TimedTranscriptFragment

Defined in: packages/client/src/lib/voice-session/voice-observations.ts:45

One continuous transcript fragment. Times are milliseconds from the provider connection’s media origin, not wall-clock or confirmed audible-playback times. Reusing an ID revises that fragment in place. Text is retained verbatim.

Properties

endMs

readonly endMs: number

Defined in: packages/client/src/lib/voice-session/voice-observations.ts:57

End on the same timeline, at or after startMs.

---

final?

readonly optional final?: boolean

Defined in: packages/client/src/lib/voice-session/voice-observations.ts:59

Provider evidence that an utterance ends here. Absence means unknown.

---

id

readonly id: string

Defined in: packages/client/src/lib/voice-session/voice-observations.ts:47

Provider-local identity, unique within a connection generation; 1–256 characters.

---

role

readonly role: "user" | "assistant"

Defined in: packages/client/src/lib/voice-session/voice-observations.ts:49

Conversation role, independent of other simultaneous speakers.

---

speakerId

readonly speakerId: string

Defined in: packages/client/src/lib/voice-session/voice-observations.ts:51

Stable speaker identity within this connection; 1–256 characters.

---

startMs

readonly startMs: number

Defined in: packages/client/src/lib/voice-session/voice-observations.ts:55

Inclusive start on the connection’s media timeline.

---

text

readonly text: string

Defined in: packages/client/src/lib/voice-session/voice-observations.ts:53

Raw fragment text, including whitespace and punctuation.