@tanstack/ai-grok 0.6.8 → 0.7.1

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 (44) hide show
  1. package/dist/esm/adapters/image.js +15 -8
  2. package/dist/esm/adapters/image.js.map +1 -1
  3. package/dist/esm/adapters/transcription.d.ts +84 -0
  4. package/dist/esm/adapters/transcription.js +109 -0
  5. package/dist/esm/adapters/transcription.js.map +1 -0
  6. package/dist/esm/adapters/tts.d.ts +70 -0
  7. package/dist/esm/adapters/tts.js +137 -0
  8. package/dist/esm/adapters/tts.js.map +1 -0
  9. package/dist/esm/audio/transcription-provider-options.d.ts +41 -0
  10. package/dist/esm/audio/tts-provider-options.d.ts +42 -0
  11. package/dist/esm/index.d.ts +8 -2
  12. package/dist/esm/index.js +17 -2
  13. package/dist/esm/index.js.map +1 -1
  14. package/dist/esm/model-meta.d.ts +6 -0
  15. package/dist/esm/model-meta.js +22 -1
  16. package/dist/esm/model-meta.js.map +1 -1
  17. package/dist/esm/realtime/adapter.d.ts +21 -0
  18. package/dist/esm/realtime/adapter.js +816 -0
  19. package/dist/esm/realtime/adapter.js.map +1 -0
  20. package/dist/esm/realtime/index.d.ts +4 -0
  21. package/dist/esm/realtime/realtime-contract.d.ts +30 -0
  22. package/dist/esm/realtime/token.d.ts +22 -0
  23. package/dist/esm/realtime/token.js +73 -0
  24. package/dist/esm/realtime/token.js.map +1 -0
  25. package/dist/esm/realtime/types.d.ts +95 -0
  26. package/dist/esm/utils/audio.d.ts +23 -0
  27. package/dist/esm/utils/audio.js +171 -0
  28. package/dist/esm/utils/audio.js.map +1 -0
  29. package/dist/esm/utils/index.d.ts +1 -0
  30. package/package.json +6 -3
  31. package/src/adapters/image.ts +16 -7
  32. package/src/adapters/transcription.ts +233 -0
  33. package/src/adapters/tts.ts +260 -0
  34. package/src/audio/transcription-provider-options.ts +54 -0
  35. package/src/audio/tts-provider-options.ts +44 -0
  36. package/src/index.ts +50 -1
  37. package/src/model-meta.ts +54 -0
  38. package/src/realtime/adapter.ts +1215 -0
  39. package/src/realtime/index.ts +18 -0
  40. package/src/realtime/realtime-contract.ts +46 -0
  41. package/src/realtime/token.ts +131 -0
  42. package/src/realtime/types.ts +105 -0
  43. package/src/utils/audio.ts +217 -0
  44. package/src/utils/index.ts +1 -0
package/dist/esm/index.js CHANGED
@@ -1,18 +1,33 @@
1
1
  import { GrokTextAdapter, createGrokText, grokText } from "./adapters/text.js";
2
2
  import { GrokSummarizeAdapter, createGrokSummarize, grokSummarize } from "./adapters/summarize.js";
3
3
  import { GrokImageAdapter, createGrokImage, grokImage } from "./adapters/image.js";
4
- import { GROK_CHAT_MODELS, GROK_IMAGE_MODELS } from "./model-meta.js";
4
+ import { GrokSpeechAdapter, createGrokSpeech, grokSpeech } from "./adapters/tts.js";
5
+ import { GrokTranscriptionAdapter, createGrokTranscription, grokTranscription } from "./adapters/transcription.js";
6
+ import { GROK_CHAT_MODELS, GROK_IMAGE_MODELS, GROK_REALTIME_MODELS, GROK_TRANSCRIPTION_MODELS, GROK_TTS_MODELS } from "./model-meta.js";
7
+ import { grokRealtimeToken } from "./realtime/token.js";
8
+ import { grokRealtime } from "./realtime/adapter.js";
5
9
  export {
6
10
  GROK_CHAT_MODELS,
7
11
  GROK_IMAGE_MODELS,
12
