Replace PyTorch scaffolds with sklearn baselines for ECGFounder + pulmonary_classifier

Wire fetal_health (CTG) + skin/retinal image-driven adapters end-to-end

Ship synthetic-baseline pickles for 7 sklearn ML adapters + fix label/prob alignment

Phase 2.2/2.3: snapshot-driven retrieval + broader planner triggers

Phase 1.5: bundle ml/ adapters + 2 trained pickles

Phase 1.2/1.3/1.4: per-specialty Groq keys + chat history + patient profile

Port complex-diagnosis graph to HF Space AI service

Walk messages backwards to find the actual user question

Wire user question into expert graph LLM call

Initial MedVerse AI deploy