This project provides an open-source game, LLM Riddles, designed to explore prompt engineering and natural language processing. It challenges users to craft specific prompts to elicit desired responses from large language models, targeting enthusiasts and developers interested in AI interaction.

How It Works

LLM Riddles operates by presenting users with a series of challenges where they must engineer prompts to achieve specific LLM outputs. The game leverages various LLM APIs (ChatGPT, ChatGLM, DeepSeek, Mistral-7B) and supports both Chinese and English language interactions. This approach allows users to directly experiment with prompt construction and observe the nuanced behavior of different models.

Quick Start & Requirements

• Install: pip3 install -r requirements.txt
• Prerequisites: Python 3, API keys for supported LLMs (ChatGPT, ChatGLM, DeepSeek). Mistral-7B can be run locally with pip3 install -r requirements-dev.txt.
• Online versions are available via Hugging Face Spaces and OpenXLab.
• Documentation: Technical blog

Highlighted Details

• Supports multiple LLMs including ChatGPT, ChatGLM, DeepSeek, and Mistral-7B.
• Offers both Chinese and English language support for game interactions.
• Allows users to submit custom level designs via Pull Requests.
• Provides online demos for immediate play.

Maintenance & Community

• Active development with a roadmap including support for Baichuan2-7B and LLaMA2-7B.
• Feedback and contributions are encouraged via GitHub Issues.
• Contact: opendilab@pjlab.org.cn

Licensing & Compatibility

• License: Apache License 2.0.
• Permissive license suitable for commercial use and integration with closed-source projects.

Limitations & Caveats

The project is primarily focused on prompt engineering challenges and does not currently include features for LLM inference speed optimization or advanced game mechanics. Support for additional models like LLaMA2-7B is planned but not yet implemented.