🛰 VOYAGR 🌗

Agentic AI: Planner → CNN Vision → RAG Retriever → LLM Drafter → Budget Critic (self-revises) → DFS Optimizer → Writer.
RAG: TF-IDF cosine retrieval over a travel knowledge base grounds the LLM prompt.
CNN: Sobel convolution → ReLU → max-pooling feature maps + colour statistics classify the photo's scene type.
DFS: Branch-and-bound depth-first search finds the shortest visiting order of each day's stops using Haversine distances.
Journey: geocoding + OSRM road routing + a multi-modal fare model estimate door-to-door travel cost, which is reserved from your budget first.
Stack: HTML/JS (Next.js-portable), Groq (auto-detects live models) / Gemini, Leaflet + OpenStreetMap, Wikipedia, Open-Meteo, Docker (nginx).