Papyrus is a minimalist, efficient, keyboard-driven AI Agent-powered SRS (Spaced Repetition System) review engine focused on high-intensity model memorization.

"The Great Way is simple."

• 🚀 Minimalist Interaction: Fully keyboard-driven workflow, no mouse needed. Helps you enter the "flow state" of deep review.
• 🧠 Intelligent Scheduling: Built-in simplified spaced repetition algorithm based on the Ebbinghaus forgetting curve logic. Automatically schedules next review based on mastery level.
• 📦 Lightweight & Portable: Single .exe file, zero dependencies. Data stored locally in Papyrusdata.json for privacy and security.

Prepare a UTF-8 encoded .txt file with the following format:

Model scenario or question A === Core trigger or answer A

Model scenario or question B === Core trigger or answer B

Tip: Each card is separated by ===. Leave an empty line between groups for clarity.

Go to the Releases page to download the latest version.

1. Extract the downloaded .zip file to any folder.
2. Double-click Papyrus.exe to launch.

Note: Do not move the .exe alone to other locations, otherwise the program cannot load icons or read previous review progress (v1.0.0 only).

Click the "⚙️ Settings" button in the sidebar and enter your API Key:

• OpenAI: Get at https://platform.openai.com/api-keys
• Anthropic: Get at https://console.anthropic.com/
• Ollama: Runs locally, no API Key needed

  # Install Ollama
  # Download: https://ollama.ai
  
  # Pull model
  ollama pull llama2

Agent Mode: AI uses tool calls to add, edit, and delete cards
Chat Mode: Chat only

In settings, you can adjust:

• Temperature: Controls creativity (0-2, higher = more random)
• Max Tokens: Maximum response length

AI configuration is saved in data/ai_config.json, including:

• API Key
• Model selection
• Parameter settings

1. API Costs: OpenAI and Anthropic charge by usage. Set a budget.
2. Local Model: Ollama is completely free but requires better hardware.
3. Network: Cloud APIs require stable internet.
4. Privacy: Local model data won't be uploaded; cloud APIs send question content.

AI Features Not Working

• Check if requests library is installed
• Check console error messages

API Call Failed

• Check if API Key is correct
• Check network connection
• Confirm API quota is sufficient

Ollama Connection Failed

• Confirm Ollama service is running
• Check if port is 11434
• Try: ollama serve

• Development Mode: Human-led, AI-assisted (Claude Sonnet 4.6, Gemini 3.0 Pro)
• Language: Python 3.8 (Tkinter)
• Engineering: Packaged with PyInstaller, handled sys._MEIPASS resource path compatibility
• License: MIT License
• Credits: Icon material from Minecraft Wiki (Mojang Studio)

The birth of this project wasn't ideal or grand.

Why make this? One day I was studying AI and downloaded Anki to memorize knowledge points. After downloading, it needed to check for updates, which required connecting to a server, but I couldn't connect.

Then I noticed Gemini was right there, so I asked it to write one for me. After some tweaking, I started using it.

I actually didn't plan to release this. Why did I eventually?

Yesterday, I used an old account to star my girlfriend's repo. GitHub identified me as a bot and removed my activity. I was furious! I use AI, but I'm not AI. To prove I'm not AI, I happened to have this on hand, so I polished it and put it here.

Story's over. It's already the 6th day of the Lunar New Year, dawn is breaking soon. Poor student, sigh, can only vent here.

I love a saying: "Knowledge makes you free" (Veritas vos liberabit). To all the great developers, this is my first open-source project. Please be kind.

Consider this a belated New Year greeting.
May you, in the new year, wield the blade of knowledge to pierce the ruins of darkness and reach the realm of dawn.

• Quick Start Guide - Get started in 5 minutes
• Version Info - Current version and updates
• Changelog - Complete version history
• Project Structure - Code organization
• AI Features - AI assistant guide
• AI Tools Demo - Real-world use cases