feat: add chatbot & 2D accident animation tabs

feat: add new metrics for accident severity and geographic location

feat: upgrade to LiquidAI/LFM2-VL-3B for balanced inference

fix: switch to Qwen2-VL-7B-Instruct (open model)

feat: upgrade model to Llama-3.2-11B-Vision-Instruct

fix: update parser for new Vehicle 1 format

feat: add time, road type, and detailed vehicle positioning to LFM prompt

fix: add strict format constraints to LFM prompt

feat: update LFM prompt and parser to reproduce static UI format

fix: move @spaces.GPU to top-level of app.py for ZeroGPU detection

refactor: switch to Gradio ZeroGPU with custom frontend + REST API

Add automatic lifespan DB fix for persistent volume staleness

Implement direct static mounted routing for images via FastAPI

Fix frontend photo URL concatenation

Fix photo rendering with robust relative path resolution

Update image paths and add /api/logs

Initial MVP deployment with baked-in data