Gemini Proxy API Documentation

This document explains how to use the Gemini Proxy API, hosted at https://gemini-proxy.hdanny.org. The API provides an interface for interacting with the Gemini generative model, allowing users to start a chat session, send messages, and end a chat session. Example Python code demonstrates usage.

---

Base URL

https://gemini-proxy.hdanny.org

---

Endpoints

1. Start a Chat

Endpoint: /start_chat Method: POST

This endpoint initializes a chat session using the specified model and configuration.

Request Parameters

• user_id (string, required): The ID of the user starting the chat. Must be a valid user as defined in valid_users.txt. If you don't know the user ID, please contact the administrator (Danny Y. Huang).
• model_name (string, optional): The name of the generative model to use (default: "gemini-1.5-flash-002").
• generation_config (object, optional): Configuration for the model's behavior. Example:

  {
      "max_output_tokens": 1000,
      "temperature": 0.7,
      "top_p": 0.9
  }

Response

• 200 OK: Returns the unique chat_id for the session.
• 403 Forbidden: Invalid or missing user_id.
• 500 Internal Server Error: Error in processing the request.

Example Request

url = "https://gemini-proxy.hdanny.org/start_chat"
payload = {
    "user_id": "user1",
    "model_name": "gemini-1.5-flash-002",
    "generation_config": {
        "max_output_tokens": 1000,
        "temperature": 0.7,
        "top_p": 0.9
    }
}
response = requests.post(url, json=payload)
chat_id = response.json()["chat_id"]
print(chat_id)

---

2. Send a Message

Endpoint: /send_message Method: POST

This endpoint sends a message to the active chat session.

Request Parameters

• chat_id (string, required): The unique ID of the chat session, returned from /start_chat.
• message (string, required): The message to send to the model.

Response

• 200 OK: Returns the model's response to the message.
• 404 Not Found: Invalid or missing chat_id.
• 429 Too Many Requests: Rate limit exceeded (at most 2 requests per second per user).
• 500 Internal Server Error: Error in processing the request.

Example Request

url = "https://gemini-proxy.hdanny.org/send_message"
payload = {
    "chat_id": "chat_id_from_start_chat",
    "message": "What is the capital of California?"
}
response = requests.post(url, json=payload)
model_response = response.json()["response"]
print(model_response)

---

3. End a Chat

Endpoint: /end_chat Method: POST

This endpoint terminates an active chat session.

Request Parameters

• chat_id (string, required): The unique ID of the chat session.

Response

• 200 OK: Confirms the chat session has ended.
• 404 Not Found: Invalid or missing chat_id.
• 500 Internal Server Error: Error in processing the request.

Example Request

url = "https://gemini-proxy.hdanny.org/end_chat"
payload = {"chat_id": "chat_id_from_start_chat"}
response = requests.post(url, json=payload)
end_message = response.json()["message"]
print(end_message)

---

Rate Limits

• Each user is limited to 2 send_message calls per second.
• Exceeding the rate limit returns a 429 Too Many Requests error.

---

Logging

All interactions with the API are logged in JSON format. Each log entry includes:

• timestamp: ISO 8601 timestamp of the interaction.
• user_id: The ID of the user making the request.
• message: The message sent by the user.
• output_length: The length of the model's response.

Example log entry:

{
    "timestamp": "2024-12-04T10:15:30.123Z",
    "user_id": "user1",
    "message": "What is the capital of California?",
    "output_length": 42
}

---

Example Usage in Python

The following Python script demonstrates a complete workflow: starting a chat, sending messages, and ending the chat. Note that the code below should work out-of-the-box, provided that you have the valid user_id. If you don't have a user_id, please contact the administrator (Danny Y. Huang).

import requests

BASE_URL = "https://gemini-proxy.hdanny.org"

# Step 1: Start a Chat
def start_chat(user_id, model_name, generation_config):
    url = f"{BASE_URL}/start_chat"
    payload = {
        "user_id": user_id,
        "model_name": model_name,
        "generation_config": generation_config,
    }
    response = requests.post(url, json=payload)
    response.raise_for_status()
    return response.json()["chat_id"]

# Step 2: Send a Message
def send_message(chat_id, message):
    url = f"{BASE_URL}/send_message"
    payload = {
        "chat_id": chat_id,
        "message": message,
    }
    response = requests.post(url, json=payload)
    response.raise_for_status()
    return response.json()["response"]

# Step 3: End the Chat
def end_chat(chat_id):
    url = f"{BASE_URL}/end_chat"
    payload = {"chat_id": chat_id}
    response = requests.post(url, json=payload)
    response.raise_for_status()
    return response.json()["message"]

# Example Usage
if __name__ == "__main__":
    USER_ID = "[ENTER_YOUR_USER_ID]"
    MODEL_NAME = "gemini-1.5-flash-002"
    GENERATION_CONFIG = {
        "max_output_tokens": 1000,
        "temperature": 0.7,
        "top_p": 0.9,
    }

    try:
        # Start a chat
        chat_id = start_chat(USER_ID, MODEL_NAME, GENERATION_CONFIG)
        print(f"Chat started with ID: {chat_id}")

        # Send messages
        response1 = send_message(chat_id, "What is the capital of California?")
        print(f"Response 1: {response1}")

        response2 = send_message(chat_id, "What about New Jersey?")
        print(f"Response 2: {response2}")

        # End the chat
        end_message = end_chat(chat_id)
        print(f"Chat ended: {end_message}")
    except requests.exceptions.RequestException as e:
        print(f"Error: {e}")

---

Notes

• Ensure the valid_users.txt file contains your user ID before making requests.
• For secure interactions, ensure HTTPS is properly configured and certificates are valid.
• If rate limits are exceeded, consider introducing a delay between consecutive send_message calls.

---

This documentation should help you effectively interact with the Gemini Proxy API.