Testing

Simulation: Dry-Run Mode

Test your security rules without spending a single cent on tokens. By passing the X-EdgeMask-Mode: dry-run header, EdgeMask will intercept the request, perform redaction, and return a simulation report.

How to Enable Dry-Run

Add a single header to your request:

X-EdgeMask-Mode: dry-run

The request will be fully intercepted and analyzed, but no tokens will be consumed and no request will be forwarded to the LLM provider.

Example Request

Send a prompt containing PII data with the dry-run header.

terminal

Bash

curl -X POST https://api.edgemask.pro/v1/chat/completions \
 -H "X-EdgeMask-Key: em_prod_92nd9..." \
 -H "X-EdgeMask-Mode: dry-run" \
 -d '{
"model": "gpt-4o",
"messages": [{"role": "user", "content": "Charging $500 to card 4532 1234 5678 9010"}]
}'

Simulation Response

EdgeMask returns a detailed analysis report instead of an LLM response.

response.json

JSON

{
  "status": "intercepted",
  "security_action": "masked_pii",
  "entities_detected": ["CREDIT_CARD"],
  "masked_prompt": "Charging $500 to card [MASKED_CC]",
  "latency_saved": "350ms",
  "token_cost_avoided": "$0.02"
}

Response Fields

Field

Type

Description

status

string

Always "intercepted" in dry-run mode. Indicates the request was caught by the gateway.

security_action

string

The action that would have been taken in production (e.g., masked_pii, blocked, allowed).

entities_detected

string[]

Array of PII/PHI entity types detected in the prompt (SSN, CREDIT_CARD, EMAIL, etc.).

masked_prompt

string

The prompt as it would appear after redaction, with sensitive data replaced by tokens.

latency_saved

string

Estimated latency saving if this had been a cached response.

token_cost_avoided

string

Estimated token cost that would be avoided via caching or blocking.