PallasBot/Pallas-Bot

What it solves

Pallas-Bot provides a social chatbot for group chats that blends traditional "repeater" behavior (mimicking user speech) with modern Large Language Model (LLM) capabilities. It solves the problem of generic AI bots feeling out of place in casual group settings by learning the specific linguistic style and common phrases of a particular community to make its interactions feel more natural and integrated.

How it works

The bot uses a hybrid approach to interaction:

• Learning Repeater: It analyzes group chat corpora to learn how to automatically respond, repeat phrases, and use emojis based on the community's actual behavior.
• Semantic Style Guidance: It tracks what is commonly said in specific contexts across different accounts and groups, using this data to guide the LLM's output so the AI matches the group's expression style.
• LLM Integration: When explicitly mentioned, it uses a Bot Provider for multi-turn conversations, utilizing long-term memory, knowledge retrieval, web search, and tool calling.
• Corpus Federation: It can connect to a community-shared corpus pool to expand its vocabulary of short responses while avoiding duplicate replies across multiple deployments.

Who it’s for

Users who want a customizable, personality-driven chatbot for group chats (specifically those using OneBot v11 protocols like NapCat) that can act as both a quirky repeater and a sophisticated AI assistant.

Highlights

• Hybrid AI/Repeater Engine: Combines corpus-based mimicry with LLM-driven conversation.
• Contextual Style Learning: Injects group-specific expression styles into LLM prompts to avoid a "generic AI" feel.
• Comprehensive Management: Includes a WebUI console for managing configurations, plugins, and AI observability.
• Extensible Ecosystem: Supports NoneBot plugins and features an internal plugin store.
• High Performance: Uses a background task delivery system and multi-worker sharding to handle high concurrency without blocking messages.