Decision Agent

Inspiration

I built Decision Agent to solve everyday choice paralysis, such as choosing one dress or pair of shoes from several options.

What it does

Users upload one photo containing multiple items or several individual photos. The agent asks about occasion, weather, comfort, and style, then selects exactly one option and explains why also we can add more about the. getting user preferences and save and then every time user upload and ask for suggestion or decision so agent according user preference and lifestlye give them best option to select.

How I built it

I built it independently using Gemini Vision, Python, FastAPI, Streamlit, Pydantic, and Docker. Gemini compares the images, while structured schemas validate the final decision.

Challenges I ran into

The main challenges were detecting multiple items in one photo, maintaining correct option numbers, and making the AI select one winner instead of giving vague suggestions.

Accomplishments I’m proud of

I created a working multimodal decision agent with explainable results, confidence scores, trade-offs, safety boundaries, and frontend-ready APIs.

What I learned

I learned how to combine visual understanding, structured AI output, user context, validation, and explainable recommendations.

What’s next

I plan to add preference memory, adaptive questions, weather integration, feedback learning, improved item detection, and Google Cloud deployment.

Built With

• agents
• docker
• fastapi
• gemini
• genai
• google
• multimodal
• pydantic
• python
• sdk
• streamlit
• vision