+ GROK_REALTIME_MODELS,
13
+ GROK_TRANSCRIPTION_MODELS,
14
+ GROK_TTS_MODELS,
8
15
  GrokImageAdapter,
16
+ GrokSpeechAdapter,
9
17
  GrokSummarizeAdapter,
10
18
  GrokTextAdapter,
19
+ GrokTranscriptionAdapter,
11
20
  createGrokImage,
21
+ createGrokSpeech,
12
22
  createGrokSummarize,
13
23
  createGrokText,
24
+ createGrokTranscription,
14
25
  grokImage,
26
+ grokRealtime,
27
+ grokRealtimeToken,
28
+ grokSpeech,
15
29
  grokSummarize,
16
- grokText
30
+ grokText,
31
+ grokTranscription
17
32
  };
18
33
  //# sourceMappingURL=index.js.map
@@ -1 +1 @@
1
- {"version":3,"file":"index.js","sources":[],"sourcesContent":[],"names":[],"mappings":";;;;"}
1
+ {"version":3,"file":"index.js","sources":[],"sourcesContent":[],"names":[],"mappings":";;;;;;;;"}
@@ -215,8 +215,14 @@ export declare const GROK_CHAT_MODELS: readonly ["grok-4-1-fast-reasoning", "gro
215
215
  * Grok Image Generation Models
216
216
  */
217
217
  export declare const GROK_IMAGE_MODELS: readonly ["grok-2-image-1212"];
218
+ export declare const GROK_TTS_MODELS: readonly ["grok-tts"];
219
+ export declare const GROK_TRANSCRIPTION_MODELS: readonly ["grok-stt"];
220
+ export declare const GROK_REALTIME_MODELS: readonly ["grok-voice-fast-1.0", "grok-voice-think-fast-1.0"];
218
221
  export type GrokChatModel = (typeof GROK_CHAT_MODELS)[number];
219
222
  export type GrokImageModel = (typeof GROK_IMAGE_MODELS)[number];
223
+ export type GrokTTSModel = (typeof GROK_TTS_MODELS)[number];
224
+ export type GrokTranscriptionModel = (typeof GROK_TRANSCRIPTION_MODELS)[number];
225
+ export type GrokRealtimeModel = (typeof GROK_REALTIME_MODELS)[number];
220
226
  /**
221
227
  * Type-only map from Grok chat model name to its supported input modalities.
222
228
  * Used for type inference when constructing multimodal messages.
@@ -48,8 +48,29 @@ const GROK_CHAT_MODELS = [
48
48
  GROK_4_20_MULTI_AGENT.name
49
49
  ];
50
50
  const GROK_IMAGE_MODELS = [GROK_2_IMAGE.name];
51
+ const GROK_TTS = {
52
+ name: "grok-tts"
53
+ };
54
+ const GROK_STT = {
55
+ name: "grok-stt"
56
+ };
57
+ const GROK_VOICE_FAST_1 = {
58
+ name: "grok-voice-fast-1.0"
59
+ };
60
+ const GROK_VOICE_THINK_FAST_1 = {
61
+ name: "grok-voice-think-fast-1.0"
62
+ };
63
+ const GROK_TTS_MODELS = [GROK_TTS.name];
64
+ const GROK_TRANSCRIPTION_MODELS = [GROK_STT.name];
65
+ const GROK_REALTIME_MODELS = [
66
+ GROK_VOICE_FAST_1.name,
67
+ GROK_VOICE_THINK_FAST_1.name
68
+ ];
51
69
  export {
52
70
  GROK_CHAT_MODELS,
53
- GROK_IMAGE_MODELS
71
+ GROK_IMAGE_MODELS,
72
+ GROK_REALTIME_MODELS,
73
+ GROK_TRANSCRIPTION_MODELS,
74
+ GROK_TTS_MODELS
54
75
  };
55
76
  //# sourceMappingURL=model-meta.js.map
@@ -1 +1 @@
1
- {"version":3,"file":"model-meta.js","sources":["../../src/model-meta.ts"],"sourcesContent":["/**\n * Model metadata interface for documentation and type inference\n */\ninterface ModelMeta {\n name: string\n supports: {\n input: Array<'text' | 'image' | 'audio' | 'video' | 'document'>\n output: Array<'text' | 'image' | 'audio' | 'video'>\n capabilities?: Array<'reasoning' | 'tool_calling' | 'structured_outputs'>\n tools?: ReadonlyArray<never>\n }\n max_input_tokens?: number\n max_output_tokens?: number\n