feat: harden review workflow and reporting

feat: harden OCR and production runtime

fix: secure authentication denial responses

fix: secure sessions behind HTTPS proxies

fix: create upload directory at startup

fix: retain static directory in deployments

feat: complete production safety and quality gates

feat: harden prescription processing and evaluation

fix: refine_model_medications was treating fallback dict as list causing KeyError(0)

fix: revert to Qwen2.5-7B - smaller models not available on HF free inference API

fix: replace error alert with graceful in-page 'no medications found' message

fix: graceful fallback when HF fails + no crash on empty filtered results

perf: prefix-indexed fuzzy match + preload lexicon and OCR at startup

fix: add junk medication filter to remove clinic names, form text, mangled entries

feat: professional details page with metadata, table, collapsible panels and print support

fix: restore refine_model_medications, improve LLM temperature and token limit

fix: improve LLM prompt accuracy and natural language frequency parsing

feat: consolidate medications into table, add patient/doctor/date header

feat: return patient_name, doctor_name, date in /api/analyze response

feat: extract patient_name, doctor_name, date from OCR via LLM prompt

perf: switch HF model to Qwen2.5-1.5B-Instruct for faster inference

Fix plural rec_texts typo in dict parser

Add OCR debug logging

Fix OCR parsing logic for PaddleX v3 outputs

Fix PaddleOCR oneDNN Crash: Disable MKLDNN

Fix PaddleOCR Deployment: Add libgomp1 and lock numpy<2

Fix unterminated heredoc in Dockerfile

Fix PaddleOCR show_log missing argument at build time in Dockerfile

Support PaddleOCR v3 and legacy init args

Install OpenCV system libs for PaddleOCR in Space

Preload Paddle OCR models and stabilize runtime deps

Disable Paddle model source check for Space runtime

Use Python 3.10 base image for PaddleOCR compatibility on Space

Fix concurrent Paddle OCR initialization in Space

Merge master into main: add parquet benchmark support

Harden parquet benchmark parsing

Add parquet benchmark support and isolate app imports

Refine endpoint extraction with OCR heuristics

Add local OCR and model benchmarking workflow

Merge GitHub main into deployment branch

Add dataset-enriched medication summaries and UI refinements

Add Space metadata to README

Prepare app for local and Docker deploy

Split app logic into modules

Initial commit