RedactiPHI - Redact

Processing Mode

Single

One document

Batch

Multiple files

De-identification Policy

Safe Harbor

Full HIPAA compliance - removes all 18 identifiers

LLM Safe

Tokenized for LLM use - enables re-identification

Analytics

Preserves dates/ages for research use

Full Redaction

Replace all PHI with [REDACTED] tags

Detection Engines Fast

Pattern Matching ~100ms - Regex patterns, identifiers

Transformer NER ~2-10s - AI entity recognition (GPU: ~500ms)

Speed Mode

Pattern matching is very fast. Transformer adds AI-based detection for better coverage.

Note: First request may take 15-20s to initialize models. Subsequent requests are much faster (~50ms).

LLM Reviewer

Use a local LLM (Ollama) to review and filter false positives from detection.

Enable LLM Reviewer

Age Handling

Redact

Remove all ages → [REDACTED_AGE]

Decade

Replace with decade → "40s", "60s"

Generalize 90+

Keep ages <90, generalize 90+ (HIPAA)

Keep All

Preserve all ages unchanged

Options

Re-identify output

Generate receipt

Show PHI highlights

Include confidence scores

Advanced Settings

Subject ID (for consistent tokens)

Same subject ID = same tokens for matching PHI across documents

Confidence Threshold 0.7

Only flag PHI with confidence above this threshold

Date Shift Range ±30 days

Random shift applied to all dates (preserves intervals)

Analytics Labels

Comma-separated tags for analytics grouping

Detected Format

📄 Plain Text

Standard text document

Chart Sessions

Link multiple documents for the same patient with consistent tokenization and date shifting.

Load Existing

Upload Files

📄

Drop files here or click to upload

PDF, C-CDA/XML, HL7, ZIP, TXT, JSON, HTML

LLM Round-Trip Demo

Test the full de-ID → LLM → re-ID workflow. Free demo uses Claude Haiku.

Enable LLM Demo

Input (Original)

Output (De-identified)

Click "Process" to de-identify the text...

Re-identification Demo

📝

Original

PHI Present

→

🔒

De-identified

Tokens Only

→

🔓

Re-identified

PHI Restored

Reason for Re-identification (required for audit)

Selective Re-identification Choose what to restore

Names Dates SSN MRN Phone Email Address Facility

All re-identification requests are logged with cryptographic receipts for HIPAA compliance. Unchecked PHI types will remain redacted.

🔒 De-identified Text

                                    Process text first to see de-identified output...
                                

🔓 Re-identified Text

                                    Click "Re-identify Text" to restore PHI...
                                

JSON Response

// Process text to see JSON response

De-identification Receipt

Document Info

Process text to generate receipt...

PHI Found: 0

Tokens Created: 0

Processing Time: 0ms

Policy: safe_harbor

Engines: -