@alquimia-ai/tools 1.3.2 → 1.4.1

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.
@@ -1,6 +1,6 @@
1
- import { W as WhisperProvider, R as RatingsProvider, C as CharacterizationProvider, S as StableDiffusionProvider } from '../providers-D6FJ6tlA.mjs';
2
- export { G as GenerativeProvider, L as LoggerProvider } from '../providers-D6FJ6tlA.mjs';
3
- import { RatingData } from '../types/index.mjs';
1
+ import { W as WhisperProvider, R as RatingsProvider, C as CharacterizationProvider, S as StableDiffusionProvider } from '../providers-0Gi78uNQ.mjs';
2
+ export { G as GenerativeProvider, L as LoggerProvider } from '../providers-0Gi78uNQ.mjs';
3
+ import { TTSResult, RatingData } from '../types/index.mjs';
4
4
  import '@elastic/apm-rum';
5
5
  import 'ai';
6
6
 
@@ -10,7 +10,7 @@ declare class AlquimiaWhisperProvider extends WhisperProvider {
10
10
  ttsRoute: string;
11
11
  sttRoute: string;
12
12
  });
13
- textToSpeech(text: string): Promise<any>;
13
+ textToSpeech(text: string): Promise<TTSResult>;
14
14
  speechToText(audio: string): Promise<string>;
15
15
  }
16
16
  declare class AlquimiaRatingsProvider extends RatingsProvider {
@@ -24,7 +24,7 @@ declare class AlquimiaRatingsProvider extends RatingsProvider {
24
24
  }
25
25
 
26
26
  declare class OpenAIWhisperProvider extends WhisperProvider {
27
- textToSpeech(text: string): Promise<Blob>;
27
+ textToSpeech(text: string): Promise<TTSResult>;
28
28
  speechToText(audio: string): Promise<string>;
29
29
  }
30
30
  declare class OpenAIAnalyzeCharProvider extends CharacterizationProvider {
@@ -58,7 +58,25 @@ declare class ElevenLabsWhisperProvider extends WhisperProvider {
58
58
  private requestSpecs;
59
59
  constructor(config: ElevenLabsConfig);
60
60
  speechToText(audio: string): Promise<string>;
61
- textToSpeech(text: string): Promise<any>;
61
+ textToSpeech(text: string): Promise<TTSResult>;
62
62
  }
63
63
 
64
- export { AlquimiaRatingsProvider, AlquimiaWhisperProvider, CharacterizationProvider, ElevenLabsWhisperProvider, OpenAIAnalyzeCharProvider, OpenAIStableDiffusionProvider, OpenAIWhisperProvider, RatingsProvider, StabilityProvider, StableDiffusionProvider, WhisperProvider };
64
+ interface OrpheusConfig {
65
+ baseURL?: string;
66
+ apiKey: string;
67
+ modelId: string;
68
+ voiceSettings?: {
69
+ top_p?: number;
70
+ voice?: string;
71
+ temperature?: number;
72
+ max_new_tokens?: number;
73
+ repetition_penalty?: number;
74
+ };
75
+ }
76
+ declare class OrpheusWhisperProvider extends WhisperProvider {
77
+ constructor(config: OrpheusConfig);
78
+ speechToText(audio: string): Promise<string>;
79
+ textToSpeech(text: string): Promise<TTSResult>;
80
+ }
81
+
82
+ export { AlquimiaRatingsProvider, AlquimiaWhisperProvider, CharacterizationProvider, ElevenLabsWhisperProvider, OpenAIAnalyzeCharProvider, OpenAIStableDiffusionProvider, OpenAIWhisperProvider, OrpheusWhisperProvider, RatingsProvider, StabilityProvider, StableDiffusionProvider, WhisperProvider };
@@ -1,6 +1,6 @@
1
- import { W as WhisperProvider, R as RatingsProvider, C as CharacterizationProvider, S as StableDiffusionProvider } from '../providers-D6FJ6tlA.js';
2
- export { G as GenerativeProvider, L as LoggerProvider } from '../providers-D6FJ6tlA.js';
3
- import { RatingData } from '../types/index.js';
1
+ import { W as WhisperProvider, R as RatingsProvider, C as CharacterizationProvider, S as StableDiffusionProvider } from '../providers-BJTXCtI3.js';
2
+ export { G as GenerativeProvider, L as LoggerProvider } from '../providers-BJTXCtI3.js';
3
+ import { TTSResult, RatingData } from '../types/index.js';
4
4
  import '@elastic/apm-rum';
5
5
  import 'ai';
6
6
 
@@ -10,7 +10,7 @@ declare class AlquimiaWhisperProvider extends WhisperProvider {
10
10
  ttsRoute: string;
11
11
  sttRoute: string;
12
12
  });
13
- textToSpeech(text: string): Promise<any>;
13
+ textToSpeech(text: string): Promise<TTSResult>;
14
14
  speechToText(audio: string): Promise<string>;
15
15
  }
16
16
  declare class AlquimiaRatingsProvider extends RatingsProvider {
@@ -24,7 +24,7 @@ declare class AlquimiaRatingsProvider extends RatingsProvider {
24
24
  }
25
25
 
26
26
  declare class OpenAIWhisperProvider extends WhisperProvider {
27
- textToSpeech(text: string): Promise<Blob>;
27
+ textToSpeech(text: string): Promise<TTSResult>;
28
28
  speechToText(audio: string): Promise<string>;
29
29
  }
30
30
  declare class OpenAIAnalyzeCharProvider extends CharacterizationProvider {
@@ -58,7 +58,25 @@ declare class ElevenLabsWhisperProvider extends WhisperProvider {
58
58
  private requestSpecs;
59
59
  constructor(config: ElevenLabsConfig);
60
60
  speechToText(audio: string): Promise<string>;
61
- textToSpeech(text: string): Promise<any>;
61
+ textToSpeech(text: string): Promise<TTSResult>;
62
62
  }
63
63
 
