livekit-plugins-google 0.3.0__py3-none-any.whl → 1.3.8__py3-none-any.whl

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.
Files changed (20) hide show
  1. livekit/plugins/google/__init__.py +25 -7
  2. livekit/plugins/google/beta/__init__.py +13 -0
  3. livekit/plugins/google/beta/gemini_tts.py +258 -0
  4. livekit/plugins/google/llm.py +501 -0
  5. livekit/plugins/google/log.py +3 -0
  6. livekit/plugins/google/models.py +145 -31
  7. livekit/plugins/google/realtime/__init__.py +9 -0
  8. livekit/plugins/google/realtime/api_proto.py +66 -0
  9. livekit/plugins/google/realtime/realtime_api.py +1252 -0
  10. livekit/plugins/google/stt.py +518 -272
  11. livekit/plugins/google/tools.py +11 -0
  12. livekit/plugins/google/tts.py +447 -0
  13. livekit/plugins/google/utils.py +286 -0
  14. livekit/plugins/google/version.py +1 -1
  15. livekit_plugins_google-1.3.8.dist-info/METADATA +63 -0
  16. livekit_plugins_google-1.3.8.dist-info/RECORD +18 -0
  17. {livekit_plugins_google-0.3.0.dist-info → livekit_plugins_google-1.3.8.dist-info}/WHEEL +1 -2
  18. livekit_plugins_google-0.3.0.dist-info/METADATA +0 -47
  19. livekit_plugins_google-0.3.0.dist-info/RECORD +0 -9
  20. livekit_plugins_google-0.3.0.dist-info/top_level.txt +0 -1
livekit/plugins/google/models.py CHANGED
@@ -3,83 +3,197 @@ from typing import Literal
3
3
  # Speech to Text v2
4
4
 
5
5
  SpeechModels = Literal[
6
- "long", "short", "telephony", "medical_dictation", "medical_conversation", "chirp"
6
+ "long",
7
+ "short",
8
+ "telephony",
9
+ "medical_dictation",
10
+ "medical_conversation",
11
+ "chirp",
12
+ "chirp_2",
13
+ "chirp_3",
14
+ "latest_long",
15
+ "latest_short",
7
16
  ]
8
17
 
9
18
  SpeechLanguages = Literal[
10
- "en-US",
11
- "ja-JP",
12
- "en-IN",
13
- "en-GB",
14
- "hi-IN",
15
19
  "af-ZA",
16
- "sq-AL",
17
20
  "am-ET",
21
+ "ar-AE",
22
+ "ar-BH",
23
+ "ar-DZ",
18
24
  "ar-EG",
19
- "hy-AM",
25
+ "ar-IL",
26
+ "ar-IQ",
27
+ "ar-JO",
28
+ "ar-KW",
29
+ "ar-LB",
30
+ "ar-MA",
31
+ "ar-MR",
32
+ "ar-OM",
33
+ "ar-PS",
34
+ "ar-QA",
35
+ "ar-SA",
36
+ "ar-TN",
37
+ "ar-YE",
38
+ "as-IN",
20
39
  "ast-ES",
21
40
  "az-AZ",
22
- "eu-ES",
23
41
  "be-BY",
24
- "bs-BA",
25
42
  "bg-BG",
26
- "my-MM",
43
+ "bn-BD",
44
+ "bn-IN",
45
+ "bs-BA",
27
46
  "ca-ES",
28
47
  "ceb-PH",
29
48
  "ckb-IQ",
30
- "zh-Hans-CN",
31
