Release candidate v1.15.0 — ready for founder review (Stage 1 automated prep)

[rosscado]

This issue is the Stage 1 automated release-candidate prep (doc/release/delegation-criteria.md Stage 1, accepted by the founder 2026-07-07 per PR #538 / issue #523, operational now that the #525 weekly routine is active). It is opened automatically by the 2026-09-07 weekly maintenance routine because the trip threshold is met: 28 user-facing merges since the last published baseline (≥5 required), with no published-release git tag found — package.json (1.14.0) was used as the baseline.

Derived version: 1.15.0 (minor bump — 50 commits, 28 user-facing / 22 internal, from npm run release:plan).

Nothing has been bumped, built, tagged, or submitted — this issue only carries the prepared packet for review. Per Stage 1, all commands run were read-only (release:plan, release:packet); the gated steps (bump/build/tag/finalize/submit) remain founder-run, per AGENTS.md.

Submission packet (dist/submission-packet-v1.15.0.md, generated 2026-09-07)

Say, Pi — submission packet v1.15.0

Date: 2026-09-07 · Previous release: v1.14.0 (assumed baseline — no published-release tag found)

Payload: 50 commits — 28 user-facing, 22 under-the-hood.

What's new (this release)

Find a voice that feels right for your conversations.

✨ More voices, easier to choose
• Explore the Voices tab, listen to free samples, and compare voices before choosing. Listening never changes the voice you've saved.
• See which voice is in use and choose it for Pi or Claude. Your choices stay separate for each assistant.
• Browse the full collection in one place. HD voices are clearly marked, with a reminder of how they use your allowance.
• Better value voices from OpenAI, so your TTS credits go further.
• Switch back to Pi's own voices directly from its voice settings or Say, Pi's Voices tab.
• Settings now open in their own browser tab, so there's room to browse, and a warmer welcome greets you on first install.

🐛 More dependable conversations
• Your saved voice survives temporary catalog problems, and playback follows along when you change voice or sign in mid-conversation.
• When Say, Pi is speaking for Pi, Pi's own voice stays quiet — and comes back the moment you switch to it.
• Dictation keeps your text field in view, and voice controls follow the latest Pi, Claude, and ChatGPT layouts.
• More reliable audio startup and response playback.

Plus the usual under-the-hood work to keep Say, Pi fast and dependable.

Appendix — factual changelog digest

✨ New

• close the resume window when the assistant becomes audible (feat(endpointing): close the turn-outcome resume window at TTS start (#510) #643)
• probe every store-listing URL before publish (feat(release): probe every store-listing URL before publish #627)
• say which voice is actually speaking (Voices studio never says when the host's own voice is what you hear #599, Pi's own Voice settings gives no hint that a SayPi voice has overridden it #600) (feat(voices): say which voice is actually speaking (#599, #600) #601)
• the audition room (feat(voices): the audition room #586)
• surface API error detail instead of discarding it (Surface API error detail instead of discarding it #588)
• open settings as a browser tab instead of a popup window (prototype) (feat(settings): open settings as a browser tab instead of a popup window (prototype) #584)

🐛 Fixes

• adopt the reply we joined mid-write on a new conversation (fix(pi): adopt the reply we joined mid-write on a new conversation (#365) #641)
• stop hiding SayPi's per-message TTS controls on phones (fix(mobile): stop hiding SayPi's per-message TTS controls on phones #637)
• submit the voice turn when the tab is in the background (fix(prompt): submit the voice turn when the tab is in the background (#101) #640)
• also reject a partial Opus encode, not just a silent one (fix(audio): also reject a partial Opus encode, not just a silent one (#630) #634)
• delete the unreachable conservative preset and lock preset reachability (fix(vad): delete the unreachable conservative preset and lock preset reachability #633)
• never upload an Opus file with no audio (fix(audio): never upload an Opus file with no audio (#630) #632)
• match the call button to Pi's native send button (fix(pi): match the call button to Pi's native send button #626)
• preserve native Pi suppression through player changes Keep Pi native speech suppressed when its audio players are inserted or coexist #620 (Keep Pi native speech quiet across audio-player changes #621)
• keep refocused fields visible Keep the dictation button visible when a field quickly regains focus #622 (Keep the dictation button visible after rapid refocus #623)
• mute the host's voice when SayPi is the one speaking (fix(audio): mute the host's voice when SayPi is the one speaking (#602) #603)
• give new replies to the observer that can actually speak them (fix(pi): give new replies to the observer that can actually speak them (#597) #598)
• make the call button's presence a maintained invariant (fix(callbutton): make the call button's presence a maintained invariant (#593) #594)
• decorate the chat UI when the SPA routes INTO it (fix(bootstrap): decorate the chat UI when the SPA routes INTO it (#460) #592)
• make the removal observer see the audio element it watches (fix(audio): make the removal observer see the audio element it watches #591)
• make the Voices studio honest and level (fix(voices): make the Voices studio honest and level #585)
• stop the sidebar shunt — fill the container instead of shrink-to-fit (fix(settings): stop the sidebar shunt — fill the container instead of shrink-to-fit #583)
• stop the tab-switch layout shunt from scrollbar re-centering (fix(settings): stop the tab-switch layout shunt from scrollbar re-centering #582)
• make a decoration miss explainable without re-running (fix(sweep): make a decoration miss explainable without re-running (#570) #576)
• distinguish "never reached the field" from "no dictation button" (fix(sweep): distinguish "never reached the field" from "no dictation button" (#569) #575)
• absorb Cowork's hidden voice-mode placeholder so SayPi controls don't overlap (fix(claude): absorb Cowork's hidden voice-mode placeholder so SayPi controls don't overlap (#580) #581)
• retire the menu shortlist on hosts with no in-chat menu (fix(voices): retire the menu shortlist on hosts with no in-chat menu #577)
• absorb Pi's voice-surface consolidation (auto-read kebab, settings-only voices) (fix(pi): absorb Pi's voice-surface consolidation (auto-read kebab, settings-only voices) #573)

🔧 Under the hood

• 22 internal change(s): tooling, tests, refactors, and docs.

Full per-store checklists (Chrome/Edge/Firefox declarations, upload targets, watch-outs) are in the packet file generated at dist/submission-packet-v1.15.0.md (git-ignored build output — regenerate with npm run release:packet from this same commit range, or ask the preparing session to attach it).

The ask: review the packet above (and the linked PRs it cites) and, when ready, run the founder-gated steps per doc/release/README.md (npm run release:build -- --yes covers bump/build/tag/finalize/submit in one command per the current v1.14.0-era division of labour). Remove the awaiting-founder label once decided (ship it, defer it, or hold for specific items to land first).

Note per doc/release/delegation-criteria.md's Stage 1 evidence bar tracking: this is being logged as one of the "3 consecutive releases with zero substantive founder corrections" data points Stage 2 needs — worth a quick note on this issue if anything here (version, payload item, store metadata) needed a substantive fix, versus typo-level only.

[github-actions]

@rosscado — this is now awaiting a founder decision or sign-off (awaiting-founder applied by @rosscado). See the latest comment/description for what's being asked. Standing queue: open items with this label.

[github-actions]

@rosscado — this is now awaiting a founder decision or sign-off (awaiting-founder applied by @rosscado). See the latest comment/description for what's being asked. Standing queue: open items with this label.