Veylan v0.2

Learn more at Hugging Face or return to the Arynwood Terminal.

Technical Insight — Building the Arynwood Bot

This project started with a simple goal: to understand what makes a chatbot feel alive. Using the free tier on Hugging Face, I worked with the Zephyr-7B-Beta model — lightweight enough to experiment with, but still powerful enough to hold a real conversation.

Because of hardware constraints, I didn’t train a new model from scratch. Instead, I focused on system message editing and adjusting runtime parameters like temperature, top-p sampling, and max new tokens.

Temperature controls how much randomness the bot uses. Higher values make it more creative and unpredictable; lower values make it more logical and focused. Top-p sampling affects how many possible next words the bot can choose from — balancing between creativity and precision. Max new tokens determines how much text the model is allowed to generate in a single response, helping it avoid memory crashes on small servers.

Inside the app.py file, I built a custom interface with sliders for these controls so every conversation could be shaped in real-time — without touching the core logic. By tuning the system message, I could alter the bot’s tone and internal assumptions without needing to retrain it.

Zephyr-7B-Beta isn’t GPT-4. But it gives me something just as important: control. It lets me explore, experiment, and build a system I fully understand.

This version of Veylan is a prototype. I plan to expand the work by hosting multiple local bots — each adapted for coding, writing, and research. Everything begins here: one deployment, one custom prompt, and one core question about intelligence.