@jambonz/schema 0.4.17 → 0.4.19
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
|
@@ -16,7 +16,7 @@
|
|
|
16
16
|
},
|
|
17
17
|
"speechModel": {
|
|
18
18
|
"type": "string",
|
|
19
|
-
"description": "AssemblyAI speech model to use for recognition."
|
|
19
|
+
"description": "AssemblyAI speech model to use for recognition, e.g. 'universal-3-6-pro', 'universal-3-5-pro' or 'u3-rt-pro'."
|
|
20
20
|
},
|
|
21
21
|
"formatTurns": {
|
|
22
22
|
"type": "boolean",
|
|
@@ -60,6 +60,18 @@
|
|
|
60
60
|
"inactivityTimeout": {
|
|
61
61
|
"type": "number",
|
|
62
62
|
"description": "Timeout (seconds) for inactivity before closing the stream."
|
|
63
|
+
},
|
|
64
|
+
"languageCodes": {
|
|
65
|
+
"type": "array",
|
|
66
|
+
"items": {
|
|
67
|
+
"type": "string"
|
|
68
|
+
},
|
|
69
|
+
"description": "Languages to bias transcription toward (e.g. ['en', 'es']); code-switching among them is still allowed. Universal-3.5 Pro and Universal-3.6 Pro only. Default: all supported languages."
|
|
70
|
+
},
|
|
71
|
+
"voiceFocus": {
|
|
72
|
+
"type": "string",
|
|
73
|
+
"enum": ["near-field", "far-field"],
|
|
74
|
+
"description": "Isolate the primary voice and suppress background noise before transcription. Use 'near-field' for phones and headsets, 'far-field' for distant microphones. Default: off."
|
|
63
75
|
}
|
|
64
76
|
},
|
|
65
77
|
"additionalProperties": false
|
|
@@ -9,6 +9,11 @@
|
|
|
9
9
|
"type": "string",
|
|
10
10
|
"description": "xAI API key. Overrides the key configured in jambonz."
|
|
11
11
|
},
|
|
12
|
+
"model": {
|
|
13
|
+
"type": "string",
|
|
14
|
+
"description": "Transcription model to use, e.g. 'grok-voice-transcribe-2.0'. Defaults to whatever xAI currently defaults to.",
|
|
15
|
+
"examples": ["grok-voice-transcribe-2.0", "grok-voice-transcribe-1.0"]
|
|
16
|
+
},
|
|
12
17
|
"endpointing": {
|
|
13
18
|
"type": "integer",
|
|
14
19
|
"description": "Milliseconds of silence before an utterance is considered final. Range 0-5000."
|
|
@@ -40,6 +40,8 @@ A `keyterms` array can boost recognition of specific names, brands, or technical
|
|
|
40
40
|
|
|
41
41
|
- See [Keyterms guide](https://www.assemblyai.com/docs/streaming/keyterms-prompting)
|
|
42
42
|
|
|
43
|
+
Universal-3.6 Pro (`speechModel: "universal-3-6-pro"`) detects and code-switches across 32 languages automatically. `languageCodes` (e.g. `["en", "es"]`) biases it toward a subset. `voiceFocus` (`near-field` for phones and headsets, `far-field` for distant microphones) suppresses background speakers and noise before transcription; it is off by default.
|
|
44
|
+
|
|
43
45
|
### OpenAI (Whisper)
|
|
44
46
|
|
|
45
47
|
- [Speech to Text Guide](https://platform.openai.com/docs/guides/speech-to-text)
|