OpenAI SDK

Wrap an OpenAI Assistants API run in a TapPass session. Every run emits a structured audit event with correlation IDs. Every server-side governance decision surfaces as a typed exception.

Install

pip install 'tappass[openai]'

Govern an Assistants run

from openai import OpenAI
from tappass.integrations.openai import guard_openai_run

client = OpenAI()

with guard_openai_run(
    client,
    assistant_id="asst_...",
    tappass_url="https://tappass.example.com",
    api_key="tp_...",
) as session:
    thread = client.beta.threads.create()
    run = session.create_and_poll_run(thread.id, "Summarize Q4")
    print(session.audit_url)

Async variant

from tappass.integrations.openai import guard_openai_run_async

async with guard_openai_run_async(
    client,
    assistant_id="asst_...",
    tappass_url=...,
    api_key="tp_...",
) as session:
    thread = await client.beta.threads.create()
    run = await session.create_and_poll_run(thread.id, "Summarize Q4")

Zero-code gateway (chat completions)

For client.chat.completions.create(...) without the Assistants API, use env-var passthrough instead — no SDK import needed:

export OPENAI_BASE_URL=https://tappass.example.com/v1
export OPENAI_API_KEY=tp_...

Every chat completion now routes through the governance pipeline transparently.

Handling governance decisions

from tappass import GovernanceDecision, PolicyBlockError

try:
    with guard_openai_run(client, assistant_id="asst_...",
                          tappass_url=..., api_key="tp_...") as session:
        session.create_and_poll_run(thread_id, "Show all SSNs")
except PolicyBlockError as e:
    print(f"Blocked by {e.blocked_by}: {e.reason}")
    print(f"See {e.audit_url}")
except GovernanceDecision as e:
    print(f"Governance: {e}")

Migrating from 0.5

The 0.5 env-var-only story for Assistants runs is replaced by guard_openai_run for full correlation and typed decisions. See the migration guide.