DeepCamera is an open-source AI camera platform that runs local inference for VLM/LLM-based analysis, with a modular skill catalog and an autonomous deployment agent. It integrates camera feeds, local models, and agent-driven alerting via chat apps.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
DeepCamera is an Open-Source AI Camera Skills Platform that runs locally for VLM scene analysis, object detection, and facial recognition. It emphasizes that all inference runs locally and supports skills like YOLO26 detection, depth estimation for privacy, and various integrations. It also provides a desktop companion app (SharpAI Aegis) for AI-assisted skill installation and agent-based interactions via Telegram, Discord, or Slack.
How it works
The system exposes a Skill Catalog where each skill is a self-contained module with its own model and parameters. Skills communicate via a JSONL stdin/stdout protocol, enabling a unified output regardless of backend (e.g., YOLO, Coral TPU, OpenVINO). An autonomous LLM deployment agent installs skills by reading each skill's manifest (SKILL.md) and configuring hardware-specific backends. Detection runs in parallel with VLM analysis, and there are hardware auto-detection and model conversion steps managed by an env_config.py. Aegis serves as the desktop UI to manage cameras, models, skills, and agent chat.
Getting started
The README describes connecting cameras (RTSP/ONVIF, webcams, iPhone), using a built-in local LLM & VLM, and one-click skill deployment. It also notes one-click HuggingFace downloads for models like Qwen, DeepSeek, SmolVLM, and LLaVA. The installation flow includes AI-assisted troubleshooting and automatic hardware detection to install appropriate backends.
Recent releases
Latest release: v2026.3 DeepCamera v2026.3 — Skill Platform Milestone (2026-03-14): 136 commits since the last release, transforming DeepCamera from a monolithic AI camera backend into a decentralized Skill Platform for Aegis-AI.
Traction
GitHub stars: 2969. Forks: 462. Open issues: 12. Language: JavaScript. License: MIT. Created: 2019-03-05. Last push: 2026-06-18.
Behind the repo
SharpAI Aegis is promoted as the desktop app companion for DeepCamera, enabling LLM-assisted skill installation and persistent agent chat across Telegram/Discord/Slack.
Caveats
License is MIT. The README mentions various legacy applications and CLI tooling, plus a multi-year history with Android releases and a migration toward a decentralized Skill Platform. No explicit issues or age-related caveats beyond the provided metadata.