context_window?: number\n knowledge_cutoff?: string\n pricing?: {\n input: {\n normal: number\n cached?: number\n }\n output: {\n normal: number\n }\n }\n}\n\nconst GROK_4_1_FAST_REASONING = {\n name: 'grok-4-1-fast-reasoning',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.05,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4_1_FAST_NON_REASONING = {\n name: 'grok-4-1-fast-non-reasoning',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.05,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_CODE_FAST_1 = {\n name: 'grok-code-fast-1',\n context_window: 256_000,\n supports: {\n input: ['text'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.02,\n },\n output: {\n normal: 1.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4_FAST_REASONING = {\n name: 'grok-4-fast-reasoning',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.05,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4_FAST_NON_REASONING = {\n name: 'grok-4-fast-non-reasoning',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.05,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4 = {\n name: 'grok-4',\n context_window: 256_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 3,\n cached: 0.75,\n },\n output: {\n normal: 15,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_3_MINI = {\n name: 'grok-3-mini',\n context_window: 131_072,\n supports: {\n input: ['text'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.3,\n cached: 0.075,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_3 = {\n name: 'grok-3',\n context_window: 131_072,\n supports: {\n input: ['text'],\n output: ['text'],\n capabilities: ['structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 3,\n cached: 0.75,\n },\n output: {\n normal: 15,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_2_VISION = {\n name: 'grok-2-vision-1212',\n context_window: 32_768,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 2,\n },\n output: {\n normal: 10,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_2_IMAGE = {\n name: 'grok-2-image-1212',\n supports: {\n input: ['text'],\n output: ['image'],\n },\n pricing: {\n input: {\n normal: 0.07,\n },\n output: {\n normal: 0.07,\n },\n },\n} as const satisfies ModelMeta\n\n/**\n * Grok Chat Models\n * Based on xAI's available models as of 2025\n */\nconst GROK_4_20 = {\n name: 'grok-4.20',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image', 'document'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 2,\n cached: 0.2,\n },\n output: {\n normal: 6,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4_20_MULTI_AGENT = {\n name: 'grok-4.20-multi-agent',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image', 'document'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 2,\n cached: 0.2,\n },\n output: {\n normal: 6,\n },\n },\n} as const satisfies ModelMeta\n\nexport const GROK_CHAT_MODELS = [\n GROK_4_1_FAST_REASONING.name,\n GROK_4_1_FAST_NON_REASONING.name,\n GROK_CODE_FAST_1.name,\n GROK_4_FAST_REASONING.name,\n GROK_4_FAST_NON_REASONING.name,\n GROK_4.name,\n GROK_3.name,\n GROK_3_MINI.name,\n GROK_2_VISION.name,\n\n GROK_4_20.name,\n GROK_4_20_MULTI_AGENT.name,\n] as const\n\n/**\n * Grok Image Generation Models\n */\nexport const GROK_IMAGE_MODELS = [GROK_2_IMAGE.name] as const\n\nexport type GrokChatModel = (typeof GROK_CHAT_MODELS)[number]\nexport type GrokImageModel = (typeof GROK_IMAGE_MODELS)[number]\n\n/**\n * Type-only map from Grok chat model name to its supported input modalities.