piai

Python port of @mariozechner/pi-ai.

Use your ChatGPT Plus/Pro subscription to access GPT models from Python — no API key, no per-token billing. Authenticates via OAuth using your existing ChatGPT account and streams completions directly from ChatGPT's backend.

---

• Python 3.12+
• uv (recommended) or pip
• A ChatGPT Plus or Pro subscription

---

uv add pi-ai-py
# or
pip install pi-ai-py

From source:

git clone https://github.com/Xplo8E/piai
cd piai
uv sync

---

Run once to authenticate. Opens a browser for you to log in with your ChatGPT account.

piai login

Credentials are saved to auth.json in the current directory. Keep it private — add to .gitignore. If you've already logged in with the JS CLI (npx @mariozechner/pi-ai login openai-codex), the same auth.json works without re-logging in.

---

piai run "Explain async/await in Python"
piai run "What is 2+2?" --model gpt-5.1
piai run "Summarize this" --system "You are a concise assistant"
piai status      # check login state
piai logout

---

Streams the model response as typed events.

import asyncio
from piai import stream
from piai.types import Context, UserMessage, TextDeltaEvent, DoneEvent

async def main():
    ctx = Context(
        system_prompt="You are a helpful assistant.",
        messages=[UserMessage(content="What is the capital of France?")]
    )
    async for event in stream("gpt-5.1-codex-mini", ctx):
        if isinstance(event, TextDeltaEvent):
            print(event.text, end="", flush=True)
        elif isinstance(event, DoneEvent):
            print()
            print(f"Tokens: {event.message.usage['input']} in, {event.message.usage['output']} out")

asyncio.run(main())

Returns the full AssistantMessage after the response finishes.