livekit-plugins-google 0.3.0__py3-none-any.whl → 1.3.11__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 +33 -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 +562 -0
  5. livekit/plugins/google/log.py +3 -0
  6. livekit/plugins/google/models.py +160 -32
  7. livekit/plugins/google/realtime/__init__.py +9 -0
  8. livekit/plugins/google/realtime/api_proto.py +68 -0
  9. livekit/plugins/google/realtime/realtime_api.py +1249 -0
  10. livekit/plugins/google/stt.py +717 -283
  11. livekit/plugins/google/tools.py +71 -0
  12. livekit/plugins/google/tts.py +455 -0
  13. livekit/plugins/google/utils.py +220 -0
  14. livekit/plugins/google/version.py +1 -1
  15. livekit_plugins_google-1.3.11.dist-info/METADATA +63 -0
  16. livekit_plugins_google-1.3.11.dist-info/RECORD +18 -0
  17. {livekit_plugins_google-0.3.0.dist-info → livekit_plugins_google-1.3.11.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
@@ -1,85 +1,213 @@
1
1
  from typing import Literal
2
2
 
3
- # Speech to Text v2
3
+ # Speech to Text (v1 and 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",
16
+ ]
17
+ # https://docs.cloud.google.com/speech-to-text/docs/transcription-model
18
+
19
+ SpeechModelsV2 = Literal[
20
+ "telephony",
21
+ "chirp_2",
22
+ "chirp_3",
7
23
  ]
8
24
 
9
25
  SpeechLanguages = Literal[
10
- "en-US",
11
- "ja-JP",
12
- "en-IN",
13
- "en-GB",
14
- "hi-IN",
15
26
  "af-ZA",
16
- "sq-AL",
17
27
  "am-ET",
28
+ "ar-AE",
29
+ "ar-BH",
30
+ "ar-DZ",
18
31
  "ar-EG",
19
- "hy-AM",
32
+ "ar-IL",
33
+ "ar-IQ",
34
+ "ar-JO",
35
+ "ar-KW",
36
+ "ar-LB",
37
+ "ar-MA",
38
+ "ar-MR",
39
+ "ar-OM",
40
+ "ar-PS",
41
+ "ar-QA",
42
+ "ar-SA",
43
+ "ar-TN",
44
+ "ar-YE",
45
+ "as-IN",
20
46
  "ast-ES",
21
47
  "az-AZ",
22
- "eu-ES",
23
48
  "be-BY",
24
- "bs-BA",
25
49
  "bg-BG",
26
- "my-MM",
50
+ "bn-BD",
51
+ "bn-IN",
52
+ "bs-BA",
27
53
  "ca-ES",
28
54
  "ceb-PH",
29
55
  "ckb-IQ",
30
- "zh-Hans-CN",
31
- "yue-Hant-HK",
32
- "zh-TW",
33
- "hr-HR",
56
+ "cmn-Hans-CN",
57
+ "cmn-Hant-TW",
34
58
  "cs-CZ",
59
+ "cy-GB",
35
60
  "da-DK",
36
- "nl-NL",
61
+ "de-AT",
62
+ "de-CH",
63
+ "de-DE",
64
+ "el-GR",
37
65
  "en-AU",
66
+ "en-CA",
67
+ "en-GB",
68
+ "en-HK",
69
+ "en-IE",
70
+ "en-IN",
71
+ "en-NZ",
72
+ "en-PK",
73
+ "en-SG",
74
+ "en-US",
75
+ "es-419",
76
+ "es-AR",
77
+ "es-BO",
78
+ "es-CL",
79
+ "es-CO",
80
+ "es-CR",
81
+ "es-DO",
82
+ "es-EC",
83
+ "es-ES",
84
+ "es-GT",
85
+ "es-HN",
86
+ "es-MX",
87
+ "es-NI",
88
+ "es-PA",
89
+ "es-PE",
90
+ "es-PR",
91
+ "es-SV",
92
+ "es-US",
93
+ "es-UY",
94
+ "es-VE",
38
95
  "et-EE",
39
- "fil-PH",
96
+ "eu-ES",
97
+ "fa-IR",
98
+ "ff-SN",
40
99
  "fi-FI",
100
+ "fil-PH",
101
+ "fr-BE",
41
102
  "fr-CA",
103
+ "fr-CH",
42
104
  "fr-FR",
105
+ "ga-IE",
43
106
  "gl-ES",
44
- "ka-GE",
45
- "de-DE",
46
- "el-GR",
47
107
  "gu-IN",
48
108
  "ha-NG",
49
- "iw-IL",
50
109
  "hi-IN",
110
+ "hr-HR",
51
111
  "hu-HU",
52
- "is-IS",
112
+ "hy-AM",
53
113
  "id-ID",
114
+ "ig-NG",
115
+ "is-IS",
116
+ "it-CH",
54
117
  "it-IT",
118
+ "iw-IL",
55
119
  "ja-JP",
56
120
  "jv-ID",
57
- "kea-CV",
121
+ "ka-GE",
58
122
  "kam-KE",
59
- "kn-IN",
123
+ "kea-CV",
60
124
  "kk-KZ",
61
125
  "km-KH",
126
+ "kn-IN",
62
127
  "ko-KR",
63
128
  "ky-KG",
64
- "lo-LA",
65
- "lv-LV",
129
+ "lb-LU",
130
+ "lg-UG",
66
131
  "ln-CD",
132
+ "lo-LA",
67
133
  "lt-LT",
68
134
  "luo-KE",
69
- "lb-LU",
135
+ "lv-LV",
136
+ "mi-NZ",
70
137
  "mk-MK",
138
+ "ml-IN",
139
+ "mn-MN",
140
+ "mr-IN",
141
+ "ms-MY",
142
+ "mt-MT",
143
+ "my-MM",
144
+ "ne-NP",
145
+ "nl-BE",
146
+ "nl-NL",
71
147
  "no-NO",
148
+ "nso-ZA",
149
+ "ny-MW",
150
+ "oc-FR",
151
+ "om-ET",
152