\n * Used for type inference when constructing multimodal messages.\n */\nexport type GrokModelInputModalitiesByName = {\n [GROK_4_1_FAST_REASONING.name]: typeof GROK_4_1_FAST_REASONING.supports.input\n [GROK_4_1_FAST_NON_REASONING.name]: typeof GROK_4_1_FAST_NON_REASONING.supports.input\n [GROK_CODE_FAST_1.name]: typeof GROK_CODE_FAST_1.supports.input\n [GROK_4_FAST_REASONING.name]: typeof GROK_4_FAST_REASONING.supports.input\n [GROK_4_FAST_NON_REASONING.name]: typeof GROK_4_FAST_NON_REASONING.supports.input\n [GROK_4.name]: typeof GROK_4.supports.input\n [GROK_3.name]: typeof GROK_3.supports.input\n [GROK_3_MINI.name]: typeof GROK_3_MINI.supports.input\n [GROK_2_VISION.name]: typeof GROK_2_VISION.supports.input\n [GROK_4_20.name]: typeof GROK_4_20.supports.input\n [GROK_4_20_MULTI_AGENT.name]: typeof GROK_4_20_MULTI_AGENT.supports.input\n}\n\n/**\n * Type-only map from Grok chat model name to its provider options type.\n * Since Grok uses OpenAI-compatible API, we reuse OpenAI provider options.\n */\nexport type GrokChatModelProviderOptionsByName = {\n [K in (typeof GROK_CHAT_MODELS)[number]]: GrokProviderOptions\n}\n\n/**\n * Type-only map from Grok chat model name to its supported provider tools.\n * Grok exposes no provider-specific tool factories, so every model gets an\n * empty tuple. This ensures that passing an Anthropic/OpenAI ProviderTool to\n * a Grok adapter produces a compile-time type error.\n */\nexport type GrokChatModelToolCapabilitiesByName = {\n [GROK_4_1_FAST_REASONING.name]: typeof GROK_4_1_FAST_REASONING.supports.tools\n [GROK_4_1_FAST_NON_REASONING.name]: typeof GROK_4_1_FAST_NON_REASONING.supports.tools\n [GROK_CODE_FAST_1.name]: typeof GROK_CODE_FAST_1.supports.tools\n [GROK_4_FAST_REASONING.name]: typeof GROK_4_FAST_REASONING.supports.tools\n [GROK_4_FAST_NON_REASONING.name]: typeof GROK_4_FAST_NON_REASONING.supports.tools\n [GROK_4.name]: typeof GROK_4.supports.tools\n [GROK_3.name]: typeof GROK_3.supports.tools\n [GROK_3_MINI.name]: typeof GROK_3_MINI.supports.tools\n [GROK_2_VISION.name]: typeof GROK_2_VISION.supports.tools\n [GROK_4_20.name]: typeof GROK_4_20.supports.tools\n [GROK_4_20_MULTI_AGENT.name]: typeof GROK_4_20_MULTI_AGENT.supports.tools\n}\n\n/**\n * Grok-specific provider options\n * Based on OpenAI-compatible API options\n */\nexport interface GrokProviderOptions {\n /** Temperature for response generation (0-2) */\n temperature?: number\n /** Maximum tokens in the response */\n max_tokens?: number\n /** Top-p sampling parameter */\n top_p?: number\n /** Frequency penalty (-2.0 to 2.0) */\n frequency_penalty?: number\n /** Presence penalty (-2.0 to 2.0) */\n presence_penalty?: number\n /** Stop sequences */\n stop?: string | Array<string>\n /** A unique identifier representing your end-user */\n user?: string\n}\n\n// ===========================\n// Type Resolution Helpers\n// ===========================\n\n/**\n * Resolve provider options for a specific model.