Skip to content

v2.10.0 — Web UI catches up · feature parity with MCP

Choose a tag to compare

View all tags
@gfargo gfargo released this 17 May 02:36
· 31 commits to main since this release
v2.10.0
8b15bdd
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Highlights

The web UI now exposes everything the MCP layer does. Until now the browser interface was built around the v2.2 surface (AI chat + scenes + console + read-only group drawer). Eight releases later the MCP layer shipped 48 tools — 35 of the underlying endpoints had no UI affordance. This release closes that gap.

If you can do it via Claude Desktop, you can now do it via http://lights.local:5000.

What's new

Always-visible header

  • BLACKOUT button (top-right, red, pulsing) — calls /api/blackout, zeros every channel including strobe/macro state. Distinct from fade(0). No confirmation prompt — operators need this fast.

Tabs (was 2, now 6)

Tab What it does
AI Control Unchanged for now — full agentic chat ships in v2.11
Virtual Console Existing per-fixture sliders + Identify button per fixture (pulses 2s × 4, restores prior state)
Cue Lists List, builder modal, GO/STOP, live elapsed-time progress bar + cue counter
Chases List, visual scene-picker builder with reorder / direction / run-order / fade timings, start/stop
Quick Tools Kelvin slider (1800–10000K with gradient track), strobe rate slider (0–20Hz), palette assignment table per group
Diagnostics Pi health cards (CPU temp / load / memory / disk / uptime / USB / services, color-coded), one-click test_dmx, journal tail for the four allowlisted services

Groups drawer (read-only → CRUD)

  • Inline + New Group form: name, fixture picker (chips), optional description
  • Rename / Delete per group
  • Target-group chips on AI tab + Quick Tools group selects refresh automatically

Scene grid affordances

  • Each scene tile gains a ⋯ hover menu: Rename / Duplicate / Delete

Implementation

  • All vanilla JS, no build step. Template stays single-file (~2500 lines, was 1011)
  • Same design vocabulary as v2.2 (#0a0a0a / #111 / #222 panels, uppercase letter-spaced labels)
  • Lazy load: each tab fetches data only when first opened
  • Cue Lists tab polls /api/cue_lists/active every 1s while visible for elapsed-time updates; pauses on tab switch

Upgrade

```bash
git pull
./lightsctl.sh control-restart # serves the new template
```

No package dependency changes. No new endpoints. Pure frontend work consuming existing API.

Compatibility

  • Zero backend changes
  • Existing UI patterns preserved (the AI Control + Virtual Console tabs work identically to v2.9)
  • New CSS classes are namespaced (.btn-primary, .tool-panel, .modal-panel) — no collisions with existing styles
  • Local storage keys unchanged — command history persists across this upgrade

Documentation

  • Wiki Control-Server page — feature list updated with the new tabs + drawer CRUD + BLACKOUT + scene affordances
  • Marketing FAQ — new entry confirming web UI feature parity with the MCP layer

What's intentionally NOT in this release

Agentic chat — the "chat with the LLM, it calls MCP tools, reports back" experience ships as v2.11. The current AI Control tab is the one-shot natural-language interpreter that was there before; v2.11 replaces it with a proper conversational agent that has access to all 48 MCP tools.

Roadmap status

Web UI gap: closed. Tier 3 remaining: #10 (SSE event stream), #11 (multi-workspace), #12 (audio-reactive). Plus the agentic chat work tracked for v2.11.

Assets 2
Loading