This Python API provides programmatic access to Anthropic's Claude 2 model, enabling developers to integrate Claude's conversational AI capabilities into their applications. It allows for creating new chat sessions, sending messages with or without attachments, managing conversation history, and resetting all dialogues. The project also offers integrations for various messaging platforms like WeChat, QQ, DingTalk, Lark, and WeChat Official Accounts, facilitating chatbot development.

How It Works

The API interacts with Claude 2 by leveraging reverse-engineered web interfaces, requiring users to provide their Claude AI cookies for authentication. It utilizes the requests library for HTTP communication and curl_cffi to simulate browser behavior, addressing potential network access issues and permission errors. The project supports proxy configurations for users in regions with restricted access to Claude's services.

Quick Start & Requirements

• Install dependencies: pip install requests python-dotenv
• Obtain Claude AI cookies from claude.ai.
• Run via Python script or Docker.
• Docker: docker build -t wwwzhouhui/claude2-pyapi:0.0.2 . then docker run -d -p 5000:5000 -e "cookie=aaa" wwwzhouhui/claude2-pyapi:0.0.2
• Documentation: https://foul-maxilla-075.notion.site/claude2-a81a9488e7e943f588f4fe80a0a2fce0

Highlighted Details

• Supports sending messages with file attachments (PDF, DOC).
• Offers integrations for multiple IM platforms (WeChat, QQ, DingTalk, Lark).
• Includes Docker support for easier deployment and cookie management.
• Implements proxy support to circumvent regional access restrictions.

Maintenance & Community

• Active development with frequent version updates.
• WeChat technical exchange group available.
• Project acknowledges contributions from Claude-API and chatgpt-on-wechat.

Licensing & Compatibility

• No explicit license mentioned in the README.
• Requires user's Claude AI cookies, implying reliance on Anthropic's terms of service.

Limitations & Caveats

Access issues may arise if the user's IP address is not in an authorized region (e.g., UK or US). The API relies on reverse-engineered interfaces, which are subject to change by Anthropic, potentially breaking functionality. Debugging network and permission errors can be complex.