A demo of my AI Chatbot is now currently available at https://aura.happymonkey.ai for users to test. * It is in a beta state though and subject to change, And may go offline at any time. Please be aware that it is currently in active development, which means you may encounter bugs, and features may change. Your data may be wiped during updates. I made a YouTube video about it 👇
Overview
I recently built an AI Chatbot as a personal project that had been on my mind for a while. It’s fairly advanced and includes several key features:
- Dynamic thought system that combines user input with external sources (e.g., BBC news feeds) it ingests regularly.
- Mood scoring that generates a positive or negative emotional state, producing a weighted emotional score that feeds back into the questions it asks.
- Proactive “push” system that asks users questions to learn about them—collecting key facts such as name, interests, etc.—and stores this data in a vector database and JSON files.
- Memory of past interactions, allowing the bot to recall previous conversations, user preferences (name, favorite colour, etc.), and broader world knowledge, then use that information to shape future dialogue.
- Idle detection: if a user doesn’t respond, the bot checks “Are you still there?” and, after no reply, says “Okay, I’ll wait.”
The goal was a more human‑like chatbot that interacts naturally, maintains context, and continuously refines its profile of each user. It’s currently a proof of concept, but I plan to deploy it on a web server and share it publicly. Testing multi‑user functionality would be valuable to see how the system scales.
This project revives an older idea inspired by classic Alice bots, now powered by modern LLMs. Initially I considered using Google’s news‑topic vectors, but with today’s agentic AI the bot can call tools to read live news and websites, generating topics, thoughts, and questions proactively.
I noticed that many chatbots are static—only responding to user queries. My bot’s mission is to initiate conversation, learn about the user, and create deeper, more natural interactions through feedback loops, emotional scoring, external inputs, and a clever push‑messaging system.