Clarity

• 

  Clarity

Inspiration

Most decision tools focus on speed and answers. I wanted to build something that focuses on clarity before action.

I noticed that many bad decisions are not caused by lack of information, but by: • unclear priorities • emotional overload • repeating the same patterns without reflection

Clarity was inspired by the idea that better questions lead to better decisions, and that reflection over time especially learning from past choices is a missing layer in most AI tools.

What it does

Clarity is an AI-powered decision companion that helps users make better decisions through structured thinking, memory, and perspective shift.

The app: • Guides users through clarifying questions before making a decision • Uses Gemini embeddings to find similar past decisions from the user’s own history • Analyzes priorities and patterns across decisions • Generates a “Future You” reflection, simulating how the decision might look 6 months later • Builds long-term self-awareness instead of giving one-off advice

The goal is not to decide for the user but to help them decide more consciously.

How we built it

Clarity is built as a lightweight web app using Python + Streamlit, with Gemini as the core intelligence layer.

Key Gemini integrations: • Gemini 3 text generation for: • clarifying questions • decision analysis • future-self reflections • Gemini embeddings for: • semantic similarity search between decisions • personal decision memory

Decisions are stored locally in a structured format, enriched with embeddings and metadata (priorities, reflections). This allows Clarity to act as a personal decision history, not just a chat interface.

Challenges we ran into

• Designing AI outputs that support thinking instead of replacing it
• Avoiding generic advice and motivational fluff
• Managing API rate limits while keeping the app responsive
• Making insights explainable and grounded, not “black-box AI”
• Structuring decision memory so it stays useful over time

A key challenge was balancing ML-powered features with human-readable explanations.

Accomplishments that we're proud of

• Successfully integrating Gemini embeddings for personal decision similarity
• Building a working “Future You” reflection feature that users find emotionally impactful
• Creating explainable insights instead of opaque predictions
• Designing a system that improves over time as more decisions are added
• Keeping the entire app simple, inspectable, and user-controlled

What we learned

• AI is most powerful when it augments reflection, not automation
• Memory + context matter more than raw intelligence
• Users trust AI more when it explains why, not just what
• Gemini works especially well for structured reasoning and perspective-taking tasks

This project reinforced the idea that decision quality compounds over time and AI can help make that compounding visible.

What's next for Clarity

Next steps include: • ML-based topic clustering using Gemini embeddings • Timeline views to track decision evolution • Optional cloud sync for long-term decision memory • Deeper pattern detection across values and outcomes • A lightweight mobile-friendly interface

Long-term, Clarity aims to become a personal decision intelligence layer, helping people learn how they think not just what they choose.

Built With

• 3
• dotenv
• embeddings
• gemini
• github
• google
• jsonl
• numpy
• python
• streamlit