Ai-chan - The Romantic AI Companion

Ai-chan is an advanced Line Messaging Bot that transcends simple automated responses. By integrating state-of-the-art Large Language Models (LLMs) with real-time sentiment analysis, Ai-chan provides a warm, empathetic, and context-aware conversational experience for users seeking companionship and emotional support.

🌟 Core Features

• Empathetic Dialogue Engine: Utilizes Google Gemini 1.5 Flash (with multi-version fallback) to generate natural, comforting, and personalized responses.
• Intelligent Sentiment Analysis: Automatically detects user emotions—such as joy, sadness, anger, stress, or boredom—to tailor the tone and personality of the conversation.
• Contextual Awareness: Maintains a conversational memory (up to 6 rounds of dialogue) to ensure continuity and depth in long-term interactions.
• Comprehensive Admin Dashboard: A full-featured management portal for monitoring conversation metrics, user engagement statistics, and sentiment distribution.
• Robust Error Handling: Implements graceful degradation with keyword-matching fallbacks to ensure availability even during LLM service interruptions.

🛠️ Technical Architecture

Tech Stack

• Framework: Next.js with TypeScript for a robust full-stack foundation.
• Messaging Interface: Line Messaging API for seamless mobile integration.
• Storage: MongoDB Atlas with Mongoose ODM for scalable conversation logging.
• AI/ML: Google Gemini API for natural language understanding and generation.
• Validation: Zod for strict schema validation across API boundaries.

System Workflow

1. Reception: User message is received via the Line Webhook.
2. Analysis: The message is processed by the sentiment engine to identify the user's emotional state.
3. Synthesis: The LLM generates a response using a dynamic prompt that incorporates the detected emotion and recent conversation history.
4. Delivery: The synthesized response is sent back through the Line Bot SDK while the transaction is logged for analytical purposes.

📊 Management & Monitoring

The integrated dashboard provides real-time insights into:

• Total conversation volume and message frequency.
• Heatmaps of user sentiment trends.
• Detailed message logs for quality assurance and system tuning.

This project was developed as part of the Web Services course (HW6) at National Taiwan University.