RTSPanda

Watch your RTSP cameras in the browser. One app. No cloud. Runs on your machine.

RTSPanda is a small app you run on your PC or server. You add your camera URLs, open a browser, and watch live—no account, no subscription, no data sent to the cloud. You can also record, take screenshots, run per-camera YOLOv8 tracking, view live overlays + event history, and send rich Discord alerts.

---

Demo

---

What is new (v0.0.4)

• YOLO UI Polishing: per-camera ignore-zone polygon editor, overlay support, and improved YOLO camera setting flows.
• Security Updates: stronger backend input validation for ignore zones and integration settings, plus centralized OpenAI key/model handling in app settings.
• UI Updates: multi-camera mode (up to 4 streams), Integrations tab, dashboard network reset action, and bundled Chrome PiP extension install flow.
• Reliability Updates: ONNX Runtime migration for lower AI-worker memory, stream keepalive health checks, and stream reset APIs.
• Connectivity: external recording sync to Local Server, Dropbox, Google Drive, OneDrive, and Proton Drive.

Release details: RELEASE_NOTES_v0.0.4.md Diff: v0.0.3...v0.0.4

---

What you need

• A computer (Windows, macOS, or Linux)
• Your camera’s RTSP URL (looks like rtsp://admin:password@192.168.1.64:554/stream)
• To build once: Git, Go, and Node.js (we’ll show you how to install them on Windows below)

Optional: the mediamtx program is what actually pulls the video from your camera. If you don’t add it, RTSPanda still runs and you can add cameras—they’ll just show “offline” until you drop in mediamtx.

---

Quick Start (any OS)

1. Get the code

git clone https://github.com/248Tech/RTSPanda.git
cd RTSPanda

2. Get mediamtx (so you can see video)

• Go to mediamtx releases and download the zip for your system.
• Put the mediamtx (or mediamtx.exe on Windows) file inside the mediamtx folder in RTSPanda.

3. Build

• Windows: open PowerShell in the RTSPanda folder and run:
  .\build.ps1
• Mac/Linux: run:
  make build

4. Run

• Windows:
  .\backend\rtspanda.exe
• Mac/Linux:
  ./backend/rtspanda

5. Open your browser

Go to http://localhost:8080. Click Settings → Cameras → Add Camera, enter a name and your RTSP URL, then go back to the dashboard and click the camera to watch.

---

One-line Docker setup

Use this if you want the fastest deploy path without local Go/Node setup.

git clone https://github.com/248Tech/RTSPanda.git && cd RTSPanda && docker compose up --build -d

Then open http://localhost:8080.

This compose setup starts both services: rtspanda (Go backend) and ai-worker (FastAPI + YOLOv8). First build can take longer because Python AI dependencies are large.

Stop it:

docker compose down

Data persists in ./data on your host.

Windows one-click helper (auto-starts Docker Desktop if needed):

.\scripts\docker-up.ps1

---

Windows: one-line install (PowerShell 7)

One command to clone the repo, install Git/Go/Node (if missing), and build. Paste into PowerShell 7:

git clone https://github.com/248Tech/RTSPanda.git RTSPanda; cd RTSPanda; .\scripts\install-windows.ps1

To have the script also download mediamtx for you:

git clone https://github.com/248Tech/RTSPanda.git RTSPanda; cd RTSPanda; .\scripts\install-windows.ps1 -DownloadMediamtx

From CMD (requires Git and PowerShell 7 installed):

git clone https://github.com/248Tech/RTSPanda.git RTSPanda && cd RTSPanda && pwsh -NoProfile -File scripts\install-windows.ps1

Requires: For the one-liners you need PowerShell 7 and Git (so we can clone). If Git isn’t installed: git-scm.com or winget install Git.Git. The script will try to install Go and Node via winget if they’re missing.

When the install finishes, run:

.\backend\rtspanda.exe

and open http://localhost:8080.

---

Windows: PowerShell 7 step-by-step (power users)

Full control: install every dependency yourself, then build and run.

1. Install PowerShell 7 (if you’re still on Windows PowerShell 5)

winget install Microsoft.PowerShell --accept-package-agreements

Close and reopen your terminal; use “PowerShell 7” or pwsh so the rest of the commands run in PS7.

2. Install Git

winget install Git.Git --accept-package-agreements

Close and reopen the terminal so git is on your PATH.

3. Install Go

winget install GoLang.Go --accept-package-agreements

Again, reopen the terminal so go is available.

4. Install Node.js (LTS)

winget install OpenJS.NodeJS.LTS --accept-package-agreements

Reopen the terminal so node and npm are on your PATH.

5. Clone RTSPanda

git clone https://github.com/248Tech/RTSPanda.git
cd RTSPanda

6. (Optional) Download mediamtx

Download the Windows zip from mediamtx releases (e.g. mediamtx_v*_windows_amd64.zip), unzip it, and copy mediamtx.exe into the mediamtx folder inside RTSPanda:

New-Item -ItemType Directory -Force -Path mediamtx
# Then copy mediamtx.exe from your Downloads into .\mediamtx\

Or use the install script to do it for you:

.\scripts\install-windows.ps1 -DownloadMediamtx -SkipBuild

(Use -SkipBuild only if you already built and just want mediamtx.)

7. Build RTSPanda

.\build.ps1

You should see the frontend build, then the Go build. The result is backend\rtspanda.exe.

8. Run RTSPanda

.\backend\rtspanda.exe

You should see something like: RTSPanda listening on :8080 (data: ./data).

9. Use it

Open http://localhost:8080 in your browser. Add a camera in Settings → Cameras, then click it on the dashboard to watch the stream.