\n * If the model has explicit options in the map, use those; otherwise use base options.\n */\nexport type ResolveProviderOptions<TModel extends string> =\n TModel extends keyof GrokChatModelProviderOptionsByName\n ? GrokChatModelProviderOptionsByName[TModel]\n : GrokProviderOptions\n\n/**\n * Resolve input modalities for a specific model.\n * If the model has explicit modalities in the map, use those; otherwise use text only.\n */\nexport type ResolveInputModalities<TModel extends string> =\n TModel extends keyof GrokModelInputModalitiesByName\n ? GrokModelInputModalitiesByName[TModel]\n : readonly ['text']\n"],"names":[],"mappings":"AA0BA,MAAM,0BAA0B;AAAA,EAC9B,MAAM;AAiBR;AAEA,MAAM,8BAA8B;AAAA,EAClC,MAAM;AAiBR;AAEA,MAAM,mBAAmB;AAAA,EACvB,MAAM;AAiBR;AAEA,MAAM,wBAAwB;AAAA,EAC5B,MAAM;AAiBR;AAEA,MAAM,4BAA4B;AAAA,EAChC,MAAM;AAiBR;AAEA,MAAM,SAAS;AAAA,EACb,MAAM;AAiBR;AAEA,MAAM,cAAc;AAAA,EAClB,MAAM;AAiBR;AAEA,MAAM,SAAS;AAAA,EACb,MAAM;AAiBR;AAEA,MAAM,gBAAgB;AAAA,EACpB,MAAM;AAgBR;AAEA,MAAM,eAAe;AAAA,EACnB,MAAM;AAaR;AAMA,MAAM,YAAY;AAAA,EAChB,MAAM;AAiBR;AAEA,MAAM,wBAAwB;AAAA,EAC5B,MAAM;AAiBR;AAEO,MAAM,mBAAmB;AAAA,EAC9B,wBAAwB;AAAA,EACxB,4BAA4B;AAAA,EAC5B,iBAAiB;AAAA,EACjB,sBAAsB;AAAA,EACtB,0BAA0B;AAAA,EAC1B,OAAO;AAAA,EACP,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,cAAc;AAAA,EAEd,UAAU;AAAA,EACV,sBAAsB;AACxB;AAKO,MAAM,oBAAoB,CAAC,aAAa,IAAI;"}
1
+ {"version":3,"file":"model-meta.js","sources":["../../src/model-meta.ts"],"sourcesContent":["/**\n * Model metadata interface for documentation and type inference\n */\ninterface ModelMeta {\n name: string\n supports: {\n input: Array<'text' | 'image' | 'audio' | 'video' | 'document'>\n output: Array<'text' | 'image' | 'audio' | 'video'>\n capabilities?: Array<'reasoning' | 'tool_calling' | 'structured_outputs'>\n tools?: ReadonlyArray<never>\n }\n max_input_tokens?: number\n max_output_tokens?: number\n context_window?: number\n knowledge_cutoff?: string\n pricing?: {\n input: {\n normal: number\n cached?: number\n }\n output: {\n normal: number\n }\n }\n}\n\nconst GROK_4_1_FAST_REASONING = {\n name: 'grok-4-1-fast-reasoning',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.05,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4_1_FAST_NON_REASONING = {\n name: 'grok-4-1-fast-non-reasoning',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.05,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_CODE_FAST_1 = {\n name: 'grok-code-fast-1',\n context_window: 256_000,\n supports: {\n input: ['text'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.02,\n },\n output: {\n normal: 1.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4_FAST_REASONING = {\n name: 'grok-4-fast-reasoning',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.05,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4_FAST_NON_REASONING = {\n name: 'grok-4-fast-non-reasoning',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.2,\n cached: 0.05,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4 = {\n name: 'grok-4',\n context_window: 256_000,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 3,\n cached: 0.75,\n },\n