refactor: centralize prompts, remove model selector, clean README

feat: filter stale jobs and add freshness indicator

chore: remove DuckDuckGo references from app description

fix: remove deprecated Groq models (gemma2-9b-it, mixtral-8x7b-32768)

feat: remove DuckDuckGo scraper and 'any' position type; add docs link to README

fix: use public agent API in app.py; update README for new structure and Groq

refactor: full OOP redesign of agent architecture

fix: always show Groq models in dropdown regardless of key presence

feat: add Groq backend as primary free inference provider

feat: make all scored positions reviewable, not just those above min_score

refactor: remove dead code, extract shared job helpers

chore: rename app to PhdScout

feat: add predoctoral position type

docs: update README and app.py description to reflect current state

feat: add mlscientist.com scraper; remove map from UI

feat: location dropdown with interactive Leaflet map (Nominatim reverse geocoding)

fix: update model list to HF Router-compatible chat models

fix: disable API info to workaround gradio_client bool schema bug

fix: launch with server_name=0.0.0.0 for HF Spaces compatibility

feat: remove per-user token input, use shared HF_TOKEN secret from env

refactor: OOP redesign — JobAgent orchestrator, shared utils, no global state mutation

fix: upgrade gradio to v5 for Python 3.13 compatibility