Skip to main content

Troubleshooting

Camera Issues​

Blink rate-limits authentication to protect accounts. Wait 60 minutes before trying again. This is enforced by Blink's servers, not by Aegis.

Blink's own servers are having problems. This isn't your account — it affects many users at once:

  • Wait 30–60 minutes
  • Check if the official Blink app on your phone also has issues
  • This will resolve on its own without any action on your part

Double-check your credentials. Make sure you can log into the official Blink app with the same email and password. Note that Blink passwords are case-sensitive.

Your saved session has gone stale. Click Clear stored data and try again, then re-enter your credentials. This commonly happens after a Blink server maintenance window.

Ring: Missing thumbnails​

Go to Ring Camera settings → Fix Previews. This regenerates thumbnails from the existing video files using FFmpeg. Check the Thumbnail Generation status indicator to confirm FFmpeg is available on your system.

Ring: "No authentication token available"​

Your Ring session may have expired. Unlink and re-link your account via the mobile app integration.

RTSP camera won't connect​

  1. Verify the URL format: rtsp://username:password@IP:554/stream
  2. Check network connectivity: your computer and camera must be on the same network
  3. Try Stream Verification: use the Stream Verification modal to test multiple URLs — ONVIF cameras often expose 3–5 URLs, but only one actually works
  4. Check the camera's web interface: browse to http://CAMERA_IP to verify the camera is responsive
  5. Check firewall settings: ensure port 554 (RTSP) is not blocked by your firewall

ONVIF cameras not discovered​

  • Ensure your camera supports ONVIF (check the manufacturer's specs page)
  • Your computer and camera must be on the same subnet (e.g., both on 192.168.1.x)
  • Some cameras require ONVIF to be explicitly enabled in their settings — log into the camera's web interface and look for an ONVIF toggle
  • If your computer has multiple network interfaces (Ethernet + WiFi), select the correct one in Advanced Settings
  • Try the Network Scanner as an alternative — it uses port scanning instead of multicast

Camera was working but stopped streaming​

The camera's IP address may have changed (common with DHCP). Re-run the ONVIF scan — Aegis automatically detects the new IP by matching the camera's MAC address and updates the stored configuration.

AI Issues​

Setup Nudge says "Set up your LLM"​

You need to connect a language model:

  1. ⚙️ Settings → Persona → LLM
  2. Paste an API key (OpenAI, Anthropic) or enable a local model from the Downloaded tab
  3. Full LLM guide →

Setup Nudge says "Vision Model"​

You need a vision model for camera analysis:

  1. ⚙️ Settings → Persona → VLM
  2. Download a model from the Staff Picks tab (start with SmolVLM2-256M)
  3. Click Load on the downloaded model
  4. Full VLM guide →

AI responses are slow​

This usually means inference is running on CPU. Check these in order:

  1. Check GPU detection: ⚙️ Settings → System → Hardware — is your GPU shown? are acceleration badges (CUDA/Metal) visible?
  2. Try a smaller model: SmolVLM2-256M is the fastest option for testing
  3. Close GPU-heavy apps: other applications (games, video editors, other AI tools) compete for GPU memory
  4. Use a cloud VLM: OpenAI Vision or Google Gemini provide fast responses without local compute
  5. Reduce analysis frame rate: lower the fps in Video Analysis settings from 5 fps to 0.5 fps

AI Engine won't start​

  1. Check ⚙️ Settings → System → AI Engine for status and error messages
  2. Restart the AI Engine from the settings page
  3. Make sure a model is actually downloaded and selected — an empty model list means nothing is available to load
  4. macOS: ensure Aegis has system permissions in System Preferences → Privacy & Security
  5. Windows: try running Aegis as administrator if the engine fails to bind to its port
  6. Linux: check that required libraries are present — ldd can help diagnose missing .so files

No AI descriptions on clips​

Verify that:

  1. Video Analysis is enabled in ⚙️ Settings → System → AI Engine → Video Analysis
  2. A VLM is loaded (check the VLM settings — should say "Active" with a green indicator)
  3. The LLaMA Server status shows Online in the VLM System tab

Alert Issues​

Event handler never fires​

  1. Check that the VLM is producing descriptions: open a clip in the player and verify AI descriptions appear
  2. Check that the handler is enabled: disabled handlers have a grayed-out toggle
  3. Review the condition: is it too specific? Try a broader condition first to verify the pipeline works
  4. Check the camera filter: make sure the handler is assigned to the correct camera (or "All Cameras")

Event handler fires too often​

  1. Check the trigger count — 100+ fires per day means the condition is too broad
  2. Make the condition more specific: add location, time, or activity context
  3. Assign to a specific camera instead of "All Cameras"
  4. Add time boundaries: "after 10pm" or "during the day" reduces noise

Skill Issues​

Skill install hangs on "Model download"​

Check your internet connection. Large models can be 1–8 GB and may take several minutes on slower connections. The download progress bar shows speed and percentage.

"Incompatible" shows on a skill​

Your system doesn't meet the skill's requirements. Common causes:

  • The skill needs a GPU you don't have (see requirements on the skill card)
  • Python version is too old — check with python3 --version
  • Insufficient RAM

Skill crashes immediately after starting​

  1. Open the Console (terminal icon in Skills panel) and check the error output
  2. The error is usually a missing dependency or a model that failed to download
  3. Try Restart — some transient errors resolve on retry
  4. Uninstall and reinstall the skill if the error persists

Performance Issues​

Running out of storage​

⚙️ Settings → System → Storage:

  • Reduce the retention period (e.g., from 30 days to 7 days)
  • Switch to a more aggressive storage mode (Free Space or Fixed Size)
  • Move clips to an external drive by changing the media path
  • Delete downloaded AI models you're not using

App feels slow​

  • Reduce the number of cameras shown simultaneously in Monitor view
  • Use a smaller VLM model (SmolVLM2-256M)
  • Close the Skills console if open — it buffers log output and can use memory
  • Close other applications that compete for GPU or RAM
  • On Windows, check Task Manager to see if CPU or GPU is at capacity

Platform-Specific​

macOS​

  • Grant permissions: Camera and screen recording access must be allowed in System Preferences → Privacy & Security
  • Apple Silicon performance: Macs get Metal GPU acceleration automatically — no configuration needed
  • Intel Mac: CPU inference only. Consider using a cloud VLM for acceptable speed.

Windows​

  • Firewall: If cameras aren't discovered, add Aegis AI as an allowed app in Windows Firewall
  • NVIDIA drivers: Ensure you have the latest NVIDIA drivers. The CUDA-enabled inference engine requires recent drivers.
  • Administrator: If the AI Engine fails to start, try running Aegis AI as administrator
  • Antivirus: Some antivirus software may flag inference engine binaries — add an exclusion for the Aegis AI data directory

Linux​

  • Camera access: May require adding your user to the video group: sudo usermod -aG video $USER (log out and back in)
  • AppImage: Make sure the file is executable: chmod +x Aegis*.AppImage
  • libfuse2: Some distributions need libfuse2 for AppImage: sudo apt install libfuse2
  • NVIDIA: Install the latest proprietary drivers (not nouveau) for CUDA support