Skip to main content

Vision Model (VLM) Setup

The VLM is what lets Aegis AI see and understand camera frames. Every time a clip is recorded, the VLM analyzes it and generates a text description. This powers alerts, timeline search, and the agent's situational awareness.

Open VLM configuration from ⚙️ Settings → Persona → VLM.


Built-in (Local Models)

Runs vision models locally using the bundled llama-server. No API key or internet needed after the model is downloaded.

Staff Picks

Curated models tested with Aegis AI, sorted by size:

ModelSizeVideo SupportNotes
SmolVLM2 256M Video0.4 GBSmallest video model, runs on almost any hardware
SmolVLM2 500M Video0.7 GBLightweight, good for quick testing
nanoLLaVA1.3 GBUltra-compact 430M parameter model
InternVL3 1B1.5 GBFast and efficient with excellent vision
LFM2.5-VL 1.6B1.6 GBTop pick — Liquid Foundation Model
SmolVLM2 2.2B1.7 GBCompact, great for resource-constrained setups
Qwen3-VL 2B Instruct1.8 GBStrong reasoning and instruction-following
Qwen2.5-VL 3B2.1 GBBalanced size and performance
LLaVA Phi-3 Mini2.7 GBMicrosoft Phi-3 backbone
MiniCPM-V 2.62.8 GBStrong OCR, excellent for security cameras
Moondream23.5 GBLatest Moondream with improved vision
Gemma 3 4B3.7 GBGoogle Gemma 3, great general-purpose
LLaVA 1.6 Mistral 7B5.0 GBMistral 7B backbone, high quality

Click Download on any staff pick card. The download progress bar shows percentage and speed.

Downloaded Tab

Lists all downloaded GGUF vision models with:

  • Model name, quantization, size
  • Load / Eject / Delete buttons (same behavior as LLM)

Search Tab

Search HuggingFace for additional vision-language models. Same filters as the LLM Search tab (sort, size presets).

System Tab

Shows llama-server status:

  • Running / Stopped indicator with PID
  • Currently loaded model ID
  • Server health check (polled every 3 seconds)

You can start/stop the server and select which downloaded model to serve.

Authentication Tab

Manage your HuggingFace token for accessing gated model repositories.


API Providers

Cloud Providers

Connect to a cloud vision API. Supported providers:

ProviderModelsAPI Key Source
OpenAIgpt-4o-mini, gpt-4o, gpt-4-turboplatform.openai.com
Anthropicclaude-3.5-sonnet, claude-3-haikuconsole.anthropic.com
Googlegemini-1.5-flash, gemini-1.5-proaistudio.google.com

Setup:

  1. Select a provider tab.
  2. Enter your API Key. Click Save Key.
  3. Select a model from the dropdown.
  4. Click Select to activate and validate the connection.

Cost Estimation

When using a cloud provider, Aegis AI displays a real-time cost estimate:

MetricDescription
Per ClipCost to analyze a single clip (e.g. $0.0043)
MonthlyProjected monthly spend based on your usage pattern
Clips/DayAverage clips recorded per day
Total ClipsTotal clips analyzed in the last 30 days
Projected MonthlyEstimated clips next month

If you have fewer than a few days of data, Aegis shows a default estimate based on 50 clips/day.

An expandable Usage Trend chart shows daily clip counts and cost over time.

External (OpenAI-Compatible)

Point Aegis to any server with an OpenAI-compatible /v1/chat/completions endpoint that supports vision.

  1. Enter the Endpoint URL (default: http://localhost:1234/v1).
  2. Click Load Models to fetch the model list.
  3. Select a model.
  4. Click Select — the connection is auto-validated, and latency is displayed (e.g. Connected (142ms)).

Compatible with: LM Studio, vLLM, LocalAI, text-generation-webui, and similar.


How the VLM Is Used

  1. A camera records a clip (triggered by motion, schedule, or a skill).
  2. Aegis extracts a frame or set of frames from the clip.
  3. The frame(s) are sent to the active VLM with the configured analysis prompt.
  4. The VLM returns a text description (e.g. "Person walking on driveway carrying a package").
  5. This description is stored with the clip and used for alerts, timeline search, and agent context.

Troubleshooting

IssueSolution
"LLaMA Server: Offline" in System tabClick Start, or check that a model is downloaded and selected.
Cloud API returns errorsVerify the API key. Check the provider's status page.
Analysis too slow with local modelUse a smaller model (SmolVLM2 256M or 500M). Reduce the frame rate in Video Analysis settings.
External endpoint returns empty model listVerify the server is running and the URL includes /v1.
Cost estimate shows "$0.00"Wait for at least 1 day of clip data, or select a cloud provider to see the default estimate.