blueprint.dev - Collaborative AI Agent Planning Workspace

A modern, real-time collaboration tool for designing software architecture with AI agents.

AI Workflow Architecture for Collaborative System Design

This workflow integrates distinct AI agents into your collaborative system design tool. The system accepts a System Design Plan (in .md format) and passes it through a pipeline of specialized agents powered by Gemini.

Each agent operates on a specific "Lens" to prevent context pollution (e.g., the Security agent doesn't care about variable naming, only vulnerabilities).

Workflow Overview

• Input: User uploads/edits a System Design Plan (design_plan.md).
• Parallel Processing: The plan is simultaneously analyzed by specialized Lenses.
• Synthesis/Action: The Plan Refiner aggregates insights and offers "Accept/Reject" modifications.

Tool 1: Security Lens Toggle 🛡️

Trigger: User toggles "Security Review Mode" in the UI. Role: Specialized Auditor.

Workflow Logic:

• When the toggle is ON, the system sends the markdown content to a strict security-focused system instruction.
• It ignores architectural elegance and focuses solely on the "CIA Triad" (Confidentiality, Integrity, Availability).

Prompt Strategy (Gemini): "You are a Security Auditor. Analyze the provided system design markdown. You are strictly forbidden from commenting on feature logic. ONLY identify: Authentication flaws, Injection risks, Secrets handling, Rate limiting gaps."

Tool 2: AI Contradiction Detector ❌

Trigger: Automatic (runs on save or edit). Role: Semantic Consistency Checker.

Workflow Logic:

• This agent parses the full document to build a "Truth Table" of architectural decisions.
• It compares early definitions against later implementation details.

Prompt Strategy (Gemini): "Analyze this system design for semantic contradictions. Example: Statelessness claimed vs. Server Sessions used."

Tool 3: AI Risk Labeler (Step-Level) 🚨

Trigger: UI Rendering (Visual Annotation). Role: Metadata Tagger.

Workflow Logic:

• This agent breaks the markdown plan into logical "steps" or "blocks".
• It assigns a risk score to each block to visually color-code the UI (e.g., Red border for High Risk).

Prompt Strategy (Gemini): "For each distinct step or module in the provided markdown, assign a risk level (low, medium, high)."

Tool 4: Plan Refiner (Auto-Remediation) 🛠️

Trigger: User clicks "Fix Issues" or "Apply Recommendations". Role: Active Editor.

Workflow Logic:

• This tool takes the outputs from Tools 1, 2, and 3.
• It generates a "Diff" or a revised version of the .md file.
• It presents the changes as a "Pull Request" style view for the user to accept or reject.

Features

Real-time Collaboration

• Multiplayer Editing: See other users' cursors and edits in real-time.
• Live Presence: Know who is currently viewing the document.

AI-Powered Planning

• Instant Plans: Generate comprehensive implementation plans from a simple prompt.
• Enhance with Gemini: Refine your plan with a single click, incorporating your comments and annotations.

Rich Annotation Platform

• Contextual Comments: Highlight text to add comments or attach images.
• Whiteboard: Draw diagrams directly in the browser and attach them to your plan.
• Redline Mode: Mark up the document for deletion or review.

Modern UI/UX

• BentoGrid Layout: A beautiful landing page showcasing features.
• Dark Mode: Native support for light and dark themes.
• Shadcn UI: Built with accessible and customizable components.

Tech Stack

Frontend (Client)

• React 19
• Vite
• TailwindCSS v4
• Shadcn UI
• Lucide React
• Socket.io-client
• Axios

Backend (Server)

• Node.js & Express
• Google Gemini API (@google/genai)
• Socket.io
• Mongoose (MongoDB)

Getting Started

Prerequisites

• Node.js 18+
• npm

Installation

1. Clone the repository

   git clone https://github.com/Om-Thanage/blueprint.dev.git
   cd blueprint.dev

2. Install dependencies

   Install Backend Dependencies:

   cd backend
   npm install

   Install Frontend Dependencies:

   cd ../frontend
   npm install

3. Set up environment variables

   Backend (.env) Create a .env file in the backend directory:

   PORT=5000
   GEMINI_API_KEY=your_gemini_api_key
   MONGODB_URI=mongodb://localhost:27017/blueprint

4. Run the application

   Terminal 1 (Backend):

   cd backend
   npm run dev

   Terminal 2 (Frontend):

   cd frontend
   npm run dev

5. Open your browser Navigate to http://localhost:5173 to start building.