Skip to content

OpenAI_Server.md

Latest commit

 

History

History
181 lines (122 loc) · 4.53 KB

OpenAI_Server.md

File metadata and controls

181 lines (122 loc) · 4.53 KB

🧠 Grok3API OpenAI-Compatible Server

🤖 Early stage of server development

The first draft of a server compatible with the OpenAI API (/v1/chat/completions) has been implemented.

⚠️ It's in the early stage and not fully debugged. Use it at your own risk!

⚙️ Running the Server

Make sure you have fastapi, uvicorn, pydantic, and grok3api installed.

# From the project root:
python -m grok3api.server
# Or you can configure the host and port
python -m grok3api.server --host 127.0.0.1 --port 9000

🎉 The server will start at http://127.0.0.1:9000.

Uses uvicorn under the hood. Make sure the project structure allows for module-based execution (via -m).

🔁 Endpoint: /v1/chat/completions

Compatible with the OpenAI request format:

📥 Example request:

from openai import OpenAI
from openai import OpenAIError

# Creating a client with the specified URL and key
client = OpenAI(
    base_url="http://localhost:9000/v1",
    api_key="dummy"
)

try:
    # Sending a request to the server
    response = client.chat.completions.create(
        model="grok-3",
        messages=[{"role": "user", "content": "What is Python?"}]
    )

    # Displaying the response from the server
    print("Server Response:")
    print(f"Model: {response.model}")
    print(f"Message: {response.choices[0].message.content}")
    print(f"Finish Reason: {response.choices[0].finish_reason}")
    print(f"Usage: {response.usage}")

except OpenAIError as e:
    print(f"Error: {e}")
except Exception as e:
    print(f"Unexpected error: {e}")

stream is not supported, a 400 Bad Request will be returned.

📤 Output:

Server Response:
Model: grok-3
Message: Python is a high-level, interpreted programming language with a simple and readable syntax. It supports multiple programming paradigms (object-oriented, functional, procedural) and is widely used for web development, data analysis, task automation, machine learning, and more. Python is popular for its versatility, extensive standard library, and large developer community.
Finish Reason: stop
Usage: CompletionUsage(completion_tokens=46, prompt_tokens=3, total_tokens=49, completion_tokens_details=None, prompt_tokens_details=None)

🧵 Endpoint: /v1/string (GET)

A simple text GET request. Accepts a string as a query parameter and returns a response from Grok without JSON wrapping.

📥 Example Request:

Open in a browser or use curl:

curl http://localhost:9000/v1/string?q=Hello

📤 Example Response:

Hello! How can I help?

🐍 Example in Python:

import requests

response = requests.get("http://localhost:9000/v1/string", params={"q": "Tell a joke"})
print(response.text)  # Just text, no JSON

🧵 Endpoint: /v1/string (POST)

A simple text POST request. Accepts a string in the request body and returns a response from Grok without JSON wrapping.

📥 Example Request:

Use curl or another tool:

curl -X POST http://localhost:9000/v1/string -d "Hello"

📤 Example Response:

Hello! How can I help?

🐍 Example in Python:

import requests

response = requests.post("http://localhost:9000/v1/string", data="Tell a joke")
print(response.text)  # Just text, no JSON

⚙️ Environment Variables (optional)

Variable Description Default Value
GROK_COOKIES Cookies file for GrokClient None
GROK_PROXY Proxy (e.g., http://localhost:8080) None
GROK_TIMEOUT Grok request timeout (in seconds) 120
GROK_SERVER_HOST IP address for running the server 0.0.0.0
GROK_SERVER_PORT Port for running the server 8000

📂 Project Structure

Grok3API/
├── grok3api/
│   ├── server.py          # <--- running the server
│   ├── client.py
│   ├── types/
│   └── ...
├── tests/
│   └── openai_test.py     # <--- compatibility test
└── README.md

❗ TODO / Known Issues

  • 🔄 Streaming support (stream=True)
  • 🧪 More tests and validation
  • 🧼 Refactor message_payload and history logic
  • 🧩 Custom instructions, images, and additional features

💬 For questions and suggestions — feel free to open an issue!