output: {\n normal: 15,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_3_MINI = {\n name: 'grok-3-mini',\n context_window: 131_072,\n supports: {\n input: ['text'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 0.3,\n cached: 0.075,\n },\n output: {\n normal: 0.5,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_3 = {\n name: 'grok-3',\n context_window: 131_072,\n supports: {\n input: ['text'],\n output: ['text'],\n capabilities: ['structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 3,\n cached: 0.75,\n },\n output: {\n normal: 15,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_2_VISION = {\n name: 'grok-2-vision-1212',\n context_window: 32_768,\n supports: {\n input: ['text', 'image'],\n output: ['text'],\n capabilities: ['structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 2,\n },\n output: {\n normal: 10,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_2_IMAGE = {\n name: 'grok-2-image-1212',\n supports: {\n input: ['text'],\n output: ['image'],\n },\n pricing: {\n input: {\n normal: 0.07,\n },\n output: {\n normal: 0.07,\n },\n },\n} as const satisfies ModelMeta\n\n/**\n * Grok Chat Models\n * Based on xAI's available models as of 2025\n */\nconst GROK_4_20 = {\n name: 'grok-4.20',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image', 'document'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 2,\n cached: 0.2,\n },\n output: {\n normal: 6,\n },\n },\n} as const satisfies ModelMeta\n\nconst GROK_4_20_MULTI_AGENT = {\n name: 'grok-4.20-multi-agent',\n context_window: 2_000_000,\n supports: {\n input: ['text', 'image', 'document'],\n output: ['text'],\n capabilities: ['reasoning', 'structured_outputs', 'tool_calling'],\n tools: [] as const,\n },\n pricing: {\n input: {\n normal: 2,\n cached: 0.2,\n },\n output: {\n normal: 6,\n },\n },\n} as const satisfies ModelMeta\n\nexport const GROK_CHAT_MODELS = [\n GROK_4_1_FAST_REASONING.name,\n GROK_4_1_FAST_NON_REASONING.name,\n GROK_CODE_FAST_1.name,\n GROK_4_FAST_REASONING.name,\n GROK_4_FAST_NON_REASONING.name,\n GROK_4.name,\n GROK_3.name,\n GROK_3_MINI.name,\n GROK_2_VISION.name,\n\n GROK_4_20.name,\n GROK_4_20_MULTI_AGENT.name,\n] as const\n\n/**\n * Grok Image Generation Models\n */\nexport const GROK_IMAGE_MODELS = [GROK_2_IMAGE.name] as const\n\n// xAI's `/v1/tts` endpoint is endpoint-addressed and does not take a `model`\n// parameter. This synthetic identifier satisfies the SDK's `TTSOptions.model`\n// contract and provides a stable value for logging and fixture matching.\nconst GROK_TTS = {\n name: 'grok-tts',\n supports: {\n input: ['text'],\n output: ['audio'],\n },\n} as const satisfies ModelMeta\n\n// xAI's `/v1/stt` endpoint is endpoint-addressed and does not take a `model`\n// parameter. This synthetic identifier satisfies the SDK's\n// `TranscriptionOptions.model` contract.