+ "or-IN",
153
+ "pa-Guru-IN",
72
154
  "pl-PL",
155
+ "ps-AF",
73
156
  "pt-BR",
74
157
  "pt-PT",
75
158
  "ro-RO",
76
159
  "ru-RU",
77
- "es-CO",
78
- "es-MX",
79
- "es-US",
160
+ "rup-BG",
161
+ "rw-RW",
162
+ "sd-IN",
163
+ "si-LK",
164
+ "sk-SK",
165
+ "sl-SI",
166
+ "sn-ZW",
167
+ "so-SO",
168
+ "sq-AL",
169
+ "sr-RS",
170
+ "ss-Latn-ZA",
171
+ "st-ZA",
172
+ "su-ID",
173
+ "sv-SE",
174
+ "sw",
175
+ "sw-KE",
176
+ "ta-IN",
177
+ "te-IN",
178
+ "tg-TJ",
80
179
  "th-TH",
180
+ "tn-Latn-ZA",
81
181
  "tr-TR",
182
+ "ts-ZA",
82
183
  "uk-UA",
184
+ "umb-AO",
185
+ "ur-PK",
186
+ "uz-UZ",
187
+ "ve-ZA",
83
188
  "vi-VN",
84
- "da-DK",
189
+ "wo-SN",
190
+ "xh-ZA",
191
+ "yo-NG",
192
+ "yue-Hant-HK",
193
+ "zu-ZA",
194
+ ]
195
+
196
+ Gender = Literal["male", "female", "neutral"]
197
+
198
+ ChatModels = Literal[
199
+ "gemini-3-pro-preview",
200
+ "gemini-3-flash-preview",
201
+ "gemini-2.5-flash",
202
+ "gemini-2.5-pro-preview-05-06",
203
+ "gemini-2.5-flash-preview-04-17",
204
+ "gemini-2.5-flash-preview-05-20",
205
+ "gemini-2.0-flash-001",
206
+ "gemini-2.0-flash-lite-preview-02-05",
207
+ "gemini-2.0-pro-exp-02-05",
208
+ "gemini-1.5-pro",
209
+ ]
210
+
211
+ GeminiTTSModels = Literal[
212
+ "gemini-2.5-flash-tts", "gemini-2.5-flash-lite-preview-tts", "gemini-2.5-pro-tts"
85
213
  ]
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 +1,68 @@
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
+ # Gemini API deprecations: https://ai.google.dev/gemini-api/docs/deprecations
9
+ # Gemini API release notes with preview deprecations: https://ai.google.dev/gemini-api/docs/changelog
10
+ # live models: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/live-api
11
+ # VertexAI retirement: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/learn/model-versions#retired-models
12
+ # Additional references:
13
+ # 1. https://github.com/kazunori279/adk-streaming-test/blob/main/test_report.md
14
+ LiveAPIModels = Literal[
15
+ # VertexAI models
16
+ "gemini-live-2.5-flash-native-audio", # GA https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/2-5-flash-live-api#live-2.5-flash
17
+ "gemini-live-2.5-flash-preview-native-audio-09-2025", # Public preview https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/2-5-flash-live-api#live-2.5-flash-preview
18
+ "gemini-live-2.5-flash-preview-native-audio", # still works, possibly an alias, but not mentioned in any docs or changelog
19
+ # Gemini API models
20
+ "gemini-2.5-flash-native-audio-preview-12-2025", # https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-live
21
+ "gemini-2.5-flash-native-audio-preview-09-2025", # https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-live
22
+ "gemini-2.0-flash-exp", # still works in Gemini API but not VertexAI
23
+ ]
24
+
25
+ Voice = Literal[
26
+ "Achernar",
27
+ "Achird",
28
+ "Algenib",
29
+ "Algieba",
30
+ "Alnilam",
31
+ "Aoede",
32
+ "Autonoe",
33
+ "Callirrhoe",
34
+ "Charon",
35
+ "Despina",
36
+ "Enceladus",
37
+ "Erinome",
38
+ "Fenrir",
39
+ "Gacrux",
40
+ "Iapetus",
41
+ "Kore",
42
+ "Laomedeia",
43
+ "Leda",
44
+ "Orus",
45
+ "Pulcherrima",
46
+ "Puck",
47
+ "Rasalgethi",
48
+ "Sadachbia",
49
+ "Sadaltager",
50
+ "Schedar",
51
+ "Sulafat",
52
+ "Umbriel",
53
+ "Vindemiatrix",
54
+ "Zephyr",
55
+ "Zubenelgenubi",
56
+ ]
57
+
58
+
59
+ ClientEvents = Union[
60
+ types.ContentListUnion,
61
+ types.ContentListUnionDict,
62
+ types.LiveClientContentOrDict,
63
+ types.LiveClientRealtimeInput,
64
+ types.LiveClientRealtimeInputOrDict,
65
+ types.LiveClientToolResponseOrDict,
66
+ types.FunctionResponseOrDict,
67
+ Sequence[types.FunctionResponseOrDict],
68
+ ]