- Add service_name and image_tag variables to nomad job
- Use ${var.image_tag} in image ref instead of hardcoded 'latest'
- CI now passes git SHA as -var='image_tag=<sha>' to nomad job run
- Align Traefik tags with DevOpsMCP pattern (service_name var, rate limiting)
- Add canary update strategy and reschedule block
- Move service block to group level (nomad best practice)
- select() now fetches /api/v1/boards/{id} to get tasks array (lazy, once)
- updateTaskStatus() merges response correctly (spread instead of replacing)
- Progress % in sidebar uses done_count/task_count from list summary,
or computed from tasks array when board detail is loaded
- body had x-cloak but no x-data, so Alpine never initialized it and page stayed display:none
- added x-data to <body> tag — Alpine now processes the element and removes x-cloak
- added Makefile: make dev (hot-reload), make install, make open, make lint/fmt
- Use r.available_agents and r.available_skills field names from MCP proxy
- Fix extractContent to unwrap agents[0].content / skills[0].content
- Fix Alpine x-for key to include item.name for agents/skills
- Fix selected state comparison to handle name-keyed items
FastMCP 2.0 requires:
1. POST initialize → get mcp-session-id header
2. POST notifications/initialized with session ID
3. POST tools/call with session ID
Also prefer structuredContent over text/content in result extraction.
The devops-mcp-data host volume is not yet configured on autobox.i80.dk,
causing 'missing compatible host volumes' placement failure.
Comment out the volume + volume_mount blocks so the job can schedule
immediately. Knowledge catalog files are accessed via MCP proxy for now.
Instructions to re-enable (when ready):
Add to /etc/nomad.d/client.hcl on autobox:
host_volume "devops-mcp-data" {
path = "/opt/devops-mcp/data"
read_only = false
}
Then: mkdir -p /opt/devops-mcp/data && systemctl restart nomad
Also: switch auth to HARBOR_ROBOT_TOKEN pattern (matches devops-mcp.nomad),
add force_pull=true, set provider=consul.
