DeepSeek4Free

You must provide a DeepSeek auth token to initialize the API. You can retrieve this from chat.deepseek.com using one of two methods:

Method 1: From LocalStorage (Recommended)

1. Log in to chat.deepseek.com.
2. Open browser developer tools (F12).
3. Go to the Application tab.
4. In the left sidebar, expand Local Storage and select https://chat.deepseek.com.
5. Find the key userToken and copy its value.

Alternatively, run this in the browser console:

JSON.parse(localStorage.getItem("userToken")).value

Method 2: From Network Tab

1. Log in to chat.deepseek.com.
2. Open developer tools (F12) and go to the Network tab.
3. Perform any action in the chat.
4. Inspect the request headers of the requests and copy the authorization token (remove the Bearer prefix).

To use this package, clone the repository and install the required dependencies using pip.

git clone https://github.com/yourusername/deepseek4free.git
cd deepseek4free
pip install -r requirements.txt

If you encounter Cloudflare challenges (e.g., "Just a moment..." page) or receive the error "Please wait a few minutes before trying again", you need to capture a cf_clearance cookie.

Run the following command to open an undetected browser, solve the challenge, and save the cookie to dsk/cookies.json:

python -m dsk.bypass

The API will automatically use the captured cookie in future requests.

You can create threaded conversations by tracking the message_id from a previous response and passing it as parent_message_id in the next call.

# 1. Start session
chat_id = api.create_chat_session()
parent_id = None

# 2. Send initial message and capture message_id
for chunk in api.chat_completion(chat_id, "Tell me about neural networks"):
    if 'message_id' in chunk:
        parent_id = chunk['message_id']

# 3. Send follow-up in the same thread
for chunk in api.chat_completion(
    chat_id, 
    "How do they compare to other ML models?",
    parent_message_id=parent_id
):
    if chunk['type'] == 'text':
        print(chunk['content'], end='', flush=True)

The core interface is DeepSeekAPI. You initialize it with your auth token, create a session, and then iterate over the streaming response from chat_completion.

from dsk.api import DeepSeekAPI

# Initialize with your auth token
api = DeepSeekAPI("YOUR_AUTH_TOKEN")

# Create a new chat session
chat_id = api.create_chat_session()

# Simple chat completion
prompt = "What is Python?"
for chunk in api.chat_completion(chat_id, prompt):
    if chunk['type'] == 'text':
        print(chunk['content'], end='', flush=True)

The /html GET endpoint visits a URL, bypasses Cloudflare, and returns the full HTML content of the page. In addition to the HTML body, it provides the session cookies and User-Agent via HTTP response headers.

Query Parameters:

• url (string, required): The target URL to visit.
• retries (integer, default: 5): Number of attempts to bypass Cloudflare.
• proxy (string, optional): A proxy string to use for the browser session.

Response Details:

• Body: The raw HTML content of the page (text/html).
• Headers:
  • cookies: A JSON-encoded string of the session cookies (e.g., {"name": "value"}).
  • user_agent: The User-Agent string used by the browser.

GET /html?url=https%3A%2F%2Fexample.com

The /cookies GET endpoint allows you to bypass Cloudflare protection for a specific URL and retrieve the resulting session cookies and the browser's User-Agent. This is useful for authenticating subsequent requests to a protected API.

Query Parameters:

• url (string, required): The target URL to visit.
• retries (integer, default: 5): Number of attempts to bypass Cloudflare.
• proxy (string, optional): A proxy string to use for the browser session.

Response Format: Returns a CookieResponse object containing:

• cookies: A dictionary of cookie names and values.
• user_agent: The User-Agent string used by the browser.

GET /cookies?url=https%3A%2F%2Fexample.com&retries=5&proxy=http%3A%2F%2Fproxy%3A8080

The CookieResponse Pydantic model defines the structure of the JSON returned by the /cookies endpoint.

{
  "cookies": {
    "cookie_name": "cookie_value"
  },
  "user_agent": "Mozilla/5.0 ..."
}

The server can be started from the command line using Python. It uses uvicorn to host the FastAPI application.

Environment Variables:

• DOCKERMODE: Set to true to enable Docker-specific Chromium configurations (e.g., --no-sandbox, --remote-debugging-port=9222).
• SERVER_PORT: The port the server will listen on (default: 8000).

CLI Arguments:

• --nolog: Disables logging during the bypass process.
• --headless: Runs the browser in headless mode (note: if DOCKERMODE is true, it will use pyvirtualdisplay to simulate a display).

python dsk/server.py --headless --nolog