PII Shield
Eighteen detectors with real validators: Luhn for cards, mod-97 for IBANs, SSN range rules, token patterns for keys and JWTs. Plus profanity screening and spam scoring, all deterministic and under 50ms.
$ curl -s -X POST "https://piishield.ayamk.in/v1/redact" -H "content-type: application/json" -d '{"text":"Call me on 415-555-0132 or jane@acme.com","style":"label"}' | jq -r .redactedText{
"redactedText": "Call me on [PHONE_1] or [EMAIL_1]",
"findings": [
{ "type": "PHONE", "start": 11, "end": 23, "confidence": 0.85 },
{ "type": "EMAIL", "start": 27, "end": 40, "confidence": 0.99 }
],
"counts": { "PHONE": 1, "EMAIL": 1 },
"risk": { "level": "medium", "score": 45, "detectedTypes": ["PHONE", "EMAIL"] },
"meta": { "costClass": "free", "durationMs": 3 }
...6 endpoints, one key
Every route below is described in the OpenAPI document. Import it into Postman or generate a client.
| Method | Path | What it does |
|---|---|---|
| POST | /v1/scan | Find PII with spans and confidence, no modification. |
| POST | /v1/redact | Return redacted text, masking, labelling or hashing. |
| POST | /v1/profanity | Profanity screening with obfuscation handling. |
| POST | /v1/spam | Heuristic spam score with weighted signals. |
| POST | /v1/moderation | PII plus profanity plus spam in one call. |
| GET | /v1/types | Every detector, what it matches, how it validates. |
What you actually get back
curl -s -X POST "https://piishield.ayamk.in/v1/redact" -H "content-type: application/json" -d '{"text":"Call me on 415-555-0132 or jane@acme.com","style":"label"}' | jq -r .redactedText
{
"redactedText": "Call me on [PHONE_1] or [EMAIL_1]",
"findings": [
{ "type": "PHONE", "start": 11, "end": 23, "confidence": 0.85 },
{ "type": "EMAIL", "start": 27, "end": 40, "confidence": 0.99 }
],
"counts": { "PHONE": 1, "EMAIL": 1 },
"risk": { "level": "medium", "score": 45, "detectedTypes": ["PHONE", "EMAIL"] },
"meta": { "costClass": "free", "durationMs": 3 }
}
Built for the boring part
Validators, not just regex
A 16 digit number is not a card unless it passes Luhn. An IBAN is not an IBAN unless mod-97 checks out. That difference is the difference between a useful signal and noise.
Spans you can trust
Overlapping matches are resolved deterministically: the more specific, higher confidence detector wins. Findings come back sorted with start and end offsets, ready for highlight UIs.
Zero egress
Regex, checksums and wordlists inside the Worker. Your text never leaves the request, which is the only version of this a compliance team will sign off.
Priced per request, not per seat
All plans are managed by the RapidAPI marketplace, which applies a marketplace fee before payout. Annual and custom limits are available on request.
Where this earns its place
Sanitise prompts before an LLM call
Clean application logs
Support macros and notes
UGC moderation queues
Questions we get asked
Is this a substitute for a DLP product?
No. It is a fast pre-filter: catch the obvious PII in text before it lands somewhere permanent. Pair it with policy and access control, not instead of them.
Can I reverse a redaction?
With style hash you get a stable 8 character digest per value, so you can correlate occurrences without storing the original. Masks and labels are not reversible.
Which languages are supported for profanity?
English is the most complete list. Spanish, French and German cover common terms. The PII detectors are language independent.
How fast is it?
Deterministic scans on 10KB of text return in single digit milliseconds. The response includes durationMs so you can measure it yourself.
Subscribe to PII Shield
Free tier included. The listing is published on the RapidAPI marketplace; the key works against every endpoint on this page.