LLM Provider

A Python library for interacting with various LLM providers (ChatGPT, Claude) and AI video generation services (Sora) through browser automation. This library provides a unified interface for managing browser profiles and interacting with different AI services.

Features

• Free API: Access ChatGPT and Claude through their web interfaces—ideal for users with budget constraints.
• Higher TPM: Web interfaces often have higher tokens-per-minute (TPM) limits compared to basic API tiers, which may allow for longer prompts.
• Automation: For services that don’t have an official API (e.g., Sora), this repository can offer a temporary Sora API to facilitate automation until an official API is published.

Service Support

• ChatGPT
• Claude
• Sora

Prerequisites

• Chrome Browser: Currently, only Chrome is fully supported and tested.
• Python Version:
  • If you’re using Python 3.12 or higher, you may encounter issues with undetected-chromedriver.
  • To resolve this, please refer to this GitHub issue or using Python 3.11 or lower until the compatibility issue is resolved.
• Accounts: You must have valid accounts for the services you want to use (ChatGPT, Claude, or Sora).
• Chrome Profile Setup:
  1. Create a new Chrome profile specifically for this automation
  2. Log in to your accounts (ChatGPT/Claude/Sora) manually in this profile
  3. Ensure you stay logged in by checking Remember me or similar options if available
  4. Disable the Ask where to save each file before downloading option in Chrome so that Sora can download files automatically.
  5. The library will use this profile for automation

Installation

1. Clone the repository:

git clone https://github.com/yourusername/llm-provider.git
cd llm-provider

2. Install required dependencies:

pip install -r requirements.txt

Configuration

The library uses a YAML-based configuration system. Default settings are provided in llm_provider/default_config.yaml, which can be overridden by providing a custom configuration file.

Browser Configuration

chrome:
  use_undetected: true
  load_profile: true
  user_data_dir: "User Data"
  profile_directory: "Profile {number}"

XPath Configuration

May need to change the xpath if frontend changed, but should careful of absolut paht or relative path.

Usage

• For detailed usage examples, please refer to the examples directory.

ChatGPT Example

from llm_provider import CONFIG, update_config, get_provider_config
from llm_provider.browsers import ChromeBrowser
from llm_provider.providers import ChatGPTProvider, ClaudeProvider

# Retrieve configuration for ChatGPT
gpt_config = get_provider_config('chatgpt')

# Initialize a Chrome browser instance
chrome_browser = ChromeBrowser()

# Create a ChatGPT provider using the browser and config
chatgpt = ChatGPTProvider(browser=chrome_browser, config=gpt_config)

# Select model
chatgpt.select_model('gpt-4o')

# Send a message
success = chatgpt.send_message_safely('Hello GPT! How are you today?')
print(f'Message sent successfully? {success}')

wait_status = chatgpt.wait_for_response_completion()
response_dict = chatgpt.get_response()
print('GPT Response (Text):', response_dict['chat']['text'])
print('GPT Response (Markdown):', response_dict['chat']['markdown'])

chatgpt.browser.close()

Claude Example

from llm_provider import CONFIG, update_config, get_provider_config
from llm_provider.browsers import ChromeBrowser
from llm_provider.providers import ChatGPTProvider, ClaudeProvider

# Retrieve configuration for Claude
claude_config = get_provider_config('claude')

# Initialize a Chrome browser instance
chrome_browser = ChromeBrowser()

# Create a Claude provider using the browser and config
claude = ClaudeProvider(browser=chrome_browser, config=claude_config)

# Select model
claude.select_model('sonnet')

# Send a message
success = claude.send_message('Hello Claude! How are you today?')
print(f'Message sent successfully? {success}')

wait_status = claude.wait_for_response_completion()
response_dict = claude.get_response()
print('Claude Response (Text):', response_dict['chat']['text'])
print('Claude Response (Markdown):', response_dict['chat']['markdown'])

claude.browser.close()

Sora Example

from llm_provider.browsers import ChromeBrowser
from llm_provider.directors import SoraDirector
from llm_provider.config import CONFIG

# Initialize browser and director
browser = ChromeBrowser()
director = SoraDirector(browser, CONFIG['sora'])

# Configure video settings
director.update_aspect_ratio('16:9')
director.update_resolution('480p')
director.update_duration(15)
director.update_variation(2)

# Create video
director.create_video('A cinematic shot of a serene lake at sunset')
director.browser.random_time_delay(30, 30)
director.download_videos()

browser.close()

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details

Disclaimer

This library is for educational purposes only. Please ensure you comply with the terms of service of all AI services when using this library.