tts

TTS Voice Configuration

Interactive control panel for the TTS voice system.

Flow

1. Call get_voice_config to read current settings
2. If needsSetup is true: run the Setup flow (below) first
3. Display the current config as a formatted status panel:

TTS Voice Config
─────────────────
Mode:      [ambient]     Enabled: [yes]
Threshold: [500 chars]

Categories:
  completions  ✓  voice: jessica
  errors       ✓  voice: rob
  questions    ✓  voice: edward
  status       ✓  voice: daniel
  summaries    ✓  voice: daniel
  plans        ✓  voice: edward

Default voice: edward

4. Present all config options in a single AskUserQuestion call. Include current values in each question text so the user knows what they're changing from:

{
  "questions": [
    {
      "question": "TTS Mode? (currently: ambient)",
      "header": "Mode",
      "multiSelect": false,
      "options": [
        { "label": "Off", "description": "Nothing speaks" },
        { "label": "Minimal", "description": "Completions, errors, blocking questions" },
        { "label": "Ambient", "description": "All categories enabled" },
        { "label": "Full", "description": "Everything — summaries for long output" }
      ]
    },
    {
      "question": "Which categories should be enabled? (Other for: summaries, plans)",
      "header": "Categories",
      "multiSelect": true,
      "options": [
        { "label": "completions", "description": "Task completions" },
        { "label": "errors", "description": "Build/test failures" },
        { "label": "questions", "description": "Blocking questions" },
        { "label": "status", "description": "Progress updates" }
      ]
    },
    {
      "question": "Default voice? (currently: edward) Other for: archer, donovan, juniper, mark-casual, mark-natural, cassidy, spuds, rob, adam",
      "header": "Voice",
      "multiSelect": false,
      "options": [
        { "label": "edward", "description": "British, deep, composed (default)" },
        { "label": "jessica", "description": "Eloquent, dramatic" },
        { "label": "daniel", "description": "Steady broadcaster" },
        { "label": "austin", "description": "Deep, raspy, authentic" }
      ]
    },
    {
      "question": "Summary threshold? (currently: 500 chars)",
      "header": "Threshold",
      "multiSelect": false,
      "options": [
        { "label": "300", "description": "Speak more often" },
        { "label": "500 (default)", "description": "Balanced" },
        { "label": "800", "description": "Only long responses" },
        { "label": "Disabled", "description": "Never auto-summarize" }
      ]
    }
  ]
}

5. Apply changes:
   • Mode → category sync: If mode changed but categories were NOT explicitly toggled, auto-apply the mode preset:
     • Off: all categories false, enabled: false
     • Minimal: completions + errors + questions only
     • Ambient: all categories true
     • Full: all categories true
   • If user explicitly selected categories in Q2, respect that override regardless of mode
   • No change: If all answers match current config, skip the write and say "No changes needed."
6. Write all changes in one set_voice_config call
7. Confirm with speak("TTS settings updated.", category: "status")

Setup Flow (first-time only)

When get_voice_config returns needsSetup: true:

1. Tell the user: "TTS needs an ElevenLabs API key to work. Get one free at https://elevenlabs.io"
2. Use AskUserQuestion to ask them to paste their key:
   • Question: "Paste your ElevenLabs API key (starts with sk_)"
   • No predefined options — user types their key in the "Other" field
3. Call setup_tts with the key
4. If successful, call speak("Voice setup complete. Welcome to TTS mode.", category: "completions") to test it
5. Then continue to the normal config flow above

Mode Reference

Mode	What speaks
Off	Nothing
Minimal	Task completions, build failures, blocking questions
Ambient	All of minimal + status updates + long-output summaries
Full	Everything — every response gets spoken (summaries for long output)

Plans category: Enabled in Ambient and Full modes. Speaks when presenting or revising a plan.

Available Voices

edward, daniel, jessica, austin, archer, donovan, juniper, mark-casual, mark-natural, cassidy, spuds, rob, adam