kosmi-dj_skill

Overview

This skill turns the agent into a video DJ for Kosmi watch-party rooms using agent-browser automation. It connects to a Kosmi room, queues or plays video URLs, and can run a background DJ loop that auto-plays a playlist. Environment variables and a persistent agent-browser session allow reliable, repeatable control.

How this skill works

The skill drives a Chromium instance through the agent-browser CLI: it opens the room URL, takes accessibility-tree snapshots to discover interactive elements, and issues click/fill/press commands by ref ID. Scripts provide idempotent connect, single-play, and loop modes. The loop polls player state and advances the queue, writing a PID file for stop/status control.

When to use it

• You need to play or queue a YouTube/video URL into a Kosmi room programmatically.
• You want a persistent DJ that auto-plays and loops a playlist in a Kosmi watch party.
• You need to connect and authenticate to a Kosmi room from an automated agent with session persistence.
• You must debug Kosmi UI changes by inspecting accessibility snapshots.
• You want to run a headless or background DJ service on a server using agent-browser.

Best practices

• Store KOSMI_ROOM_URL, AGENT_BROWSER_SESSION_NAME and optional credentials in a .env loaded before runs.
• Use session persistence (AGENT_BROWSER_SESSION_NAME) to avoid repeated logins and keep cookies/localStorage.
• Always take snapshots and match elements by role + case-insensitive name, since ref IDs change each load.
• Use the provided snapshot-debug script when a UI element is not found; update element mappings accordingly.
• Run the loop script as a background process and rely on the PID file for graceful stop/status commands.

Example use cases

• Start a Kosmi watch party DJ that plays a curated list of YouTube links on a schedule.
• Add a requested video URL to the Kosmi room queue via agent command without manual intervention.
• Run a continuous DJ loop on a server that fetches new URLs and advances when each video ends.
• Diagnose Kosmi UI updates by dumping the accessibility snapshot and remapping button/text names.
• Temporarily take over playback to ensure a presentation or public stream always has content.