Documentation
¶
Overview ¶
Package assets embeds vendored binary model assets so the telephony package never needs a runtime file read to load them.
Index ¶
Constants ¶
This section is empty.
Variables ¶
var AudioForcedStopULaw []byte
AudioForcedStopULaw is the forced-stop clip, μ-law encoded at 8 kHz (1 byte/sample). It plays when a single utterance exceeds MaxUtteranceMS and the caller is cut off, then the call terminates through the same mark-echo flow as the farewell. Like FarewellULaw, nothing depends on its length: MarkEchoTimeout derives the post-playout wait from it.
var FarewellULaw []byte
FarewellULaw is the call-termination farewell clip, μ-law encoded at 8 kHz (1 byte/sample, so len/8000 == seconds). It says "Call me anytime... Bye!" in the voice-out TTS's F5 voice.
voice-out is a supertonic TTS sidecar, the kind of /v1/tts endpoint the driver package's Speak synthesizes through, so an agent whose replies are voiced in voice-out's F5 voice ends the call in the voice it spoke in. The telephony package itself never talks to voice-out. A reply reaches a call as audio frames a consumer hands to ReplyRouter.Route; handleSTTResult does not reply at all: it accumulates FullPass transcripts in the turn buffer and flushes that buffer to the TurnSink when a result ends the turn.
PROVENANCE.md records how to regenerate it. Nothing depends on its length: MarkEchoTimeout derives the post-playout wait from len(FarewellULaw), so a longer or shorter clip retimes itself.
var LLMThinkingULaw []byte
LLMThinkingULaw is a loopable "thinking" bed, μ-law at 8 kHz. It was first embedded with no playback path, for a future LLM-composed reply to loop while it waits for the real response.
Functions ¶
This section is empty.
Types ¶
This section is empty.