AI SWOT: Practical demonstration of the System Prompt Learning paradigm.

中文说明

SWOT (meaning "Small-Town Overachiever") is an AI training system focused on Self-Prompt Training, following the System Prompt Learning paradigm. It provides a comprehensive platform for users to train and manage AI models through the following features.

Core Features

• Training and Answering:
  • Users can load question sets, and the AI will answer based on the provided notes.
  • The system records the AI's answering performance, including accuracy, error analysis, etc.
  • Users can control the training process, such as starting and pausing training, and adjusting training parameters.
• Note Management:
  • Current Notes: The system displays notes autonomously learned and recorded by the AI during the problem-solving process. Users can view these notes and observe how the AI iterates and optimizes its notes based on problem-solving results and error analysis.
  • Note History: The system automatically saves historical versions of AI notes. Users can easily view the evolution of notes and restore to previous versions if needed.
  • Note Import/Export: Supports importing and exporting note data.
• Prompt Configuration:
  • Users can edit and manage prompt templates used in various stages of the training process to optimize AI training effectiveness.
• Question Bank Configuration:
  • Users can manage question bank data for training or testing, supporting the import of processed data.
• Model Interface Configuration:
  • Manage AI model providers, API keys, and selected models.
• Conversation History:
  • Saves historical conversation records with the AI model, allowing users to easily review and analyze previous interactions.
• Storage Management:
  • View and manage various data stored locally by the system, including trainer status, question sets, prompt templates, etc. Supports data import and export.
• Debugging Tools:
  • Provides a series of debugging tools for developers to perform data operations and status checks.

Design Philosophy

The SWOT system aims to enhance AI models' capabilities in specific knowledge domains by simulating a cycle of "problem-solving - learning - improving notes - problem-solving again." Users provide question sets, and the AI autonomously records and iterates on its notes during the problem-solving process. Through this learning cycle, the goal is to ultimately improve the model's performance. This design philosophy aligns with the System Prompt Learning paradigm proposed by Andrej Karpathy.

Tech Stack

• Vue.js (Frontend Framework)
• PrimeVue (UI Component Library)
• TypeScript (Programming Language)
• Vite (Build Tool)
• UnoCSS (CSS Engine)

How to Run

Please refer to the HOW_TO_RUN-cn.md (Chinese) or HOW_TO_RUN-en.md (English) files in the project for detailed running instructions. The project provides various startup scripts, such as start-project-macos-en.sh, start-project-linux-en.sh, start-project-en.bat, etc., for launching the project on different operating systems.