Skip to content

Ker102/Crosswind-Console

Repository files navigation

Crosswind Console

Status License Frontend Backend AI

Intelligent Orchestration for the Modern Explorer. A unified research dashboard that blends immersive 3D aesthetics with powerful AI agents to surface opportunities in Travel, Jobs, and Social Trends.


🚀 Overview

Crosswind Console is not just a dashboard; it's an AI-driven orchestrator. It solves the problem of information fragmentation by deploying specialized agents to scour the web, analyze data, and present actionable insights in a single, cohesive interface.

Whether you're looking for the cheapest flight to Tokyo, the next big career move, or a viral trend before it peaks, Crosswind Console coordinates the complex work of data gathering and reasoning behind the scenes, delivering results through a stunning, glassmorphic UI.


✨ Key Capabilities

🧠 Intelligent Orchestration

At its core, the system uses advanced LLM Reasoning to understand user intent. It doesn't just search; it plans.

  • Context-Aware: Understands "Find me a trip to Paris" vs. "Find me a tech job in Paris".
  • Tool Chaining: Automatically selects the right tools (Search, Scrapers, APIs) to fulfill a request.

🌍 Multi-Modal Travel Agent

A next-generation travel assistant that goes beyond simple flight search.

  • Hybrid Ground Transport Engine: Combines Google Search discovery with Firecrawl deep scraping to find bus and train routes (FlixBus, Rome2Rio) that standard APIs miss.
  • Fallback Reliability: Automatically switches to the Kiwi API for standard routes if scraping fails, ensuring you never get a "no results" error.
  • Flight Discovery: Real-time flight pricing and route analysis.

💼 Career & Trend Agents

  • Jobs Agent: Aggregates opportunities from major platforms, filtering by relevance and potential.
  • Trends Agent: Monitors social signals to identify rising topics and viral content.

🎨 Immersive Experience

  • Dynamic Island Navigation: A sleek, iOS-inspired navigation bar that expands and adapts to your context.
  • 3D Visuals: Interactive 3D elements (Icon Cloud) powered by Three.js for a premium feel.
  • Glassmorphism: A modern, translucent design language that feels alive.

🛠️ How It Works

The architecture is built for speed, modularity, and intelligence.

  1. The Interface (Frontend):

    • Built with Svelte 5 and Vite for blazing fast performance.
    • Handles user interactions and renders the immersive 3D environment.
    • Communicates with the backend via typed API contracts.
  2. The Brain (Backend):

    • Powered by FastAPI (Python).
    • Acts as the central coordinator, receiving user prompts and dispatching them to the appropriate agents.
  3. The Tools (MCP Servers):

    • We use the Model Context Protocol (MCP) to standardize how the AI interacts with external tools.
    • Travel Server: Handles flight/ground transport logic (Search + Scrape).
    • Trends/Jobs Servers: Specialized connectors for their respective domains.
    • Docker Gateway: Manages these tools securely and efficiently.

💻 Tech Stack

  • Frontend: Svelte, TypeScript, TailwindCSS, Three.js
  • Backend: Python, FastAPI, Uvicorn
  • Data & AI: Model Context Protocol (MCP), Firecrawl, Playwright, Google Custom Search, RapidAPI
  • Infrastructure: Docker

⚡ Getting Started

Prerequisites

  • Node.js & npm
  • Python 3.10+
  • Docker (optional, for MCP Gateway)

1. Backend Setup

cd backend
python -m venv .venv
# Activate virtual env (Windows: .venv\Scripts\activate, Mac/Linux: source .venv/bin/activate)
pip install -e .
uvicorn app.main:app --reload

2. Frontend Setup

cd frontend
npm install
npm run dev

Visit http://localhost:5173 to launch the console.

3. MCP Servers (Optional)

To enable full AI capabilities, ensure your .env file is configured with necessary API keys (RAPIDAPI_KEY, GOOGLE_SEARCH_CX, etc.) and run the specific server scripts in mcp/.


Built with ❤️ by the Crosswind Team.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published