---

What RTSPanda can do

Feature	What it means
Live view	Dashboard with all cameras; click one for full-screen live video.
On-demand	The app only connects to a camera when someone is watching.
Recording	Turn on “Record to disk” per camera; get 1-hour MP4 files you can browse and download in the app.
Screenshots	While watching, hover over the video and click to save a PNG.
YOLOv8 tracking UI	Configure tracking per camera and run test detections directly in camera view.
Live overlays + history	Show bounding boxes on live video and browse grouped detection snapshots/events.
Discord alerts + triggers	Send rich webhook alerts with configurable detection/interval triggers, media options, cooldown, and mention per camera.
Manual Discord media	Send instant screenshot or clip from camera view with one click.
Legacy alert-rule API	Optional compatibility webhooks remain available for external automation flows.
REST API	Manage cameras, get stream status, list recordings, and trigger alerts from code or scripts.

---

Configuration

You can change behaviour with environment variables (no config file needed):

Variable	Default	What it does
PORT	8080	Port the web server uses.
DATA_DIR	./data	Where the database and recordings are stored.
MEDIAMTX_BIN	auto	Full path to mediamtx.exe if it’s not in the mediamtx folder.
RCLONE_BIN	rclone	rclone binary path for cloud video storage sync (Dropbox/Drive/OneDrive/Proton Drive).
FFMPEG_BIN	ffmpeg	FFmpeg path for frame capture used by object detection sampling.
DETECTOR_URL	http://127.0.0.1:8090	URL of the async detector worker (/detect, /health).
DETECTION_SAMPLE_INTERVAL_SECONDS	30	Global sample interval for camera frame capture.
DETECTION_WORKERS	2	Concurrent async detection worker requests from backend queue.
DETECTION_QUEUE_SIZE	128	Max queued snapshots waiting for detector service.
DISCORD_MOTION_CLIP_SECONDS	4	Default motion-clip length used when camera-specific value is missing.

AI worker (YOLO) tuning variables:

Variable	Default	What it does
YOLO_MODEL_PATH	/model/yolov8n.onnx	ONNX model path used by ai-worker runtime.
YOLO_CONFIDENCE	0.25	Base confidence threshold before backend camera filters are applied. Lower for more detections.
YOLO_IOU	0.45	NMS IoU threshold. Lower values suppress overlapping boxes more aggressively.
YOLO_MAX_DETECTIONS	100	Maximum objects returned per frame.
YOLO_LOG_LEVEL	INFO	AI worker logging verbosity (DEBUG, INFO, WARNING, etc).

Example (different port and data folder):

$env:PORT = "9000"; $env:DATA_DIR = "C:\rtspanda-data"; .\backend\rtspanda.exe

External Video Storage

RTSPanda can auto-sync recordings to:

• Local Server (NAS/SMB/NFS path)
• Dropbox
• Google Drive
• OneDrive
• Proton Drive

Setup guide: docs/EXTERNAL_VIDEO_STORAGE.md

---

REST API (for scripts and power users)

Everything the web UI does can be done over HTTP. Base URL: http://localhost:8080/api/v1

What	Method	Path
List cameras	GET	/cameras
Add camera	POST	/cameras
Get one camera	GET	/cameras/{id}
Update camera	PUT	/cameras/{id}
Delete camera	DELETE	/cameras/{id}
Stream status + HLS URL	GET	/cameras/{id}/stream
List recordings	GET	/cameras/{id}/recordings
Download recording	GET	/cameras/{id}/recordings/{filename}
Alert rules	GET / POST	/cameras/{id}/alerts
Send alert event (webhook)	POST	/alerts/{id}/events
Detection health	GET	/detections/health
Trigger test frame capture	POST	/cameras/{id}/detections/test-frame
Trigger test detection	POST	/cameras/{id}/detections/test
Send screenshot to Discord	POST	/cameras/{id}/discord/screenshot
Send recording to Discord	POST	/cameras/{id}/discord/record
List recent detection events	GET	/detection-events
Get snapshot for event	GET	/detection-events/{id}/snapshot
Health check	GET	/health

Example: add a camera with PowerShell:

Invoke-RestMethod -Method POST -Uri "http://localhost:8080/api/v1/cameras" -ContentType "application/json" -Body '{"name":"Front Door","rtsp_url":"rtsp://admin:password@192.168.1.10:554/stream","enabled":true}'

More examples and details: human/USER_GUIDE.md.

---

Security

• RTSPanda has no login screen. Use it on a trusted network, behind a VPN, or behind a reverse proxy (e.g. nginx with password).
• Camera passwords are stored in the SQLite database. Keep the data folder (and DATA_DIR) only readable by people you trust.
• Don’t expose the app directly to the internet without something in front of it (proxy, VPN, etc.).

---

Development

• Run backend only: cd backend; go run ./cmd/rtspanda
• Run frontend with hot reload: cd frontend; npm install; npm run dev (then open http://localhost:5173; API and HLS are proxied to the backend.)
• Lint: cd backend; go vet ./... and cd frontend; npm run lint

Full guide, RTSP URL tips, and troubleshooting: human/USER_GUIDE.md.

---

Project layout

RTSPanda/
├── backend/          # Go server and embedded web UI
├── frontend/         # React app (built and embedded into backend)
├── mediamtx/         # Put mediamtx.exe (or mediamtx) here
├── Dockerfile        # Docker image build
├── docker-compose.yml# One-command container run
├── scripts/          # install-windows.ps1, dev helpers
├── human/            # User guide
├── build.ps1         # Windows build
└── Makefile          # Mac/Linux build

---

License

MIT. See LICENSE.

RTSPanda — self-hosted, no cloud, no fuss.