POST/v1/inbound/tokenize

Inbound Tokenization

Scrub and tokenize Protected Health Information from clinical text before it reaches any cloud LLM.

<10ms latency

Description

This endpoint uses a two-stage pipeline to detect and replace PHI entities:

Local Regex Patterns — High-speed pattern matching for SSNs, MRNs, phone numbers, and dates of birth.
spaCy NER Model — A fine-tuned Named Entity Recognition model trained on clinical corpora to identify patient names, physician names, and locations.

Detected entities are replaced with deterministic tokens (e.g., "Patient Forgive" → [ID_001]). Tokens are reversible only by the originating EnorAI instance.

Important

Tokenization runs locally. No raw PHI is transmitted to EnorAI cloud services.

Request body

Parameters

Name	Type	Required	Description
`input_text`	string	Required	The raw clinical text containing PHI to be scrubbed and tokenized.
`fhir_context`	string	Optional	Optional EHR system reference for FHIR R4 compliance audit linking.

Response

Response fields

Name	Type	Required	Description
`tokenized_text`	string	Required	Input text with PHI replaced by deterministic tokens.
`entities_found`	integer	Required	Number of PHI entities detected and tokenized.
`processing_time_ms`	number	Required	Processing time in milliseconds.
`token_map_id`	string	Required	UUID reference to stored token map for de-tokenization.

Request

curl -X POST \
  https://api.enorai.com/v1/inbound/tokenize \
  -H "Authorization: Bearer $ENORAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input_text": "Patient Forgive presents with a fever of 39.2°C.",
    "fhir_context": "Patient/p-28491"
  }'
from enorai import EnorAI
 
client = EnorAI(api_key="ENORAI_API_KEY")
 
result = client.inbound.tokenize(
  input_text="Patient Forgive presents with a fever.",
  fhir_context="Patient/p-28491"
)
print(result.tokenized_text)

Response · 200 OK

{
  "status": "success",
  "scrubbed_text": "Patient [ID_001] presents with a fever of 39.2°C.",
  "entities_found": 1,
  "processing_time_ms": 6.2,
  "token_map_id": "tmap_a1b2c3d4-e5f6-7890"
}