fix: remove few-shot examples from SuggestMissions to prevent parroting

fix: tune SuggestMissions prompt for visual inference missions

fix: replace model names with analyst-style labels in reasoning trace

fix: chat tooltip clipping and inspect empty panel on track click

feat: auto-suggest ISR missions from video preview frame

refactor: clean up backend-frontend connection layer

fix: decouple PlanMission from AI postprocessing toggle

feat: improve ISR mission planner and assessor prompts

fix: race conditions in job cleanup and MJPEG frame drop visibility

revert: restore original single-video output with baked annotations

fix: serve annotated video for segmentation mode (masks are the output)

fix: resolve inspect endpoint errors — frame fallback search and explain error handling

feat: separate clean video from annotated video for interactive playback

fix: resolve backend-frontend integration issues across API contract, error handling, and resource management

chore: relax pinned versions in pyproject.toml, let uv.lock handle exact pins

refactor: migrate to uv, prompt-tune BAML mission planner and assessor

feat: add detailed BAML logging for mission planner and assessment verdicts

fix: add missing baml-py dependency to requirements.txt

refactor: merge verdicts into track responses and remove dead depth code

fix: add missing /detect/verdicts endpoint and clean up imports

test: add AssessDetections BAML tests with real SAR frame

feat: integrate BAML for mission planning and detection assessment

refactor: remove depth estimation pipeline entirely

fix(demo): auto-detect API base URL instead of hardcoding HF Space

feat(demo): add AI Postprocessing toggle for vision-only mode

refactor(pipeline): decouple vision stage from AI postprocessing

refactor(demo): extract shared canvas helpers in inspect panel

refactor: clean up backend utilities and background job handler

refactor(tracker): clean up ByteTracker implementation

refactor(detectors): extract shared YOLO logic into YoloBaseDetector base class

refactor: anonymize LLM labels, merge fragmented tracks, and tune defaults

fix(demo): improve 3D graph node label clarity and initial camera framing

feat(demo): threat-weighted timeline waveform

fix(demo): double-rAF in inspect panel to ensure layout before rendering

refactor: remove /app frontend, consolidate on demo as sole UI

feat(demo): port Tripo3D generative 3D support to demo inspection panel

fix(inspect): show errors instead of blank panel + make 3D on-demand

feat: integrate Tripo3D generative 3D models into inspection panel

fix(demo): don't auto-expand chat panel on track click

feat(demo): friendly track names — "Truck #3" instead of "3 truck"

fix(demo): skip SVG bbox overlay in segmentation mode

style(demo): elevate chat panel visual design

fix(demo): use Three.js 0.160 which has UMD build (0.175 dropped it)

feat(demo): add text labels to 3D graph nodes + spread out hierarchy

feat(demo): chip hover tooltip + single-object suggestion queries

feat(demo): collapsible chat panel toggle, matching event log pattern

fix(demo): extract glow rect helper, fix selectedTrackIds desync

fix(demo): address code review — double-push, double-execute, dedup toggle

feat(demo): multi-object context + expandable chat panel

fix(demo): restructure 3D graph as top-down hierarchy + remove duplicate Three.js