BARK is inspired by the man's best friend. Built to protect your packages from porch pirates, BARK holds small packages and protects large mail with its fangs. BARK is a state-aware robotic sentry that merges edge computer vision, cloud-based LLM reasoning, and blockchain logging to secure package deliveries. The system continuously polls a live webcam feed using YOLO-World to detect people and packages. When triggered, it captures a frame and passes it to the Gemini 2.5 Flash API for multimodal intent verification—ensuring the person is actually interacting with the camera.

Once visually verified, the robot handles voice authentication with password protection using Python’s speech_recognition and parses the user's intent using Gemini's text generation (e.g., determining if a package is "TOO_BIG" to fit). Dynamic audio responses are streamed back in real-time using the ElevenLabs API for low-latency, emotive Text-to-Speech. If an unauthorized user approaches or fails authentication, the Python backend executes a dual-layer defense: it builds and signs an instruction to the Solana Devnet Memo Program to permanently log the intrusion timestamp to the blockchain, and it transmits a serial byte command via USB to a Seeed XIAO RP2040 microcontroller to trigger a hardware-level kinetic motor defense.

Built With

Share this project:

Updates