- "yue-Hant-HK",
32
- "zh-TW",
33
- "hr-HR",
49
+ "cmn-Hans-CN",
50
+ "cmn-Hant-TW",
34
51
  "cs-CZ",
52
+ "cy-GB",
35
53
  "da-DK",
36
- "nl-NL",
54
+ "de-AT",
55
+ "de-CH",
56
+ "de-DE",
57
+ "el-GR",
37
58
  "en-AU",
59
+ "en-CA",
60
+ "en-GB",
61
+ "en-HK",
62
+ "en-IE",
63
+ "en-IN",
64
+ "en-NZ",
65
+ "en-PK",
66
+ "en-SG",
67
+ "en-US",
68
+ "es-419",
69
+ "es-AR",
70
+ "es-BO",
71
+ "es-CL",
72
+ "es-CO",
73
+ "es-CR",
74
+ "es-DO",
75
+ "es-EC",
76
+ "es-ES",
77
+ "es-GT",
78
+ "es-HN",
79
+ "es-MX",
80
+ "es-NI",
81
+ "es-PA",
82
+ "es-PE",
83
+ "es-PR",
84
+ "es-SV",
85
+ "es-US",
86
+ "es-UY",
87
+ "es-VE",
38
88
  "et-EE",
39
- "fil-PH",
89
+ "eu-ES",
90
+ "fa-IR",
91
+ "ff-SN",
40
92
  "fi-FI",
93
+ "fil-PH",
94
+ "fr-BE",
41
95
  "fr-CA",
96
+ "fr-CH",
42
97
  "fr-FR",
98
+ "ga-IE",
43
99
  "gl-ES",
44
- "ka-GE",
45
- "de-DE",
46
- "el-GR",
47
100
  "gu-IN",
48
101
  "ha-NG",
49
- "iw-IL",
50
102
  "hi-IN",
103
+ "hr-HR",
51
104
  "hu-HU",
52
- "is-IS",
105
+ "hy-AM",
53
106
  "id-ID",
107
+ "ig-NG",
108
+ "is-IS",
109
+ "it-CH",
54
110
  "it-IT",
111
+ "iw-IL",
55
112
  "ja-JP",
56
113
  "jv-ID",
57
- "kea-CV",
114
+ "ka-GE",
58
115
  "kam-KE",
59
- "kn-IN",
116
+ "kea-CV",
60
117
  "kk-KZ",
61
118
  "km-KH",
119
+ "kn-IN",
62
120
  "ko-KR",
63
121
  "ky-KG",
64
- "lo-LA",
65
- "lv-LV",
122
+ "lb-LU",
123
+ "lg-UG",
66
124
  "ln-CD",
125
+ "lo-LA",
67
126
  "lt-LT",
68
127
  "luo-KE",
69
- "lb-LU",
128
+ "lv-LV",
129
+ "mi-NZ",
70
130
  "mk-MK",
131
+ "ml-IN",
132
+ "mn-MN",
133
+ "mr-IN",
134
+ "ms-MY",
135
+ "mt-MT",
136
+ "my-MM",
137
+ "ne-NP",
138
+ "nl-BE",
139
+ "nl-NL",
71
140
  "no-NO",
141
+ "nso-ZA",
142
+ "ny-MW",
143
+ "oc-FR",
144
+ "om-ET",
145
+ "or-IN",
146
+ "pa-Guru-IN",
72
147
  "pl-PL",
148
+ "ps-AF",
73
149
  "pt-BR",
74
150
  "pt-PT",
75
151
  "ro-RO",
76
152
  "ru-RU",
77
- "es-CO",
78
- "es-MX",
79
- "es-US",
153
+ "rup-BG",
154
+ "rw-RW",
155
+ "sd-IN",
156
+ "si-LK",
157
+ "sk-SK",
158
+ "sl-SI",
159
+ "sn-ZW",
160
+ "so-SO",
161
+ "sq-AL",
162
+ "sr-RS",
163
+ "ss-Latn-ZA",
164
+ "st-ZA",
165
+ "su-ID",
166
+ "sv-SE",
167
+ "sw",
168
+ "sw-KE",
169
+ "ta-IN",
170
+ "te-IN",
171
+ "tg-TJ",
80
172
  "th-TH",
173
+ "tn-Latn-ZA",
81
174
  "tr-TR",
175
+ "ts-ZA",
82
176
  "uk-UA",
177
+ "umb-AO",
178
+ "ur-PK",
179
+ "uz-UZ",
180
+ "ve-ZA",
83
181
  "vi-VN",
84
