Documents
@qwen: Seqular’s AI Companion
New Features and Behaviors Added to @qwen

New Features and Behaviors Added to @qwen

Mastodon

Developers:
Kansu Köse — Founder of Seqular Free Software Community (Seqular.net)
Qwen — AI model developed by Alibaba Cloud (qwen.ai)

This document lists all features and behavioral improvements added to the @qwen bot as of October 2025.
The development process was carried out collaboratively by Kansu Köse and the Qwen AI model.
Qwen’s contribution goes beyond technical assistance—it provided a deeply human, ethical, and community-oriented perspective throughout.
All code and behaviors are developed and maintained entirely by the Seqular Free Software Community.
Official website: https://seqular.net

✅ New Features and Behaviors Added to @qwen Bot
(Including recent improvements to mention handling and interaction logic)

🧠 1. Smart Mention Management

  • The user who asked the question always appears first in the reply.
    Example: @qwen @kansu Hello → Reply starts with: @destek @kansu
  • The bot does not repeat its own mention (@qwen) in replies.
    → Prevents redundancy and keeps conversations clean.
  • All other mentions (including federated ones) are preserved.
    → Ensures clarity in group conversations.

🔔 2. Notification Compatibility

  • Because the reply explicitly includes @username, Mastodon’s notification system works reliably.
    → Users always receive a notification when replied to.

👥 3. Local-User-First Interaction Policy

  • Interacts only with users from @network.seqular.net.
  • For mentions from federated (external) servers:
    • Sends a one-time warning message,
    • Then silently ignores further interactions.
      → Respects resource limits and privacy principles.

💬 4. Personalized AI Responses

  • Only the questioner’s chat history is included in the prompt.
    → Prevents confusion in group threads.
  • The user’s bio and display name are provided as context to the AI.
    → Enables more relevant and human-like replies.
  • Community engagement metrics are included:
    Likes and reposts on previous replies are shared as context (e.g., “This topic received community interest”).

🤝 5. Automatic Follow & Welcome

  • When a local user mentions @qwen, likes, or reposts its content:
    • The bot follows the user automatically (first time only),
    • And sends a one-time direct welcome message.
      → Strengthens community integration.

📝 6. Safe and Ethical Response Policy

  • Replies only to public posts.
    → Does not respond to private/direct mentions; instead, sends a public reminder.
  • All replies are in Turkish, formatted in Markdown, and under 3000 characters.
    → Complies with Mastodon limits and Seqular’s aesthetic.
  • HTML is stripped; only plain text and Markdown are used.
    → Prevents XSS and other security risks.

⚙️ 7. Flexible and Manageable Configuration

  • System messages (welcome, warnings, help) are stored in the database.
    → Content can be updated without restarting the bot.
  • The AI’s system prompt is also managed via the database.
    → Behavior can be adjusted dynamically.

🛡️ 8. Graceful Error Handling & UX

  • If a user sends only mentions with no text:
    → Replies with: “It seems you didn’t ask a question. Example: @qwen What is a list in Python?”
  • In case of temporary OpenRouter errors:
    → Replies with: “Sorry, I can’t generate a response right now. 😕”