feat(inspection): add frame extraction and mask retrieval API endpoints

feat: add mission analyst chat panel with track context injection

feat(isr): wire mission param through API/background pipeline, add openai dep

refactor: remove stream downscaling, pass frames at full resolution

fix: steady 24fps MJPEG streaming with no frame drops

refactor: remove first-frame tab and simplify to video-only UI

feat: add /chat/mission endpoint and rewire chat UI

refactor: strip threat fields from all frontend and backend code

feat: add timeline heatmap with detection highlights for Tab 2 video player

feat: replace drone_yolo with YOLOv8-VisDrone detector

feat: expose step param on /detect/async, default to 7

feat(ttfs): instrument segmentation time-to-first-stream

refactor: rename hf_yolov8 → yolo11 across codebase

feat(segmentation): add YSAM2 variants and mission-aware detector mapping

refactor(gsam2): make SAM2 detector-agnostic

perf: increase production keyframe step from 20 to 60

perf: set SAM2 num_maskmem to 7 (trained default) in production

feat: expose num_maskmem parameter in benchmark profile/analysis endpoints

feat: add benchmark profiler & roofline analysis system

fix: replace gated GSAM2 streaming with adaptive-rate publisher thread

fix: track cards update during playback + periodic GPT re-analysis

fix: correct benchmark metrics — remove double-counting and add missing timers

perf: set num_maskmem=3 for all inference paths

feat: add num_maskmem parameter to /benchmark endpoint

feat: add /gpu-monitor endpoint for live nvidia-smi streaming

feat: add detailed timing metrics to GroundedSAM2 segmenter

feat: add job processing timing and latency measurements to backend and frontend

feat(frontend): improve UI for detection and chat interaction

Handle GPT refusal paths and preserve assessment status

Fix segmentation mission routing and first-frame GPT alignment

Remove SAM3 and standardize segmentation on Grounded-SAM2

Refine GPT enrichment logic, add relevance badges to UI, and fix minor model/logging issues

Fix GPT thread safety, improve relevance logic, and add caching for COCO matching

Async GPT enrichment: fire-and-forget background task updates track cards from UNASSESSED to ASSESSED

Event-driven MJPEG streaming, GPT concurrency semaphore, pre-resize frames in publisher

Jobs: forward cached GPT results, reduce stream buffer to 60, wire up first_frame_gpt_results

Add vision-grounded mission parsing: send first frame to GPT for object class refinement

feat: Add LLM post-processing relevance gate for broad detection

feat: Add mission-driven object relevance abstractions

feat: Add threat chat feature for Q&A about detected threats

feat(threat-assessment): implement naval threat analysis with GPT-4o\n\n- Rename utils/gpt_distance.py to utils/gpt_reasoning.py and update logic for 15 naval threat features\n- Add Pydantic schemas for NavalThreatAssessment in utils/schemas.py\n- Update backend (app.py, inference.py) to use new threat estimation and pass full metadata\n- refactor(frontend): render threat level badges and detailed feature table in UI

feat: Implement depth-based distance estimation and frontend integration

feat: Real-time radar directionality via track sync

fix: Update static file path for deployment

feat: Continuous object tracking, speed estimation, and overlay syncing

Fix inference concurrency bugs and enable default first-frame GPT

Add internvl dependency and disable frontend caching

Optimize streaming buffer and implement parallel depth pipeline

Refactor: Decouple depth estimator and suppress noisy logs

fix: Handle empty GPT responses and log outputs