@jambonz/schema 0.4.1 → 0.4.3
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.
|
@@ -29,6 +29,22 @@
|
|
|
29
29
|
"type": "string",
|
|
30
30
|
"description": "Language code for recognition."
|
|
31
31
|
},
|
|
32
|
+
"languages": {
|
|
33
|
+
"type": "array",
|
|
34
|
+
"items": { "type": "string" },
|
|
35
|
+
"minItems": 1,
|
|
36
|
+
"description": "Language hints for 'gpt-live-transcribe', which takes a list rather than a single code (e.g. ['en','fr']). Overrides the recognizer language; do not combine with 'language'."
|
|
37
|
+
},
|
|
38
|
+
"keywords": {
|
|
39
|
+
"type": "array",
|
|
40
|
+
"items": { "type": "string" },
|
|
41
|
+
"description": "Literal term hints (product names, acronyms, medications) for 'gpt-live-transcribe'. Hints, not required output. Defaults to the recognizer's 'hints' when unset."
|
|
42
|
+
},
|
|
43
|
+
"delay": {
|
|
44
|
+
"type": "string",
|
|
45
|
+
"enum": ["minimal", "low", "medium", "high", "xhigh"],
|
|
46
|
+
"description": "Latency/accuracy trade-off for 'gpt-live-transcribe': lower emits partial text sooner, higher gives the model more audio context and can improve accuracy."
|
|
47
|
+
},
|
|
32
48
|
"input_audio_noise_reduction": {
|
|
33
49
|
"type": "string",
|
|
34
50
|
"enum": ["near_field", "far_field"],
|
|
@@ -58,17 +74,17 @@
|
|
|
58
74
|
"type": "integer",
|
|
59
75
|
"minimum": 0,
|
|
60
76
|
"maximum": 3,
|
|
61
|
-
"description": "Local VAD aggressiveness (0-3). Used only with 'gpt-realtime-whisper'."
|
|
77
|
+
"description": "Local VAD aggressiveness (0-3). Used only with the client-endpointed models ('gpt-live-transcribe', 'gpt-realtime-whisper'), which reject server turn detection and rely on jambonz to end each turn."
|
|
62
78
|
},
|
|
63
79
|
"vadSilenceMs": {
|
|
64
80
|
"type": "integer",
|
|
65
81
|
"minimum": 1,
|
|
66
|
-
"description": "Milliseconds of silence to detect end-of-speech for local VAD. Used only with
|
|
82
|
+
"description": "Milliseconds of silence to detect end-of-speech for local VAD. Used only with the client-endpointed models."
|
|
67
83
|
},
|
|
68
84
|
"vadVoiceMs": {
|
|
69
85
|
"type": "integer",
|
|
70
86
|
"minimum": 1,
|
|
71
|
-
"description": "Milliseconds of voice to detect start-of-speech for local VAD. Used only with
|
|
87
|
+
"description": "Milliseconds of voice to detect start-of-speech for local VAD. Used only with the client-endpointed models."
|
|
72
88
|
}
|
|
73
89
|
},
|
|
74
90
|
"additionalProperties": false
|
package/package.json
CHANGED
|
@@ -63,6 +63,10 @@
|
|
|
63
63
|
"$ref": "../components/actionHook",
|
|
64
64
|
"description": "A webhook invoked for Dialogflow events during the conversation."
|
|
65
65
|
},
|
|
66
|
+
"toolHook": {
|
|
67
|
+
"$ref": "../components/actionHook",
|
|
68
|
+
"description": "A webhook invoked when the agent requests a client-side tool call (CX). Respond with {outputParameters} (or {error}) to resume the conversation."
|
|
69
|
+
},
|
|
66
70
|
"events": {
|
|
67
71
|
"type": "array",
|
|
68
72
|
"items": {
|