LINE-Agent

Requirements

• Get LINE_CHANNEL_ACCESS_TOKEN and LINE_CHANNEL_SECRET. Tutorial
• Fill up .env information
• Put your domain knowledge in MongoDB. Tutorial

LLM Multi-Agent System with Semantic Kernel, MongoDB, and Redis

This repository hosts a multi-agent system powered by Large Language Models (LLMs), built using Microsoft's Semantic Kernel, MongoDB, and Redis. Designed for real-time interaction on LINE, our system coordinates multiple intelligent agents to handle diverse tasks such as information retrieval, conversation management, and personalized responses.

Key Components

• Semantic Kernel: Orchestrates and coordinates agent skills, memory, and planning logic.
• MongoDB: Serves as persistent storage for user profiles, chat history, and agent metadata.
• Redis: Enables fast, in-memory storage for session management and context tracking across conversations.
• LINE Integration: Delivers multi-agent interaction capabilities directly to users via LINE messaging.

This system is built for scalability, modularity, and extensibility—ideal for research, prototyping, or production use in real-world LLM-powered assistant applications.

How to use:

• Step 1. Terminal one: docker-compose up --build
• Step 2. Terminal two: ngrok http 8501
• Step 3. Fill up LINE Webhook URL: https://xxxx-xxx-xx-xxx-xx.ngrok-free.app/webhook

Acknowledge

• semantic-kerne