OpenAI

OpenAI's Python SDK uses httpx internally, making it fully compatible with Rdakt AI's transport middleware.

Setup

import httpx
from openai import AsyncOpenAI
from rdakt_ai import RdaktMiddleware

client = AsyncOpenAI(
    http_client=httpx.AsyncClient(
        transport=RdaktMiddleware(inner=httpx.AsyncHTTPTransport()),
    ),
)

Supported Models

All OpenAI chat models work with Rdakt AI: GPT-4o, GPT-4o Mini, GPT-4 Turbo, GPT-3.5 Turbo, and any future chat completion models.

Message Format

OpenAI uses the messages[].content format:

{
  "model": "gpt-4o",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Tell me about John Smith"}
  ]
}

Rdakt AI scans the content field of each message for PII.

Streaming

OpenAI streaming responses use SSE with this format:

data: {"choices":[{"delta":{"content":"Hello"},"index":0}]}

data: {"choices":[{"delta":{"content":" world"},"index":0}]}

data: [DONE]

Rdakt AI extracts content from choices[0].delta.content and deanonymizes tokens in real-time.

Streaming Example

stream = await client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "user", "content": "Write about Jane Doe (jane@acme.com)"}
    ],
    stream=True,
)

async for chunk in stream:
    content = chunk.choices[0].delta.content
    if content:
        print(content, end="", flush=True)

Non-Streaming Example

response = await client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "user", "content": "Summarize the case for John Smith (SSN 123-45-6789)"}
    ],
)
print(response.choices[0].message.content)

Sync Client

import httpx
from openai import OpenAI
from rdakt_ai import RdaktSyncMiddleware

client = OpenAI(
    http_client=httpx.Client(
        transport=RdaktSyncMiddleware(inner=httpx.HTTPTransport()),
    ),
)