GDPR prompt AI

The masking, as an API.

Your code sends text as it is. Names, addresses and numbers are replaced with placeholders before a model sees them, and put back in the answer. The API speaks the OpenAI format, so changing the address and the key is the whole integration.

Read the docsRequest beta access
Your request
{
  "model": "auto",
  "messages": [{
    "role": "user",
    "content": "Reply to Ana Novak, tax number 41270517, about invoice 2024-117."
  }]
}
What the model receives
{
  "messages": [{
    "role": "user",
    "content": "Reply to [PERSON_1], tax number [TAX_ID_1], about invoice 2024-117."
  }]
}
What you get back
{
  "choices": [{
    "message": {
      "role": "assistant",
      "content": "Dear Ms Novak, thank you for your patience regarding invoice 2024-117 ..."
    }
  }]
}

Three endpoints

Everything goes through the same masking as the chat, with the same rules, measurements and audit trail.

Start in a minute

A company admin creates a key in the console. Keep it on your servers; each key can have its own monthly cap and can be revoked at any time.

Sessions
Send back the X-Session-Id header and a person keeps one placeholder for the whole conversation.
Streaming
Server-sent events in the OpenAI format; placeholders are restored as the text arrives.
Fail closed
If the masking cannot run, nothing is sent and you get 503.
from openai import OpenAI

client = OpenAI(base_url="https://api.gdprprompt.ai/v1",
                api_key="eucc_sk_…")
reply = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user",
               "content": "Reply to Ana Novak about invoice 2024-117."}],
)
print(reply.choices[0].message.content)