64
- export { AlquimiaRatingsProvider, AlquimiaWhisperProvider, CharacterizationProvider, ElevenLabsWhisperProvider, OpenAIAnalyzeCharProvider, OpenAIStableDiffusionProvider, OpenAIWhisperProvider, RatingsProvider, StabilityProvider, StableDiffusionProvider, WhisperProvider };
64
+ interface OrpheusConfig {
65
+ baseURL?: string;
66
+ apiKey: string;
67
+ modelId: string;
68
+ voiceSettings?: {
69
+ top_p?: number;
70
+ voice?: string;
71
+ temperature?: number;
72
+ max_new_tokens?: number;
73
+ repetition_penalty?: number;
74
+ };
75
+ }
76
+ declare class OrpheusWhisperProvider extends WhisperProvider {
77
+ constructor(config: OrpheusConfig);
78
+ speechToText(audio: string): Promise<string>;
79
+ textToSpeech(text: string): Promise<TTSResult>;
80
+ }
81
+
82
+ export { AlquimiaRatingsProvider, AlquimiaWhisperProvider, CharacterizationProvider, ElevenLabsWhisperProvider, OpenAIAnalyzeCharProvider, OpenAIStableDiffusionProvider, OpenAIWhisperProvider, OrpheusWhisperProvider, RatingsProvider, StabilityProvider, StableDiffusionProvider, WhisperProvider };
@@ -1819,6 +1819,7 @@ __export(providers_exports, {
1819
1819
  OpenAIAnalyzeCharProvider: () => OpenAIAnalyzeCharProvider,
1820
1820
  OpenAIStableDiffusionProvider: () => OpenAIStableDiffusionProvider,
1821
1821
  OpenAIWhisperProvider: () => OpenAIWhisperProvider,
1822
+ OrpheusWhisperProvider: () => OrpheusWhisperProvider,
1822
1823
  RatingsProvider: () => RatingsProvider,
1823
1824
  StabilityProvider: () => StabilityProvider,
1824
1825
  StableDiffusionProvider: () => StableDiffusionProvider,
@@ -1876,7 +1877,7 @@ var AlquimiaWhisperProvider = class extends WhisperProvider {
1876
1877
  },
1877
1878
  { responseType: "blob" }
1878
1879
  );
1879
- return response.data;
1880
+ return { type: "blob", data: response.data };
1880
1881
  } catch (error) {
1881
1882
  console.error("Error converting text to speech:", error);
1882
1883
  throw error;
@@ -1976,7 +1977,7 @@ function generateTranslatePrompt(text) {
1976
1977
  var import_openai = __toESM(require("openai"));
1977
1978
  var OpenAIWhisperProvider = class extends WhisperProvider {
1978
1979
  async textToSpeech(text) {
1979
- return new Blob();
1980
+ return { type: "url", data: "" };
1980
1981
  }
1981
1982
  async speechToText(audio) {
1982
1983
  return "";
@@ -2128,13 +2129,65 @@ var ElevenLabsWhisperProvider = class extends WhisperProvider {
2128
2129
  },
2129
2130
  { responseType: "blob" }
2130
2131
  );
2131
- return response.data;
2132
+ return { type: "blob", data: response.data };
2132
2133
  } catch (error) {
2133
2134
  console.error("Error converting text to speech:", error);
2134
2135
  throw error;
2135
2136
  }
2136
2137
  }
2137
2138
  };
2139
+
2140
+ // src/providers/orpheus.ts
2141
+ var import_axios4 = __toESM(require("axios"));
2142
+ var defaultVoiceSettings2 = {
2143
+ top_p: 0.95,
2144
+ voice: "tara",
2145
+ temperature: 0.6,
2146
+ max_new_tokens: 1200,
2147
+ repetition_penalty: 1.1
2148
+ };
2149
+ var OrpheusWhisperProvider = class extends WhisperProvider {
2150
+ constructor(config) {
2151
+ super(config);
2152
+ }
2153
+ async speechToText(audio) {
2154
+ return "";
2155
+ }
2156
+ async textToSpeech(text) {
2157
+ const axiosClient = import_axios4.default.create({
2158
+ baseURL: this.config.baseURL || "https://api.replicate.com",
2159
+ headers: {
2160
+ "Authorization": `Bearer ${this.config.apiKey}`,
2161
+ "Content-Type": "application/json",
2162
+ "Prefer": "wait"
2163
+ }
2164
+ });
2165
+ try {
2166
+ const response = await axiosClient.post(
2167
+ "/v1/predictions",
2168
+ {
2169
+ version: this.config.modelId,
2170
+ input: {
2171
+ text,
2172
+ ...defaultVoiceSettings2,
2173
+ ...this.config.voiceSettings || {}
2174
+ }
2175
+ }
2176
+ );
2177
+ if (response.data?.output) {
2178
+ return { type: "url", data: response.data.output };
2179
+ }
2180
+ throw new Error("No output URL found in Replicate response");
2181
+ } catch (error) {
2182
+ console.error("Error in textToSpeech:", {
2183
+ message: error.message,
2184
+ response: error.response?.data,
2185
+ status: error.response?.status
2186
+ });
2187
+ throw error;
2188
+ }
2189
+ }
2190
+ };
2138
2191
  /*! Bundled license information:
2139
2192
 
2140
2193
  ieee754/index.js: