offload: speech-to-text stays two stages (V-486)

The one-call audio path is refused by measurement, so the inventory says
so where a future caller would read it.
This commit is contained in:
2026-08-05 17:06:05 +04:00
parent b789676244
commit 42a7bd88b2
+8 -1
View File
@@ -1,6 +1,6 @@
# Offloading model work to the workstation # Offloading model work to the workstation
*Last verified: 2026-08-03 @ 12530c8. Living doc: correct it in place, do not append.* *Last verified: 2026-08-05 @ b789676. Living doc: correct it in place, do not append.*
Owner's call, 2026-08-02. Vikunja #483 is the umbrella. Tasks #484 to #487 are the Owner's call, 2026-08-02. Vikunja #483 is the umbrella. Tasks #484 to #487 are the
work, and this file holds the shape and the rules all four must obey. work, and this file holds the shape and the rules all four must obey.
@@ -149,6 +149,13 @@ flips. It is wired anyway: `PhraseReminder` is on the same transport and is on.
Then the embedder above, **whisper.cpp** in `mavsttd`, and **piper** in `mavttsd`. Then the embedder above, **whisper.cpp** in `mavsttd`, and **piper** in `mavttsd`.
`mavwaked` uses no model at all: an energy-threshold VAD over 30ms frames. `mavwaked` uses no model at all: an energy-threshold VAD over 30ms frames.
Speech-to-text stays two stages when it moves. One call carrying both a clip and the router
prompt was measured on 05-08-2026. It scores 54.2% intent-only against 84.7% for whisper on
homesrv, on the same 72 cases. The model transcribes clips it then routes wrong, so a long
classification prompt and an audio part compete for attention. Transcribing on the
workstation and routing the text scores 83.3% at p50 997ms. So the transfer buys 375ms and
cleaner transcripts, not accuracy. See `docs/evals/2026-08-05-audio-in-routing.md`.
## Order ## Order
1. **Transport** (#484). Nothing else is possible until a seam can cross a host. 1. **Transport** (#484). Nothing else is possible until a seam can cross a host.