Skip to main content

Meilisearch Index Definition for HIPAA + FINRA

You already have pgvector for semantic search. Meilisearch (or similar) can own the text + metadata faceting for UI search filters.

Index Settings​

{
  "uid": "documents",
  "primaryKey": "id",

  "searchableAttributes": [
    "title",
    "summary",
    "body",
    "keywords",
    "tags"
  ],

  "filterableAttributes": [
    "document_type",
    "domain",
    "jurisdiction",
    "regulations",
    "security_class",
    "contains_phi",
    "contains_pii",
    "contains_financial",
    "status",
    "retention_category",
    "business_unit",
    "desk",
    "facility",
    "owner_role",
    "owner_user_id",
    "effective_date",
    "review_due_date",
    "retain_until",
    "created_at",
    "updated_at"
  ],

  "sortableAttributes": [
    "effective_date",
    "review_due_date",
    "retain_until",
    "created_at",
    "updated_at"
  ]
}

Document Structure​

Each indexed document maps from documents + document_metadata:

{
  "id": "550e8400-e29b-41d4-a716-446655440000",
  "title": "HIPAA Privacy Officer Policy",
  "summary": "Defines the role and responsibilities of the Privacy Officer",
  "body": "This policy establishes the requirements for...",
  "keywords": ["hipaa", "privacy", "officer", "compliance"],
  "tags": ["policy", "hipaa", "privacy"],

  "document_type": "policy",
  "domain": "security-privacy",
  "jurisdiction": ["US"],
  "regulations": ["HIPAA-164.316", "HIPAA-164.530"],
  "security_class": "confidential",
  "contains_phi": true,
  "contains_pii": true,
  "contains_financial": false,
  "status": "effective",
  "retention_category": "HIPAA-6Y",

  "business_unit": "Compliance",
  "desk": null,
  "facility": "Hospital-A",
  "owner_role": "Privacy Officer",
  "owner_user_id": "u123",

  "effective_date": "2025-01-01",
  "review_due_date": "2027-01-01",
  "retain_until": "2031-01-01",
  "created_at": "2025-01-01T10:00:00Z",
  "updated_at": "2025-01-10T11:00:00Z"
}

Search Integration with CODITECT API​

Hybrid Search Flow​

  1. User query → CODITECT /api/v1/search/hybrid
  2. pgvector → k-NN over embeddings → candidate doc IDs
  3. Meilisearch → text search + compliance filters → refined results
  4. RRF fusion → combined scoring
  5. Return → ranked results with snippets

Filter Syntax Examples​

# HIPAA PHI documents in specific facility
contains_phi = true AND facility = "Hospital-A"

# FINRA documents approaching review
regulations = "FINRA-4511" AND review_due_date < 1735689600

# Confidential documents by domain
security_class = "confidential" AND domain = "finance"

# Documents on legal hold
legal_hold = true

Integration with Existing Endpoints​

Enhanced Search Request​

POST /api/v1/search/hybrid

{
  "query": "breach notification procedure",
  "mode": "hybrid",
  "filters": {
    "regulations": ["HIPAA-164.316"],
    "status": "effective",
    "contains_phi": true
  },
  "limit": 20,
  "offset": 0
}

Response Mapping​

{
  "results": [
    {
      "id": "uuid",
      "title": "...",
      "summary": "...",
      "score": 0.95,
      "highlights": {...},
      "compliance": {
        "regulations": ["HIPAA-164.316"],
        "classification": "confidential",
        "retention": "HIPAA-6Y",
        "retain_until": "2031-01-01"
      }
    }
  ],
  "total": 45,
  "processing_time_ms": 23
}

Sync Strategy​

On Document Create/Update​

def sync_to_meilisearch(document, metadata):
    """Sync document to Meilisearch after DB update."""
    index = meilisearch_client.index('documents')

    doc = {
        'id': str(document.id),
        'title': document.title,
        'summary': document.summary,
        'body': get_document_text(document),
        'keywords': list(document.keywords),
        'tags': list(document.tags),
        'document_type': document.document_type,
        'domain': metadata.domain,
        'jurisdiction': metadata.jurisdiction,
        'regulations': metadata.regulations,
        'security_class': metadata.security_class,
        'contains_phi': metadata.contains_phi,
        'contains_pii': metadata.contains_pii,
        'contains_financial': metadata.contains_financial,
        'status': metadata.status,
        'retention_category': metadata.retention_category,
        'business_unit': metadata.business_unit,
        'facility': metadata.facility,
        'owner_role': metadata.owner_role,
        'owner_user_id': metadata.owner_user_id,
        'effective_date': metadata.effective_date.isoformat() if metadata.effective_date else None,
        'review_due_date': metadata.review_due_date.isoformat() if metadata.review_due_date else None,
        'retain_until': metadata.retain_until.isoformat(),
        'created_at': document.created_at.isoformat(),
        'updated_at': document.updated_at.isoformat()
    }

    index.add_documents([doc])

References​