hanszhu/Dense-Captioning-Platform

feat(api): expose MedSAM chained call as named endpoint /medsam for remote invocation

feat(medsam): prompt-only segmentation (bboxes/points JSON); skip if none; polygons by default; optional raw masks

feat(medsam): set min_area=100 with max_regions=20 for more fine segments

feat(ui): re-enable MedSAM overlay rendering

feat(medsam): UI shows polygons only; optional raw masks via hidden flag to avoid UI crashes

chore(ui): temporarily disable MedSAM overlay to prevent page crashes

chore(logging): add per-request timestamped logs and exception tracebacks for analyze paths

chore(docker): set PYTHONUNBUFFERED=1 for immediate log flushing

chore(runtime): remove self-ping keepalive thread per request

chore(ui): add elem_id to Gradio components; align MedSAM path with original (no auto SAM); perf tweaks

feat(medsam): follow original behavior; remove SAM auto from inference path; MedSAM boxes only (max 8)

perf(medsam): reduce fallback candidate boxes to 8

chore(space): trigger rebuild

fix(sam): use default SamAutomaticMaskGenerator; still cap to top-12 masks to avoid crashes

perf(medsam): cap SAM masks to top-12; reduce fallback regions to 12 for faster requests

feat(runtime): add lightweight self-ping keepalive thread to prevent idle shutdowns

fix(runtime): disable Gradio SSR and increase keep-alive for Docker Space stability

fix(custom_models): add __init__.py and custom_hooks stub; set PYTHONPATH in Dockerfile for mmengine imports

fix(deps): pin numpy==1.26.4 in requirements; remove duplicate numpy install from Dockerfile

fix(space): use writable HF cache dir; set envs; run as user 1000; pin numpy<2; avoid permission + ABI issues

fix(deps): pin numpy<2 and scikit-image==0.21.0; avoid NumPy 2 ABI issues; keep installs only in Dockerfile

chore(docker): dedupe deps; restore Gradio app; move installs to Dockerfile; no runtime installs

chore(space): switch to Docker SDK; add Dockerfile; minimal FastAPI app; trim requirements

deploy(space): push working Gradio app with API /predict, MedSAM auto-overlay, HF model downloads, cleaned requirements

build(space): initial Docker Space with Gradio app, MMDet, SAM integration