PromptStream

Inspiration

The idea came from a simple frustration: while reading online, we’d often wonder, What does this mean in simpler terms? or How does this connect to a deeper concept? — Yet exploring those thoughts meant copying, switching tabs, and typing prompts into Gemini manually.

We wanted to bridge that gap — to create an experience where intelligence appears exactly where curiosity sparks.

That vision led to this extension: a multi-modal Gemini assistant that instantly turns any text into three levels of understanding — Simple, Deep, and Expert.

What it does

This Chrome extension transforms any selected text on a webpage into three intelligent AI prompts of varying depth — Simple, Deep, and Expert. Each prompt represents a different level of reasoning and context expansion. It offers multiple modes of prompts: professional, neutral, friendly, and persuasive.

When users highlight text, a small floating panel appears, offering three clickable buttons. Selecting a button allows users to open Gemini with the corresponding prompt pre-filled.

The extension is multi-modal, capable of generating neutral, professional, persuasive and friendly prompts enabling Gemini to produce rich prompt outputs directly in the browser. Its goal is to make knowledge exploration seamless — bringing AI understanding directly to the moment of curiosity.

How we built it

Frontend: Vanilla JavaScript + Chrome Extension APIs for lightweight injection and DOM manipulation.
Core Logic: A content script that detects text selection, builds three structured prompts of increasing analytical depth, and displays clickable floating buttons near the cursor. The extension uses Prompt API behind the scenes to generate multi-prompt modes.
Integration: Each button * opens Gemini in a new tab with the selected prompt pre-filled.
Storage: chrome.storage.sync for user preferences — base URL, prefix phrases, and prompt length limits.
Design: Minimalist grey floating UI that feels native, not intrusive.

Challenges we ran into

DOM Consistency: Handling dynamically changing sites, shadow roots, and frameworks that rebuild the DOM.
Permission Management: Balancing power and safety — avoiding overreach on sensitive pages like Gemini itself.
Prompt Quality Tuning: Engineering prompt templates that feel distinctly different in depth yet remain concise and relevant.
Performance: Ensuring smooth selection detection without adding event lag, even on text-heavy pages.

Accomplishments that we're proud of

Successfully built a multi-modal Chrome extension that transforms any text selection into three intelligent, depth-based Gemini prompts.
Engineered a seamless AI experience, allowing users to view responses directly in the browser.
Designed an elegant, non-intrusive UI that enhances curiosity-driven learning without disrupting the reading flow.
Created a foundation for adaptive prompt generation, where the extension scales from quick summaries to expert-level insights with a single click.
Balanced performance and functionality, ensuring smooth real-time behavior even on dynamic, content-heavy websites.

What we learned

How to design context-aware Chrome extensions that interact with page selections and manage event lifecycles efficiently.
Techniques for multi-depth AI prompt generation, tuning the language of each layer for curiosity, clarity, and complexity.
How to integrate Prompt API.
The importance of UX minimalism — balancing intelligence with non-intrusiveness.

What's next for PromptStream

Enhanced Multimodality: Extend support beyond text — enabling prompt generation from images, videos, and embedded media for richer AI interactions.
Collaborative Mode: Allow multiple users to share and compare AI interpretations of the same webpage in real time, fostering collective learning.
Voice Interaction: Integrate speech-to-text input and AI voice output, turning PromptStream into a hands-free, conversational companion.
Prompt Personalization: Introduce user-defined prompt profiles — e.g., Summarizer, Teacher, or Critic — to adapt depth and tone to individual goals.
AI Feedback Loop: Use reinforcement feedback from users to continuously improve the quality, depth, and accuracy of generated prompts.
Open AI Ecosystem Integration: Connect PromptStream with other AI APIs, making it a universal gateway for on-page reasoning and creativity.