💡 Inspiration
While reading long research papers and online articles, we often highlight key points, copy summaries into note apps, and search for definitions or translations all manually. We wanted a smarter way to summarize, translate, and learn directly inside Chrome, powered entirely by on-device AI for privacy and speed. This inspired MarkSense, which is a Chrome Extension that combines summarization, translation, and intelligent highlighting using Chrome’s built-in Gemini Nano APIs.
🦾 What it does
We built this Chrome-extension with five features.
| Feature | Description | Problems we Solve | API Used |
|---|---|---|---|
| Summarize the contents on the website | When users interact with MarkSense, they can generate a customized website summary with just a few clicks. After clicking the button, the tool automatically creates a summary of the webpage, which can then be tailored by selecting a preferred tone and desired length. These options allow users to shape the summary’s style and level of detail to best fit their needs. | Webpages often contain too much content for users to process quickly. MarkSense condenses long or complex pages into concise, easy-to-digest summaries—saving users time and cognitive effort. | Summarize API |
| Translate the summary that is generated by Summarize API | MarkSense supports multilingual functionality. When users select a specific language, the summary is automatically translated into that language, ensuring seamless accessibility and clear communication across diverse audiences. | Ensuring users can access concise, relevant summaries in their preferred language. | Translator API |
| Extract keywords from the website | MarkSense automatically extracts key keywords from the original content. This enables users to identify the most relevant topics and concepts, enhancing their understanding, content analysis, and overall productivity. | MarkSense solves the challenge of identifying key insights by automatically extracting the most relevant keywords from content, helping users quickly understand and analyze the main topics. | Prompt API |
| Generate related article | MarkSense generates related articles based on the original content, providing users with additional context and insights. Each related article is displayed as an interactive card, and when users click on a card, they are seamlessly directed to the corresponding website for further reading. | MarkSense solves the problem of limited context by automatically suggesting related articles, helping users explore deeper insights and discover relevant information effortlessly. | Prompt API |
| Highlight something important or confusing words or sentences | MarkSense allows users to highlight key sections of an article to emphasize important insights or flag confusing parts. This interactive highlighting feature helps users engage more deeply with the content, making it easier to revisit, analyze, or share specific information later. | MarkSense addresses the challenge of retaining and organizing key information by enabling users to highlight important or unclear sections, helping them focus on what matters most and engage more effectively with the content. | None |
🏗️ How we built it
We built MarkSense using Manifest V3, Side Panel API, and Chrome’s new AI SDK.
Our stack includes
- Summarizer API → Generates concise summaries of articles or PDFs.
- Prompt API → Defines contextual prompts to generate keyword definitions and related resources.
- Translator API (planned) → Enables multilingual summaries.
- Local Storage API → Persists highlights and notes per webpage.
- Third-party libs:
- Mozilla Readability for clean text extraction
All AI processing happens locally on-device, which means no API keys, no network calls, and zero data sharing.
⚙️ Challenges we ran into
- Chrome's built-in AI APIs (Summarizer, LanguageModel, Translation) are experimental and have inconsistent availability checks across different Chrome versions
- Summarizer API has input length limits (~4000 characters), but articles can be 50,000+ characters
- When users refresh the page, DOM elements change completely, making it impossible to restore highlights using simple element references or XPath
🎯 Accomplishments that we're proud of
MarkSense turns your browser into an intelligent study assistant, including summarizing, translating, and highlighting content instantly, all offline. No cloud processing, no logins, just smarter reading. This extension truely help students/researchers to improve their efficiency of collecting information online.
🧠 What we learned
- How to integrate Chrome’s on-device AI APIs (Gemini Nano) under strict Content Security Policy restrictions.
- Techniques for real-time text anchoring and persistent highlighting with chrome.storage.local.
- Designing a UI that feels natural within the Side Panel instead of intrusive overlays.
- Balancing privacy, performance, and usability — ensuring all logic runs offline.
🚀 What's next for MarkSense
Smart Highlight Analytics
- Track the frequency and sentiment of highlights to auto-identify learning patterns.
- Introduce a “Focus Mode” where summaries emphasize your most-highlighted concepts.
- Enable exporting highlights to Markdown or Notion for deeper study integration.
Context-Aware Summaries
- Dynamically adapt summary tone and length based on document type — academic, news, blog, or technical.
- Use contextual memory so the extension “learns” your reading habits and tailors outputs accordingly.
Log in or sign up for Devpost to join the conversation.