NickCharlie/astrbot_plugin_self_learning

What it solves

This plugin enables an AstrBot instance to autonomously learn from group chat interactions. It solves the problem of static bot personalities by allowing the bot to evolve its expression style, understand group-specific slang (jargon), track social relationships, and develop long-term memory based on real-time conversations.

How it works

The plugin intercepts messages at three key stages: when a message is received, after the bot sends a response, and before the next LLM request. It uses a pipeline to capture raw messages, extract expression patterns and jargon, and analyze social dynamics. These learned elements are stored in a database (PostgreSQL, MySQL, or SQLite) and pass through a review queue where an administrator can approve or reject the learned traits. Once approved, the plugin injects these styles, slang definitions, and social contexts as temporary context parts into the LLM request to shape the bot's response.

Who it’s for

Users of the AstrBot framework who want their AI agents to feel more integrated into specific social circles, adopt the local vernacular of a chat group, and maintain a persistent, evolving persona.

Highlights

• Style & Jargon Learning: Extracts expression patterns and high-frequency group slang, providing the LLM with definitions to ensure the accurate use of local terminology.
• Persona Evolution Review: A human-in-the-loop system where learned personality traits and styles must be reviewed and approved before being applied.
• Social Graphing: Tracks interaction relationships, affection levels, and psychological states within the group.
• Knowledge & Memory Graphs: Builds and visualizes memory and knowledge graphs for structured information retrieval.
• Comprehensive Dashboard: A WebUI for managing settings, reviewing learning results, monitoring model calls, and and visualizing graphs.
• Multi-DB Support: Supports PostgreSQL (default), MySQL, and SQLite with automatic table creation and lightweight migration.