#README.md

AI Travel Concierge & ARC-AGI Task Benchmark Suite

Overview

An advanced AI-powered platform that plans, optimizes, and adapts travel itineraries in real-time while integrating an automated ARC-AGI-1 Task Generation Runner for frontier reasoning model evaluation. It provides users with a robust web interface to generate custom travel schedules and trigger complex AI evaluation tasks dynamically.

---

Tech Stack

• Cloud PaaS: Render PaaS Server (Automated deployment via render.yaml and build.sh)
• Process Manager: Gunicorn WSGI Application Server
• Backend Framework: Django 5.x Python Web Framework
• Database: SQLite (ORM-driven structured job & itinerary storage)
• AI & Integration: Ollama & OpenAI-compatible endpoints with robust fallback generation routines
• Frontend Experience: Responsive HTML5/JavaScript dashboard featuring WebGL sky background textures, interactive mouse-tracking (mousemove), falling geometric/smog particle simulation, digital terminal styling, and ASCII shell animation
• Version Control: GitHub

---

Setup

1. Repository Configuration: Ensure your project root contains manage.py, requirements.txt, build.sh, and render.yaml.

2. Environment Variables: Configure necessary environment variables including OPENAI_BASE_URL, OLLAMA_MODEL, and execution parameters.

3. Render Deployment: Link your GitHub repository to a new Web Service on Render with the start command:

   gunicorn mysite.wsgi:application

4. Build & Execution: Push your code changes to GitHub to initiate the automatic build and WhiteNoise static asset collection workflow.

   Features

• AI Travel Concierge: Dynamic destination, date range, and budget selection generating personalized day-by-day itineraries with built-in cost optimization.
• ARC-AGI-1 Task Generator: Creates fresh, distribution-matched tasks via backend python scripts (generate_tasks.py and generate_tasks_stratified.py) designed to evaluate reasoning models.
• Interactive Dashboard UI: Immersive frontend featuring WebGL shaders, particle simulations, and terminal-style feedback.

---

Technical Workflow

1. Request Reception: Django views handle incoming HTTP POST requests from the HTML dashboard for task generation or itinerary planning (trigger_task_generation, ai_concierge_plan).
2. Subprocess & API Execution: The backend dynamically invokes internal generation scripts or communicates with OpenAI-compatible/Ollama endpoints via requests/httpx.
3. Database Persistence: Processed outputs, job states, standard logs, and travel itineraries are securely structured and saved using Django's SQLite ORM models (GenerationJob, TravelItinerary).
4. JSON Response Delivery: Results and execution statuses are returned asynchronously to update the front-end dashboard in real time.