Pipeless Agents

Pipeless Agents automates tasks by taking real-time video streams as input, connecting video sources, and automatically creating an actionable data stream that represents what is happening on the video. It emphasizes a serverless development experience for embedded vision tasks and supports building and deploying custom agents that process data from pipelines in real time.

A project consists of an agent and one or more video pipelines; the agent processes all pipelines simultaneously. Users can add pipelines by providing video sources and filters to extract structured data, and then the agent receives the filtered data and can trigger webhooks, store data, or perform other coded actions.

The system allows you to deploy agents with code stored in your own repository, enabling deployments to be as simple as pushing to a repository. It also offers a catalogue of pre-built agents for common use cases and supports end-to-end encryption and on-premise deployments for sensitive environments.

Pipeless Agents presents concepts like video sources, filters, agents, and projects, and provides a guided workflow to set up and customize agents by linking to Git repositories and adding pipelines to process live video streams. The documentation explains how agents receive data through filters and how developers can tailor agent behavior by modifying their repository code.

Use cases

• PPE monitor: Receive notifications when a worker does not wear PPE. Customizable PPE rules.
• Table occupancy monitor: Track how long people stay at a restaurant table.
• Face blurring: Blur faces in video streams.
• Checkout line analytics: Export metrics about time spent in checkout lines.
• People recognition: Recognize people and get notifications.