n8n-nodes-rehydra

n8n community node for PII anonymization and rehydration using the Rehydra SDK.

Strip personal data before sending text to LLMs, translation APIs, or analytics — then restore it afterward.

Installation

In your n8n instance:

1. Go to Settings → Community Nodes
2. Enter n8n-nodes-rehydra
3. Click Install

Or via CLI:

npm install n8n-nodes-rehydra

Operations

Anonymize

Replaces PII with placeholder tags like <PII type="EMAIL" id="1"/>.

• Mode: Pseudonymize (reversible, default) or Anonymize (irreversible)
• NER Mode: Disabled (regex only, fast), Quantized (~280 MB model), or Standard (~1.1 GB model)
• PII Types: Select which types to detect (email, phone, IBAN, names, orgs, etc.)
• Options: Locale, allowlist terms, tag format, reuse IDs

Output fields:

| Field | Description | |-------|-------------| | anonymizedText | Text with PII replaced by tags | | piiMap | Encrypted PII map (pass to Rehydrate step) | | entities | Array of detected entities with type, position, confidence | | stats | Detection statistics (counts by type, processing time) |

Rehydrate

Restores original PII from placeholder tags using the encrypted PII map.

• PII Map: The piiMap JSON object from a prior Anonymize step
• Credential: Must use the same encryption key from the Anonymize step

Output fields:

| Field | Description | |-------|-------------| | rehydratedText | Text with original PII restored |

Inspect

Detects PII without modifying the text (dry run).

Same configuration as Anonymize, but returns the original text alongside detected entities and stats.

Credentials

The Rehydra API credential stores the AES-256 encryption key used for pseudonymize mode.

• Required for Rehydrate (must match the key used during Anonymize)
• Optional for Anonymize (a random key is generated if not configured, but then the PII map can't be decrypted later)
• Not needed for Inspect

Generate a key:

npx rehydra generate-key

Workflow Examples

Anonymize → Translate → Rehydrate

[Webhook] → [Rehydra: Anonymize] → [OpenAI: Translate] → [Rehydra: Rehydrate] → [Respond]

1. Incoming text is anonymized, stripping all PII
2. The safe text is sent to OpenAI for translation
3. The translated text is rehydrated, restoring original PII
4. The result is returned with PII intact in the translated text

PII Audit Pipeline

[Read CSV] → [Rehydra: Inspect] → [IF: totalEntities > 0] → [Slack: Alert]

Scan incoming data for PII and alert when sensitive data is detected.

LLM Summarization with Privacy

[Database] → [Rehydra: Anonymize] → [Claude: Summarize] → [Rehydra: Rehydrate] → [Save]

Summarize customer records through an LLM without exposing PII.

NER Model

When NER mode is set to quantized or standard, the ONNX model is auto-downloaded on first execution (~280 MB or ~1.1 GB). Subsequent runs use the cached model.

NER detects soft PII that regex can't catch:

• Person names
• Organization names
• Locations and places
• Addresses
• Dates of birth

For production workflows where latency matters, consider pre-downloading the model on your n8n server:

npx rehydra setup-ner

License

MIT