GitHub - rNetAi/rnet-oauth-python: Python library for rNe...

A Python backend library for integrating RNet OAuth and AI Provider services. This library allows users to authenticate via RNet and pay for AI model token costs directly using their RNet account.

Features

OAuth2 PKCE Support: Secure authorization code flow with automatic code verifier and challenge generation.
Token Management: Exchange authorization codes for tokens and refresh expired tokens.
UserInfo Endpoint: Fetch the authenticated user's RNet profile with an access token.
AI Integration: Easy methods to chat with AI models using standard or streaming responses.

Installation

Quick Start

1. Initialize the Clients

from rnet_oauth import RNetAuth, RNetAi

auth = RNetAuth(
    client_id='client-id',
    client_secret='client-secret',
    redirect_uri='redirect-uri'
)
ai = RNetAi()

2. Generate Authorization URL (OAuth2 PKCE)

# 1. Generate PKCE
pkce = auth.generate_pkce()
verifier = pkce['verifier']
challenge = pkce['challenge']

# 2. Get Authorization URL
# challenge: PKCE code challenge (optional)
# state: An optional string to maintain state between the request and callback (recommended for security)
auth_url = auth.get_authorization_url(challenge, state='optional-state')

3. Exchange Code for Tokens

# 3. Exchange code for tokens
tokens = auth.exchange_code_for_token(code, verifier)
access_token = tokens['access_token']

4. Get User Info

user_info = auth.get_user_info(access_token)
print(user_info['email'])
print(user_info['name'])

The UserInfo response comes from RNet's /userinfo endpoint and may include: sub, email, email_verified, name, preferred_username, user_id, role, and status.

5. Chat with AI

response = ai.chat({
    "contents": [
        {
            "role": "user",
            "parts": [{"text": "Hello!"}]
        }
    ]
}, access_token, "gemini-2.5-flash-lite")

6. Streaming AI Response (Untested)

for chunk in ai.chat_stream({
    "contents": [
        {
            "role": "user",
            "parts": [{"text": "Hello!"}]
        }
    ]
}, access_token, "gemini-2.5-flash-lite"):
    print(chunk)

7. File Upload (Untested)

with open("document.pdf", "rb") as f:
    file_buffer = f.read()

# Upload to Gemini
gemini_upload = ai.gemini_file_upload(access_token, "gemini-2.5-flash-lite", file_buffer, "application/pdf", "document.pdf")
print(gemini_upload['fileReference']) # Use this in chat payload

# Upload to OpenAI
openai_upload = ai.openai_file_upload(access_token, "gpt-4o", file_buffer, "application/pdf", "document.pdf")

8. File Deletion (Untested)

# Gemini files auto-delete after 48 hours, so there is no delete method.
# Delete an OpenAI file:
ai.openai_file_delete(access_token, "gpt-4o", openai_upload['fileReference'])

9. AI Chat with File & Tools (Untested)

payload = {
    "contents": [
        {
            "role": "user",
            "parts": [
                { "text": "Based on this document, what is my name? Also search the web for the current weather in London." },
                { "fileData": { "fileUri": gemini_upload['fileReference'], "mimeType": gemini_upload['mimeType'] } }
            ]
        }
    ],
    "tools": [
        { "googleSearch": {} } # Enable Google Search tool
    ]
}

response = ai.chat(payload, access_token, "gemini-2.5-flash-lite")
print(response)

License

MIT

推荐订阅源

Show HN