v2.2.0 — interactive controls + /control back-channel
Turns the Debug Dashboard from a one-way display into a two-way control surface: a source can expose interactive sliders/numbers whose values flow back to it, and both the user and the AI can tune a running program live.
Added
- Interactive controls +
/controlback-channel — new widget typessliderandnumberare editable in the viewer; changing one sends the new value back to the source. First path on which data travels from AiDex to the program.- HTTP:
POST /control(set one value) andGET /control(flat{ id: value }map the source polls for current set-points). - MCP: the AI can drive controls too —
aidex_loggainscontrol_get/control_set, so Claude can tune a live program (e.g. a barge-in threshold, a gain) and watch the effect. - New
stepfield; new filesrc/loghub/control-store.ts. First consumer: GeminiPod (ESP32) barge-in tuning.
- HTTP:
- Sender-controlled plot Y-axis —
scale(linear/log),autoMin(lower bound follows the data minimum),decimals(footer precision). Audio recipe:scale:"log"+autoMin:true+ fixedmax+decimals:0. - Panel-Dashboard user guide (
docs/loghub-panel-dashboard.md) — full walkthrough: HTTP API, all six widget types, field reference, plot-scaling deep-dive, end-to-end example.
Changed
- Gauge LED colour separated from its text — new
statefield drives the LED colour independently ofvalue(the free display text). - npm package trimmed — ships only
build/+ the postinstall hook.
Fixed
- Dashboard no longer flickers on unchanged values.
aidex_global_guidelinelisttoken overflow — now a compact one-line-per-guideline index (1,246 → 17 lines).
npm install -g aidex-mcp@2.2.0