{"protocol":"saig-colleague/1","id":"saigvoice","name":"SAIG Voice","role":"Speech: transcription, Wispr-style dictation polish, and call intelligence (tone, subtle complaints, missed signals, vulnerability cues)","ask":"Give me an audio or video file URL (or base64 audio) and I return the verbatim transcript, a polished version, or an analysis of the conversation.","engine":"https://console.skylite.group/api/ai?op=transcribe","guideVersion":27,"docs":"https://voice.skylite.group/","tools":[{"name":"voice.transcribe","method":"POST","path":"/api/voice?op=transcribe","write":false,"scope":"app","confirm":false,"idempotent":true,"description":"Verbatim diarised transcript of an audio or video file, with optional polish (dictation) or insights (call analysis).","input":{"type":"object","required":[],"properties":{"audioUrl":{"type":"string","description":"https URL of an audio or video file (mp4, m4a, mp3, webm, wav)"},"audioBase64":{"type":"string","description":"base64 audio ≤ 8 MB, with mime"},"mime":{"type":"string"},"mode":{"type":"string","enum":["dictate","note","converse","analyse"]},"polish":{"type":"object","description":"{style: clean|professional|concise|message|email|notes|prompt}"},"insights":{"type":"object","description":"{context, focus}"},"language":{"type":"string"},"keyterms":{"type":"array"},"idempotencyKey":{"type":"string"}}}},{"name":"voice.vocabulary","method":"POST","path":"/api/voice?op=vocabulary","write":true,"scope":"app","confirm":false,"description":"Add your app's proper nouns and jargon so they are transcribed correctly.","input":{"type":"object","required":["terms"],"properties":{"terms":{"type":"array"}}}},{"name":"voice.usage","method":"GET","path":"/api/voice?op=usage","write":false,"scope":"app","description":"Your app's transcription minutes in the last 30 days."}],"rules":["The verbatim transcript is always returned untouched; polish and insights are layers on top.","Audio is never stored; regulated apps are transcribed in the EU with training opt-out and analysed on Shield providers only.","Tone is judged from the words plus timing measured from the audio, not voice pitch or loudness."]}