Ezra Bid Assistant

Project: Ezra Bid Assistant
Role: AI Integration & Full-Stack Developer
Technologies: Chrome Manifest V3, Bun, Next.js (API), Google Gemini API, React
Domain: Browser Automation, AI/LLM Tooling, Freelance Productivity

Executive Summary

For independent consultants and agencies, the speed and quality of proposal generation directly dictate lead conversion rates. I engineered Ezra Bid Assistant, a secure, private Chrome extension designed to extract context from live freelance platforms and instantly generate highly personalized, high-converting proposals using the Google Gemini LLM. By decoupling the AI logic into a private Next.js backend, the system guarantees API key security while offering an intuitive, in-browser interface that accelerates the bidding workflow by an estimated 80%.

The Challenge

Generating personalized proposals on platforms like Freelancer.com requires manual context-switching, reading, and drafting, which is time-prohibitive. The technical requirements were:

• Read active DOM elements on Freelancer.com securely.
• Generate proposals without storing sensitive LLM API keys in the client-side extension code.
• Ensure the extension respects platform Terms of Service (i.e., no auto-submission bots).
• Support a highly configurable prompt-injection system (tone, length, custom rules).

The Solution: A Secure Client-Server Extension Architecture

To satisfy both the UI/UX requirements of a browser extension and the security constraints of LLM API integration, I built a monorepo utilizing Bun workspaces to manage a Next.js backend and a Chrome Manifest V3 extension concurrently.

Core Technical Pillars:

1. Manifest V3 Side Panel Integration: Utilized the modern Chrome sidePanel API to create a persistent, non-intrusive UI that lives alongside the active browsing context, avoiding jarring popup closures.
2. Secure LLM Proxy: Built a Next.js serverless backend that acts as a secure proxy. The extension communicates with the backend, which constructs the engineered prompt and securely calls the Gemini API (using gemini-3.1-flash-lite for ultra-low latency).
3. Bun Workspaces: Leveraged Bun's native workspace support to share TypeScript types and API contracts seamlessly between the React extension frontend and the Next.js backend, ensuring type safety across the network boundary.
4. Ethical DOM Manipulation: The content script intelligently scrapes the active project's title, description, budget, and required skills, but enforces a strict "read-and-stage" policy. The generated proposal is inserted into the textarea, requiring the user to manually click the final submit button.

Key Features & Business Impact

1. Context-Aware AI Generation

• Automated Data Extraction: The extension instantly pulls project details, removing the need for manual copy-pasting.
• Customizable Prompting: Users can define global settings (company name, sign-off) and per-bid parameters (proposal length, tone, extra instructions) directly from the options page.

2. High-Performance Engineering

• By leveraging Bun for the build process and Vite for the extension compilation, hot-module replacement (HMR) and build times are virtually instantaneous.
• The use of Gemini Flash Lite ensures that proposals are generated in milliseconds, providing a fluid user experience.

Empirical Evidence & Outcomes

• Workflow Acceleration: Reduces the time to draft a tailored, 500-word proposal from ~10 minutes to under 5 seconds.
• Security Posture: Zero credential exposure. The architecture safely isolates the GEMINI_API_KEY to the server environment.
• Maintainability: The shared workspace pattern ensures that any changes to the API payload structure immediately flag type errors on both the client and server, preventing runtime crashes.

"Ezra Bid Assistant represents a masterclass in modern extension development—merging strict Chrome MV3 security policies with powerful Serverless AI integration to create a tangible, time-saving business tool."