\nconst GROK_STT = {\n name: 'grok-stt',\n supports: {\n input: ['audio'],\n output: ['text'],\n },\n} as const satisfies ModelMeta\n\nconst GROK_VOICE_FAST_1 = {\n name: 'grok-voice-fast-1.0',\n supports: {\n input: ['audio', 'text'],\n output: ['audio', 'text'],\n capabilities: ['tool_calling'],\n tools: [] as const,\n },\n} as const satisfies ModelMeta\n\nconst GROK_VOICE_THINK_FAST_1 = {\n name: 'grok-voice-think-fast-1.0',\n supports: {\n input: ['audio', 'text'],\n output: ['audio', 'text'],\n capabilities: ['reasoning', 'tool_calling'],\n tools: [] as const,\n },\n} as const satisfies ModelMeta\n\nexport const GROK_TTS_MODELS = [GROK_TTS.name] as const\n\nexport const GROK_TRANSCRIPTION_MODELS = [GROK_STT.name] as const\n\nexport const GROK_REALTIME_MODELS = [\n GROK_VOICE_FAST_1.name,\n GROK_VOICE_THINK_FAST_1.name,\n] as const\n\nexport type GrokChatModel = (typeof GROK_CHAT_MODELS)[number]\nexport type GrokImageModel = (typeof GROK_IMAGE_MODELS)[number]\nexport type GrokTTSModel = (typeof GROK_TTS_MODELS)[number]\nexport type GrokTranscriptionModel = (typeof GROK_TRANSCRIPTION_MODELS)[number]\nexport type GrokRealtimeModel = (typeof GROK_REALTIME_MODELS)[number]\n\n/**\n * Type-only map from Grok chat model name to its supported input modalities.\n * Used for type inference when constructing multimodal messages.\n */\nexport type GrokModelInputModalitiesByName = {\n [GROK_4_1_FAST_REASONING.name]: typeof GROK_4_1_FAST_REASONING.supports.input\n [GROK_4_1_FAST_NON_REASONING.name]: typeof GROK_4_1_FAST_NON_REASONING.supports.input\n [GROK_CODE_FAST_1.name]: typeof GROK_CODE_FAST_1.supports.input\n [GROK_4_FAST_REASONING.name]: typeof GROK_4_FAST_REASONING.supports.input\n [GROK_4_FAST_NON_REASONING.name]: typeof GROK_4_FAST_NON_REASONING.supports.input\n [GROK_4.name]: typeof GROK_4.supports.input\n [GROK_3.name]: typeof GROK_3.supports.input\n [GROK_3_MINI.name]: typeof GROK_3_MINI.supports.input\n [GROK_2_VISION.name]: typeof GROK_2_VISION.supports.input\n [GROK_4_20.name]: typeof GROK_4_20.supports.input\n [GROK_4_20_MULTI_AGENT.name]: typeof GROK_4_20_MULTI_AGENT.supports.input\n}\n\n/**\n * Type-only map from Grok chat model name to its provider options type.\n * Since Grok uses OpenAI-compatible API, we reuse OpenAI provider options.\n */\nexport type GrokChatModelProviderOptionsByName = {\n [K in (typeof GROK_CHAT_MODELS)[number]]: GrokProviderOptions\n}\n\n/**\n * Type-only map from Grok chat model name to its supported provider tools.\n * Grok exposes no provider-specific tool factories, so every model gets an\n * empty tuple. This ensures that passing an Anthropic/OpenAI ProviderTool to\n * a Grok adapter produces a compile-time type error.\n */\nexport type GrokChatModelToolCapabilitiesByName = {\n [GROK_4_1_FAST_REASONING.name]: typeof GROK_4_1_FAST_REASONING.supports.tools\n [GROK_4_1_FAST_NON_REASONING.name]: typeof GROK_4_1_FAST_NON_REASONING.supports.tools\n [GROK_CODE_FAST_1.name]: typeof GROK_CODE_FAST_1.supports.tools\n [GROK_4_FAST_REASONING.name]: typeof GROK_4_FAST_REASONING.supports.tools\n [GROK_4_FAST_NON_REASONING.name]: typeof GROK_4_FAST_NON_REASONING.supports.tools\n [GROK_4.name]: typeof GROK_4.supports.tools\n [GROK_3.name]: typeof GROK_3.supports.tools\n [GROK_3_MINI.name]: typeof GROK_3_MINI.supports.tools\n [GROK_2_VISION.name]: typeof GROK_2_VISION.supports.tools\n [GROK_4_20.name]: typeof GROK_4_20.supports.tools\n [GROK_4_20_MULTI_AGENT.name]: typeof GROK_4_20_MULTI_AGENT.supports.tools\n}\n\n/**\n * Grok-specific