- Android native startup script for Termux deployments:
scripts/android-up.sh(no Docker, no root). - New thermal monitoring package:
backend/internal/thermal/monitor.go. - Thermal band reporting in
GET /api/v1/system/statsviathermal_band. - Discord system alerts on thermal Hot-band entry for cameras with webhook configuration.
- Runtime startup now conditionally enables thermal monitoring on
arm64 + pimode (orTHERMAL_MONITOR_ENABLED=true). - Added thermal transition logging with WARN/ERROR/CRITICAL severity labels.
- README and quickstart updated for Android no-Docker setup and current release guidance.
- Stream readiness gate for
/api/v1/cameras/:id/streamwith explicitinitializingstatus and delayedhls_urlexposure until HLS is reachable. - Idempotent mediamtx path synchronization logic to avoid repeated recreate loops during keepalive.
- Expanded production walkthrough in README covering every supported setup method and operator checklist.
- mediamtx stream profile now defaults to proactive startup (
sourceOnDemand=false) while continuing to enforcertspTransport=tcp. - Keepalive behavior now uses grace and backoff windows before path repair, and no longer triggers immediate readiness-based repair churn.
- Frontend stream status handling now supports
initializingacross camera card, single-camera view, multi-camera view, and shared status badge types.
- First-class runtime deployment modes (
pi,standard,viewer) with startup auto-detection and mode-gated subsystem startup. - Snapshot Intelligence Engine for Pi mode (
backend/internal/snapshotai) with Claude/OpenAI vision providers and structured event persistence. - Shared frame capture helper (
CaptureFrameToPath) for external snapshot pipelines. - New architecture and decision documentation for deployment mode guarantees and Pi AI constraints.
- Backend boot flow now initializes YOLO detection only when mode allows it, and runs degraded detection handles in non-YOLO modes for API compatibility.
docker-compose.ymlnow sets explicitRTSPANDA_MODEdefaults and includes Snapshot AI environment controls for Pi profile runs.scripts/pi-up.shnow hard-blocks unsupported local AI-worker paths on ARM, clarifies supported Pi paths, and improves post-deploy guidance.- README was fully rewritten around the three deployment modes and explicit Pi constraints.
- Pi-first deployment architecture with a lightweight
rtspanda-piservice and standaloneai-worker-standaloneprofile. - Remote AI mode via
AI_MODEandAI_WORKER_URL, allowing Raspberry Pi ingest nodes to forward frames to a second machine. - New docs for raspberry-pi.md and cluster-mode.md.
- Backend tests for AI mode resolution and remote/local detector targeting.
- AI worker tests covering
MODEL_SOURCE, missing local model behavior, and health reporting.
ai_worker/Dockerfileis now deterministic and ONNX-only: noultralyticsinstall, no export fallback, no runtime model conversion.- AI worker runtime now prefers local models, reports
model_source, and fails or degrades explicitly when a model is missing. docker-compose.ymlnow supports additivefull,pi, andai-workerprofiles while keepingdocker compose up --build -dunchanged.- Pi helper scripts were refactored for three clear modes: lightweight Pi, full local stack, and standalone AI worker.
- README was rewritten with employer-friendly architecture framing and full setup guides for Standard, Pi Standalone, and Pi + AI deployments.
- Frontend lockfile was refreshed so the current frontend test toolchain installs cleanly.
- Backend frontend-embed behavior now compiles from a clean checkout without requiring prebuilt web assets.
- Token-based authentication flow across backend and frontend (login/session + protected API routes).
- Initial automated test baseline for backend, frontend, and AI worker.
- Raspberry Pi deployment helpers (scripts/pi-preflight.sh, scripts/pi-up.sh, scripts/pi-down.sh).
- New operational docs for streaming tuning, Pi deployment, and AI compatibility.
- Multi-arch container build flow to remove hard amd64 assumptions.
- mediamtx runtime defaults aligned to reduce configuration drift and improve latency/stability balance.
- AI worker runtime controls expanded for Pi-safe CPU-only inference and fallback behavior.
- README regenerated with consolidated setup, usage, and architecture guidance.
- This release contains coordinated changes across auth, streaming, AI, tests, and deployment. Run a full post-release smoke test on target hardware.