Launches Pi as an active live-call sidekick when Tuple transcription starts.

When call-transcription-started fires, this trigger ships a small Pi extension into a working directory and opens a terminal that hands off to tuple connect --harness pi. Connect resolves call state and gives Pi a context prompt; the bundled extension follows the call in the background.

Requires a Tuple build whose tuple CLI exposes connect --harness and transcription show. If your tuple CLI doesn't have them yet, this trigger won't work.

What Pi does on the call

The bundled extension (tuple-call-sidekick.ts) does one job well: it owns the live transcript feed in the background so Pi stays responsive to you. It reads the call through tuple transcription show --wait, looping with a fresh per-process --cursor for a gap-free, repeat-free catch-up and then live batches, and pushes new speech into Pi. So Pi is an active listener: it leaves a one-line · summary per batch, escalating to a 👋 interjection only when something matters — without ever running its own transcript loop. (The extension overrides connect's "follow the transcript yourself" instruction on startup so the call is never read twice.)

It surfaces the watch state on Pi's toolbar (mode and batches seen), and adds one tool — set_watch_mode — to trade responsiveness for quiet (realtime / balanced / low_noise) as the call's shape changes. Nothing is hard-coded about the model: Pi uses whichever provider and model you have configured as your default.

Unlike the other sidekick triggers (which are thin tuple connect launchers), Pi ships this extension because it benefits from owning the feed in the background rather than running the follow-loop in its own turn.

Customizing

Pi answers to a small set of watch words — its name plus common Whisper mis-hearings — so it reliably notices when it's addressed on the call. Edit the WATCH_WORDS constant near the top of tuple-call-sidekick.ts to add your own name and its likely mistranscriptions.

Prerequisites

• macOS
• Pi installed so pi works in a new terminal, with a provider authenticated (pi, then /login)
• The tuple CLI on your interactive shell PATH, with the subcommands listed above
  • Install it from the Tuple app: its Transcription settings have an Install button that links tuple onto your PATH.
• Tuple transcription enabled for the call

Installation

Drop this directory into your Tuple triggers folder:

~/.tuple/triggers/sidekick-pi/

The trigger fires the next time call transcription starts.

How it works

call-transcription-started fires with no call-specific arguments. This trigger:

1. Creates a working directory per start, ${TMPDIR:-/tmp}/tuple-pi-sidekick/<timestamp>-<pid>.
2. Copies tuple-call-sidekick.ts into .pi/extensions/ there. Pi auto-discovers .pi/extensions/*.ts from its working directory, so the extension is active the moment Pi starts.
3. Writes an executable launch-pi-sidekick.command wrapper into that directory.
4. Opens it in your preferred terminal via open (LaunchServices). With PREFERRED_TERM empty it opens in your default handler for .command files; set it to one of ghostty | iterm | alacritty | terminal to force one. No AppleScript, so it triggers no macOS accessibility prompt.
5. The wrapper starts a login-interactive zsh, cds to the working directory, and runs tuple connect --harness pi.

There is no dedup: each transcription-start gets its own directory, so stopping and restarting transcription in one call spawns a fresh sidekick while older ones keep running and stay queryable.

For local script testing without opening a terminal, set SIDEKICK_PI_DRY_RUN=1. It still installs the extension and writes the launcher, then exits without launching it.