• fix: increase API call timeout to 5 minutes for stack actions (update, restart, start, stop, check-updates)

Pair with Dockge v2.1.5 for the server-side timeout fix.

• feat: refresh burst after service actions — polls every 30s for 5 minutes for near-real-time entity updates without SSE

SSE Push Notifications

Real-time push updates from Dockge via Server-Sent Events.

• SSE client in coordinator with auto-reconnect and 90s heartbeat watchdog
• Processing state (blinking blue) now works for ALL trigger sources
• Polling interval increased to 600s (SSE handles real-time, polling is fallback)
• iot_class changed to local_push
• 18 unit tests + GitHub Actions CI

Requires dockge v2.2.0 or later for SSE support. Falls back to polling gracefully with older versions.

What's New

• Stack processing state: Stacks now show a processing attribute on their binary sensor when an operation is in progress
• Reliable per-server matching: Stack binary sensors now expose stack_name and agent_name attributes, enabling correct matching without relying on entity_id patterns
• System prune service: New dockge.system_prune service to clean up unused Docker images, containers, and networks

Bug Fixes

• Fixed cross-server false positives where operating on one server's stack would incorrectly mark the same-named stack on another server as processing
• Demoted debug logging from warning to debug level

• Version sensor: New sensor.dockge_server_{agent}_version per agent, showing Dockge version
• Device sw_version: Agent device info now includes software version
• Remove auto-cleanup: Stale devices/entities are no longer auto-deleted (prevents agents disappearing during restarts). Users can manually delete from HA Devices page.

Paired with dockge-managed v2.1.1.

Changes

Bug Fixes

• Preserve stack devices for temporarily disconnected agents — stale device cleanup no longer removes devices/entities for agents that are known but returned no stacks

Features

• Refresh entities after service calls — all service handlers now trigger an immediate coordinator refresh so entity states update right away

Recommended

Pair with dockge-managed v2.1.0 for reliable multi-agent support without requiring a browser session.

Full Changelog: v1.2.0...v1.3.0

What's New

• Global summary sensor — aggregates running container count across all agents, with per-agent breakdown in attributes (total_stacks, total_containers, running_containers, agents)

Fixes

• Reconnection resilience — coordinator preserves stacks from known agents that temporarily return no data, preventing entities from going unavailable after a brief disconnect
• Increased API timeout — stacks endpoint timeout raised from 10s to 30s to handle multi-agent proxy latency

Fix

• Fix stack device names not showing agent name (e.g. "agents (Voltorb)") — binary_sensor and switch platforms were not passing multi_agent flag to stack_device_info, causing device names to be overwritten without the agent suffix.

What's New

• Agent summary sensor per server — state = running container count, with attributes: stacks (list), total_stacks, total_containers, running_containers
• Container sensor attributes — stack_name and agent_name now exposed, enabling simple Jinja2 filtering
• Consistent device naming — stack devices include agent name in multi-agent setups (e.g. agents (Voltorb))
• Stale entity cleanup — automatically removes orphaned entities (not just devices) on startup and data refresh