One of the most significant challenges we encountered during the development of our mental health assistant website, Cereflow, was navigating and integrating various platforms and technologies as beginners. From the very beginning, building a visually clean, responsive, and accessible website was a learning curve, particularly when it came to customizing UI components and managing functional elements like navigation, interactive buttons, and chatbot containers. Linking our frontend codebase with GitHub and deploying through Vercel introduced additional complexity. Understanding how real-time syncing between these platforms worked—such as when a change made in the local development environment would reflect correctly on GitHub and then on the live site—required a lot of hands-on experimentation. We faced version control issues, such as cloning into existing directories, handling pull/push conflicts, and properly using git branches, especially while working across both the cloud and local environments with VS Code. Another major hurdle was the chatbot integration. We trained a chatbot using Botpress to act exclusively as a mental health assistant. However, to our surprise, the bot began answering unrelated queries such as solving math problems or handling programming questions, which was outside its intended scope. This challenged us to dig deep into prompt engineering and persona restrictions, refining the instructions and system prompts so that the assistant politely declined unrelated topics and redirected users to meaningful mental health discussions. Setting up the Botpress Webchat itself wasn't as straightforward as expected. Even after successfully training the bot, deploying it to the web interface and generating a usable script for embedding into our website took time. We had to manually identify the right options to generate web integration code and then carefully place it into our site’s HTML, ensuring it would load properly and remain responsive.