provider options\n * Based on OpenAI-compatible API options\n */\nexport interface GrokProviderOptions {\n /** Temperature for response generation (0-2) */\n temperature?: number\n /** Maximum tokens in the response */\n max_tokens?: number\n /** Top-p sampling parameter */\n top_p?: number\n /** Frequency penalty (-2.0 to 2.0) */\n frequency_penalty?: number\n /** Presence penalty (-2.0 to 2.0) */\n presence_penalty?: number\n /** Stop sequences */\n stop?: string | Array<string>\n /** A unique identifier representing your end-user */\n user?: string\n}\n\n// ===========================\n// Type Resolution Helpers\n// ===========================\n\n/**\n * Resolve provider options for a specific model.\n * If the model has explicit options in the map, use those; otherwise use base options.\n */\nexport type ResolveProviderOptions<TModel extends string> =\n TModel extends keyof GrokChatModelProviderOptionsByName\n ? GrokChatModelProviderOptionsByName[TModel]\n : GrokProviderOptions\n\n/**\n * Resolve input modalities for a specific model.\n * If the model has explicit modalities in the map, use those; otherwise use text only.\n */\nexport type ResolveInputModalities<TModel extends string> =\n TModel extends keyof GrokModelInputModalitiesByName\n ? GrokModelInputModalitiesByName[TModel]\n : readonly ['text']\n"],"names":[],"mappings":"AA0BA,MAAM,0BAA0B;AAAA,EAC9B,MAAM;AAiBR;AAEA,MAAM,8BAA8B;AAAA,EAClC,MAAM;AAiBR;AAEA,MAAM,mBAAmB;AAAA,EACvB,MAAM;AAiBR;AAEA,MAAM,wBAAwB;AAAA,EAC5B,MAAM;AAiBR;AAEA,MAAM,4BAA4B;AAAA,EAChC,MAAM;AAiBR;AAEA,MAAM,SAAS;AAAA,EACb,MAAM;AAiBR;AAEA,MAAM,cAAc;AAAA,EAClB,MAAM;AAiBR;AAEA,MAAM,SAAS;AAAA,EACb,MAAM;AAiBR;AAEA,MAAM,gBAAgB;AAAA,EACpB,MAAM;AAgBR;AAEA,MAAM,eAAe;AAAA,EACnB,MAAM;AAaR;AAMA,MAAM,YAAY;AAAA,EAChB,MAAM;AAiBR;AAEA,MAAM,wBAAwB;AAAA,EAC5B,MAAM;AAiBR;AAEO,MAAM,mBAAmB;AAAA,EAC9B,wBAAwB;AAAA,EACxB,4BAA4B;AAAA,EAC5B,iBAAiB;AAAA,EACjB,sBAAsB;AAAA,EACtB,0BAA0B;AAAA,EAC1B,OAAO;AAAA,EACP,OAAO;AAAA,EACP,YAAY;AAAA,EACZ,cAAc;AAAA,EAEd,UAAU;AAAA,EACV,sBAAsB;AACxB;AAKO,MAAM,oBAAoB,CAAC,aAAa,IAAI;AAKnD,MAAM,WAAW;AAAA,EACf,MAAM;AAKR;AAKA,MAAM,WAAW;AAAA,EACf,MAAM;AAKR;AAEA,MAAM,oBAAoB;AAAA,EACxB,MAAM;AAOR;AAEA,MAAM,0BAA0B;AAAA,EAC9B,MAAM;AAOR;AAEO,MAAM,kBAAkB,CAAC,SAAS,IAAI;AAEtC,MAAM,4BAA4B,CAAC,SAAS,IAAI;AAEhD,MAAM,uBAAuB;AAAA,EAClC,kBAAkB;AAAA,EAClB,wBAAwB;AAC1B;"}
@@ -0,0 +1,21 @@
1
+ import { RealtimeAdapter } from './realtime-contract.js';
2
+ import { GrokRealtimeOptions } from './types.js';
3
+ /**
4
+ * Creates a Grok realtime adapter for client-side use.
5
+ *
6
+ * Uses WebRTC for browser connections (default). Mirrors the OpenAI realtime
7
+ * adapter because xAI's Voice Agent API is OpenAI-realtime-compatible — the
8
+ * only differences are the endpoint URL and default model.
9
+ *
10
+ * @example
11
+ * ```typescript
12
+ * import { RealtimeClient } from '@tanstack/ai-client'
13
+ * import { grokRealtime } from '@tanstack/ai-grok'
14
+ *
15
+ * const client = new RealtimeClient({
16
+ * getToken: () => fetch('/api/realtime-token').then(r => r.json()),
17
+ * adapter: grokRealtime(),
18
+ * })
19
+ * ```
20
+ */
21
+ export declare function grokRealtime(options?: GrokRealtimeOptions): RealtimeAdapter;