Bedtime-Wind-Down

Bedtime Wind-Down

A dedicated "goodnight" ritual that closes out the user's day. Triggered by a simple bedtime command, this ability fetches tomorrow's first calendar event, generates a personalized soothing message, reads a calming quote, and plays ambient sleep sounds in a loop to help the user drift off. It operates entirely in a linear sequence with no conversation loops, prioritizing a peaceful, low-friction experience.

Trigger Words

• "bedtime"
• "bed time"
• "time for bed"
• "sleep mode"
• "night mode"
• "tuck me in"
• "going to sleep"

Setup

To fully enable the dynamic calendar schedule feature, you must configure the Composio API:

1. Open main.py.
2. Locate the COMPOSIO_API_KEY and COMPOSIO_USER_ID variables at the top of the BedtimeWindDownCapability class.
3. Replace "YOUR_COMPOSIO_API_KEY" and "YOUR_COMPOSIO_USER_ID" with your actual Composio credentials.

(Note: If the API keys are not set, the ability will gracefully default to a "no events tomorrow" state and still function perfectly.)

No API key is required for the ZenQuotes integration (it uses the free tier).

How It Works

1. User triggers the Ability with a hotword (e.g., "Time for bed").
2. Loads Preferences: The ability silently reads bedtime_prefs.json to check user settings (preferred ambient sound, duration, quote toggles). If it's the first run, it generates the default settings.
3. Fetches Tomorrow's Schedule: Calls the Google Calendar API (via Composio) to find the first event of the next day and calculates a suggested wake-up time.
4. LLM Generation: The LLM crafts a brief, calming 3-4 sentence wind-down message based on the schedule.
5. Speaks Message: The ability speaks the message using a specific, soft meditation voice (GBv7mTt0atIp3Br8iCZE) rather than the default personality voice.
6. Reads a Quote: Fetches a calming quote from the ZenQuotes API (or a local fallback list) and speaks it.
7. Plays Ambient Sounds: Enters "music mode" and dynamically loops the chosen ambient track (e.g., rain, ocean, white noise) for the configured duration (default 30 minutes).
8. Smart/Silent Exit: If the user says "Stop", it breaks the loop immediately. Otherwise, when the timer finishes, the ability exits silently without waking the user.

Key SDK Functions Used

• text_to_speech() — Speaks text utilizing a hardcoded, specific voice ID for a calming tone.
• text_to_text_response() — Generates the personalized wind-down message using the LLM synchronously.
• play_audio() — Streams the downloaded ambient MP3 bytes directly to the speaker.
• send_data_over_websocket() — Toggles music-mode on and off during audio playback to manage device state.
• check_if_file_exists(), read_file(), write_file() — Manages persistent JSON preferences locally on the device.
• resume_normal_flow() — Quietly returns control to the main personality.

Example Sequence

User: "Time for bed."
AI: (In a soft, calming voice) "Tomorrow you have a team standup at 9 AM, so waking up around 8 would give you plenty of time. Everything else can wait until morning. Rest well."
AI: "The best bridge between despair and hope is a good night's sleep. E. Joseph Cossman."
AI: "Starting ocean sounds. They'll play for 30 minutes. Sleep well."
(Ocean waves play seamlessly for 30 minutes... then the device silently powers down the session)