Lexa

Lexa is the context layer for everyday AI tools

Comment

Inspiration

A lot of time and credits are wasted due to inefficient prompting and loss of context when using GenAI tools. Each session starts from scratch, context gets fragmented across chats, and users often don’t know how to phrase prompts for the best results. We wanted to solve this by building an assistant that understands your intent, optimizes your prompts, and remembers your context across tools — making GenAI faster, cheaper, and smarter to use.

What it does

Lexa acts as your prompt enhancer and context manager. It improves your interaction with GenAI tools in two key ways:

  1. Prompt Enhancement: It asks clarifying questions and injects context from your previous projects or workspace to generate optimized prompts that get better results.
  2. Format Optimization: For image/video generation, it automatically structures prompts into a JSON format including lighting, aesthetics, and technical attributes for high-quality outputs.

The result — smarter prompts, consistent context, and reduced token/credit usage without sacrificing quality.

How we built it

  • Frontend: Chrome Extension built with React + Tailwind
  • Backend: Powered by Gemini Nano and a lightweight retrieval layer to store and recall contextual data from user projects.
  • Prompt Engine: Uses a two-step reasoning process — (1) clarify and enrich prompt, (2) output in optimal format (text or JSON).

Challenges we ran into

  • Designing a clean UX that feels invisible yet powerful while typing. Still a work in progress and is the focus area for the future
  • Integrating multiple built-in-AI APIs and doing QA testing to ensure a bug free experience
  • Managing context retrieval efficiently without high latency.
  • Balancing between automation and creative control — ensuring the enhancer helps but doesn’t override user intent.
  • JSON schema standardization across different GenAI tools (image vs video).

Accomplishments that we're proud of

  • Built a working Chrome Extension prototype with real-time prompt enhancement.
  • Created a consistent schema for image/video prompt optimization.
  • Designed a universal context layer that can inject project-specific information dynamically.
  • Made GenAI prompting feel effortless and human-aware.

What's next for Lexa

  • UI: Improving the UI of the chrome extension to make the experience better
  • Custom Model: : Training a custom model for better prompt enhancement
  • Team Mode: Shared context across collaborators to make GenAI projects multi-user.
  • Custom Prompt Libraries: Learn from your past projects to auto-suggest style or tone.
  • Deeper Integrations: Extend beyond Chrome to IDEs, Notion, and Figma.

Built With

Share this project:

Updates