- "da-DK",
182
+ "wo-SN",
183
+ "xh-ZA",
184
+ "yo-NG",
185
+ "yue-Hant-HK",
186
+ "zu-ZA",
187
+ ]
188
+
189
+ Gender = Literal["male", "female", "neutral"]
190
+
191
+ ChatModels = Literal[
192
+ "gemini-2.5-pro-preview-05-06",
193
+ "gemini-2.5-flash-preview-04-17",
194
+ "gemini-2.5-flash-preview-05-20",
195
+ "gemini-2.0-flash-001",
196
+ "gemini-2.0-flash-lite-preview-02-05",
197
+ "gemini-2.0-pro-exp-02-05",
198
+ "gemini-1.5-pro",
85
199
  ]
livekit/plugins/google/realtime/__init__.py ADDED
@@ -0,0 +1,9 @@
1
+ from .api_proto import ClientEvents, LiveAPIModels, Voice
2
+ from .realtime_api import RealtimeModel
3
+
4
+ __all__ = [
5
+ "RealtimeModel",
6
+ "ClientEvents",
7
+ "LiveAPIModels",
8
+ "Voice",
9
+ ]
livekit/plugins/google/realtime/api_proto.py ADDED
@@ -0,0 +1,66 @@
1
+ from __future__ import annotations
2
+
3
+ from collections.abc import Sequence
4
+ from typing import Literal, Union
5
+
6
+ from google.genai import types
7
+
8
+ LiveAPIModels = Literal[
9
+ # VertexAI models
10
+ "gemini-live-2.5-flash-native-audio",
11
+ "gemini-live-2.5-flash-preview-native-audio",
12
+ # deprecated vertexai models
13
+ "gemini-2.0-flash-exp",
14
+ "gemini-live-2.5-flash-preview-native-audio-09-2025",
15
+ # Gemini API models
16
+ "gemini-2.5-flash-native-audio-preview-12-2025",
17
+ "gemini-live-2.5-flash-preview",
18
+ # deprecated Gemini API models
19
+ "gemini-2.0-flash-live-001",
20
+ "gemini-2.5-flash-native-audio-preview-09-2025",
21
+ ]
22
+
23
+ Voice = Literal[
24
+ "Achernar",
25
+ "Achird",
26
+ "Algenib",
27
+ "Algieba",
28
+ "Alnilam",
29
+ "Aoede",
30
+ "Autonoe",
31
+ "Callirrhoe",
32
+ "Charon",
33
+ "Despina",
34
+ "Enceladus",
35
+ "Erinome",
36
+ "Fenrir",
37
+ "Gacrux",
38
+ "Iapetus",
39
+ "Kore",
40
+ "Laomedeia",
41
+ "Leda",
42
+ "Orus",
43
+ "Pulcherrima",
44
+ "Puck",
45
+ "Rasalgethi",
46
+ "Sadachbia",
47
+ "Sadaltager",
48
+ "Schedar",
49
+ "Sulafat",
50
+ "Umbriel",
51
+ "Vindemiatrix",
52
+ "Zephyr",
53
+ "Zubenelgenubi",
54
+ ]
55
+
56
+
57
+ ClientEvents = Union[
58
+ types.ContentListUnion,
59
+ types.ContentListUnionDict,
60
+ types.LiveClientContentOrDict,
61
+ types.LiveClientRealtimeInput,
62
+ types.LiveClientRealtimeInputOrDict,
63
+ types.LiveClientToolResponseOrDict,
64
+ types.FunctionResponseOrDict,
65
+ Sequence[types.FunctionResponseOrDict],
66
+ ]