NehaS98/mental-health-discourse-analyzer

feat: serve discovered topics via /topics API for deploy compatibility

fix: remove side-effect mkdir from config import (broke non-root container)

chore: enforce LF line endings for shell scripts and Dockerfile

feat: HF Hub model loading + Dockerfile for Spaces deployment

refactor: pivot to 3-class severity matching real dataset; add --max-rows flag

docs: add README with architecture, setup, and design notes

feat(dashboard): add Streamlit dashboard with classification and theme views

feat(api): add FastAPI service with /predict and /predict_batch endpoints

feat(inference): add shared predict() module with cached model loading

feat(topic): add BERTopic theme extraction with c-TF-IDF

feat(train): fine-tune RoBERTa with weighted loss and MLflow tracking

feat(data): add Reddit MH data prep with weak-supervision severity labels

chore: scaffold project structure, dependencies, and central config