_migration_status; print(get_migration_status())"

📊 Performance

• Resource Operations: <50ms for standard CRUD
• Vector Queries: <100ms for similarity search
• Batch Operations: 1000+ records per second
• Memory Footprint: Minimal overhead

🔐 Security Features

• Connection Encryption - SSL/TLS support
• Role-based Access - Healthcare provider permissions
• Audit Trails - Complete operation logging
• Data Isolation - Organization-specific schemas

🛠️ Advanced Usage

Vector Operations

# Store clinical embedding
adapter.store_vector(
    resource_id="patient_123",
    embedding=[0.1, 0.2, ...],  # Clinical text embedding
    metadata={"type": "patient", "department": "cardiology"}
)

# Similarity search
similar_patients = adapter.vector_search(
    query_embedding=[0.1, 0.2, ...],
    resource_type="patient",
    top_k=5
)

Batch Operations

# Bulk insert for large datasets
patients = [Patient(...) for _ in range(1000)]
results = adapter.bulk_save(patients)

📄 License

Apache-2.0 License - see LICENSE for details.