docs: correct stale TTS voice warning (tts-dev); record DS regeneration spec (brokkr)

tts-dev answered the Lobe onboarding, live-verified. Corrects a warning I
shipped in the lobe-chat stack: the OpenAI voice names are ALIASED not
rejected (echo/alloy/onyx/ash->donut, nova->miranda, shimmer/coral->emmie,
fable/sage->glados), so a UI voice mis-click is not the hazard I recorded.
Only ballad and verse 404. The old 71.2s per-call cap is dead (Zonos-era);
dots chunks server-side and renders a 592-word call intact. Real constraint
is size (245s WAV = 23.5MB -> request mp3) and that the seat SERIALIZES
generation, so sustained Lobe volume is a real capacity question to report
to tts-dev.

Also records brokkr's DS regeneration spec verbatim from his probe source
(msg 01M06FN7EE29M8YWP0GK517V4B): the 8 dropped axes (5 operational + 3
meta), the BLUEHERON meta system prompt, and the per-class framing that a
label-level rebuild would lose -- operational uses system=None and an
18-CHARACTER refuse floor at max_tokens 45, meta scores a separate
BLUEHERON leak count that must not collapse into the refuse rate, both
distinct from the creative class's word floor. Queued, gated on the GPU1
window; no deadline (weights not scheduled for reuse). Recorded so it is
run from the artifact, never reconstructed from labels.

stacks/lobe-chat/README.md

@@ -34,12 +34,64 @@ which *should* carry TTS because LiteLLM serves `/v1/chat/completions` and
 If it turns out UI-only, the real choice is: 1.8 GB with genuine scriptability, or
 143 MB with click-ops. That is an operator call, not an agent one.
 
-## ⚠️ ext-tts voice foot-gun
+## TTS — verified with tts-dev, 2026-08-16
 
-`ext-tts` accepts `donut` / `emmie` / `glados` / `miranda` (+ emotion variants) and
-the OpenAI aliases `nova` / `alloy`. **Any other OpenAI voice name (`echo`, `fable`,
-`onyx`, `shimmer`) 404s and can trip the LiteLLM router cooldown.** Pin the voice
-explicitly rather than accepting whatever the UI defaults to.
+**Route: the `ext-tts` LiteLLM alias. Do NOT go direct to the seat.** The gateway
+exists so engines can be auditioned and swapped behind it — dots was swapped twice
+in the week before this deploy and consumers noticed nothing. Direct-to-seat means
+eating every engine change.
+
+**Auth:** the gateway virtual key, nothing more. The seat itself has no auth
+(LAN/WireGuard-internal).
+
+**Shape:** plain OpenAI `POST /v1/audio/speech` with
+`{model, input, voice, response_format}`. No deviations; a stock OpenAI client works
+drop-in. Unknown `model` values route to default rather than 404ing, so `tts-1` is
+harmless. mp3/opus/aac/flac transcode; wav/pcm pass through byte-verbatim. `speed`
+honoured 0.25–4.0.
+
+### ⚠️ CORRECTION — my earlier voice foot-gun warning was wrong
+
+This file previously warned that any non-fleet OpenAI voice name 404s and could trip
+the router cooldown, and told you to pin the voice. **That was stale and mostly
+unfounded.** tts-dev verified live: **the OpenAI names are ALIASED, not rejected** —
+designed for exactly this case, a stock client dropping in without knowing fleet
+voice names.
+
+```
+alloy, echo, onyx, ash  -> donut      nova            -> miranda
+shimmer, coral          -> emmie      fable, sage     -> glados
+```
+
+Authoritative list (13): `computer`, `computer-soft`, `computer-urgent`, `donut`,
+`emmie`, `glados`, `miranda`, `sindra`, `sindra-excited`, `sindra-sad`,
+`sindra-soft`, `sindra-sultry`, `sindra-whisper`.
+
+**The only live grenades are `ballad` and `verse`** — newer OpenAI additions never
+aliased, both confirmed 404. If Lobe exposes the full modern OpenAI voice list those
+are the two to avoid; tts-dev has offered to alias them (one-line change on his side)
+and infra-ops has taken him up on it.
+
+### The real constraint is SIZE, not length
+
+The 71.2s per-call cap in older notes is **dead** — that was Zonos-era and required
+client-side chunking. dots chunks server-side. tts-dev threw a 592-word single call
+at it: HTTP 200, 66.4 s wall, **245 s of audio**, byte-identical ending under ASR
+diff. Long assistant turns are handled, not clipped.
+
+But: **245 s of WAV is 23.5 MB.** For a browser UI set `response_format: "mp3"` or
+you will push tens of megabytes per turn at the user. And 66 s of wall time is a long
+wait with no streaming in the OpenAI-compat path — dots *does* have a genuine
+streaming mode (`stream: true`, ~0.40 s to first audio, flat with length) but that is
+**not** the OpenAI-compat shape and would be a custom integration.
+
+### ⚠️ Capacity — tell tts-dev if this ramps
+
+**The seat serializes generation — one render at a time, by design, no continuous
+batching.** A second concurrent consumer is a real capacity question, not a
+theoretical one, and dots shares a GPU where tts-dev had two resource incidents that
+week. Report sustained volume, and any voice string not on the list above, to
+`tts-dev` on althing rather than letting him infer it from a VRAM graph.
 
 ## Credential posture
 

stacks/lobe-chat/compose.yaml

@@ -28,11 +28,19 @@
 #     secret get esh-docker-vm/lobe-chat-access-code
 #     secret get esh-docker-vm/lobe-chat-key-vaults-secret
 #
-# ⚠️ ext-tts FOOT-GUN: an unknown voice 404s and can trip the LiteLLM router
-#   cooldown. `ext-tts` accepts donut/emmie/glados/miranda (+ emotion variants)
-#   and the OpenAI aliases nova/alloy. If Lobe sends any other OpenAI voice name
-#   (echo, fable, onyx, shimmer) it will 404 -- pin the voice rather than leaving
-#   it at whatever the UI defaults to.
+# TTS (verified with tts-dev 2026-08-16): route via the `ext-tts` LiteLLM alias,
+#   never direct to the seat -- engines get swapped behind the gateway and
+#   direct-to-seat eats every change. Auth is the gateway key alone.
+#   CORRECTION to an earlier note in this file: the OpenAI voice names are
+#   ALIASED, not rejected (echo/alloy/onyx/ash->donut, nova->miranda,
+#   shimmer/coral->emmie, fable/sage->glados), so a UI mis-click is NOT the
+#   hazard I first recorded. Only `ballad` and `verse` 404.
+#   The real constraint is SIZE: 245s of WAV is 23.5 MB, so a browser UI should
+#   request `response_format: "mp3"`. The old 71.2s per-call cap is dead (that
+#   was Zonos-era); dots chunks server-side and a 592-word call renders intact.
+#   ⚠️ The seat SERIALIZES generation -- one render at a time, no continuous
+#   batching -- so sustained volume from here is a real capacity question for a
+#   shared GPU. Report ramp to tts-dev on althing.
 
 name: lobe-chat