Tastemate.AI

Stage 3 (Mental Well-Being)
Stage 5 (Academic Goals)
Stage 1 (Enter Your Name)
Stage 4 (Personal Goals)
Stage 6 (Dashboard)
Stage 2 (Physical Health)

Inspiration

I vividly remember my first few months using AI chatbots. I was amazed by their intelligence, but I quickly grew frustrated by their ignorance. No matter which app I downloaded, they were all strangers to me. Not a single chatbot asked me the questions that actually matter: "What are you studying?" "What are your spiritual beliefs and values?" "What is the specific dream you are chasing right now?" I realized that without this context—my education, my religion, my hobbies—the advice I got was just generic noise. It felt like asking a stranger for life advice. I decided that I didn't need another smart search engine; I needed an accountability partner. I wanted to build an AI that acts like a mentor—one that knows who I am, understands my background, and actively reminds me: "Hey, to achieve your goal of becoming a developer, remember to finish that module today." Tastemate.AI was born from the desire to turn AI from a reactive tool into a proactive, personalized companion.

What it does

It simply takes your information manually from you and allows you to customize your chatbot to work in accordance with your information while being specific.

How we built it

The application is a React single-page application (SPA) built with TypeScript. It utilizes a modular component architecture and integrates with Google Gemini via a dedicated service layer (services/gemini.ts) to handle AI context and responses.

Challenges we ran into

The biggest challenge in building Tastemate.AI was bridging the gap between a stateless React frontend and a context-heavy AI. I struggled to make the bot 'remember' user details across a conversation without hallucinating. By implementing a custom session-state manager in TypeScript and strict 'Context Guardrails' in the system prompts, I was able to transform a forgetful API into a consistent, long-term mentor.

Accomplishments that we're proud of

The Secret Sauce: Context InjectionThe real magic happens in how the prompt works. I treated the System Instruction not as code, but as a "Persona Injection."Usually, a chatbot works like this: $$Response = f(question)$$ But for Tastemate, I forced the math to look like this: $$Response = f(question \mid \text{Physical} \cup \text{Mental} \cup \text{Career})$$