reverse-chatgpt

A reverse-engineered, no-API-key ChatGPT client that streams responses in real time. It mimics a real browser session to bypass the authentication wall, solves proof-of-work and Turnstile challenges automatically, and exposes a clean HTTP streaming endpoint via FastAPI.

---

📽️ Proof of Concept

Demo video: poc.mp4 (included in repo)

To embed it on GitHub, upload the file via a GitHub issue or release and replace this line with the returned asset URL.

---

✨ Features

Feature	Details
No API key required	Works with anonymous ChatGPT sessions — no OpenAI account needed
Real-time stream response	Responses are streamed token-by-token using server-sent events (SSE), so you see output immediately as ChatGPT generates it
Browser fingerprint spoofing	Uses curl_cffi to impersonate a real Chrome browser, including TLS fingerprints
Auto session bootstrapping	Automatically fetches the latest build number, device ID, and chat-requirements token on startup
Proof-of-Work solver	Automatically solves OpenAI's proofofwork challenge embedded in chat requirements
Turnstile bypass	Handles Cloudflare Turnstile token negotiation silently
FastAPI REST endpoint	Drop-in /conversation POST endpoint that returns a text/event-stream response
Direct class usage	Use ChatGPT directly in Python without spinning up a server

---

📁 Project Structure

reverse-chatgpt/
├── app.py          # FastAPI application & /conversation endpoint
├── chat.py         # ChatGPT class — builds payloads, streams responses
├── gpt_session.py  # Session bootstrap (cookies, build number, sentinel tokens)
├── build.py        # Proof-of-work solver & token utilities
├── tunsile.py      # Turnstile challenge handler
├── utils.py        # Helpers: build number extraction, config loading, etc.
├── config.json     # Static configuration
├── requirements.txt
└── test.py         # Quick CLI smoke test

---

⚙️ Setup Guide

1. Clone the repository

git clone https://github.com/yourname/reverse-chatgpt.git
cd reverse-chatgpt

2. Create a virtual environment (recommended)

python3 -m venv venv
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

Note: curl_cffi requires a C compiler and libcurl. On Debian/Ubuntu:

sudo apt install build-essential libcurl4-openssl-dev

4. Run the test script (CLI)

python test.py

This instantiates ChatGPT directly and streams a response to stdout — no server needed.

5. Start the API server

uvicorn app:app --host 0.0.0.0 --port 5000

The server will be available at http://localhost:5000.

---

🚀 API Usage

POST /conversation

Send a prompt and receive a real-time streamed response.

Request

POST /conversation
Content-Type: application/json

{
  "text": "Write a brief Vue.js tutorial"
}

Response — text/event-stream

Tokens are streamed back as plain text chunks as soon as ChatGPT generates them.

Example with curl:

curl -X POST http://localhost:5000/conversation \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -d '{"text": "Explain async/await in JavaScript"}' \
  --no-buffer

Example with Python requests:

import requests

response = requests.post(
    "http://localhost:5000/conversation",
    json={"text": "Write a brief Vue.js tutorial"},
    stream=True,
)

for line in response.iter_lines(decode_unicode=True):
    if line:
        print(line)

---

🐍 Direct Python Usage

from chat import ChatGPT

gpt = ChatGPT()

for chunk in gpt.reply_chat("What is the capital of France?"):
    print(chunk, end="", flush=True)