@brotu/ai 0.1.0 → 0.2.0

package/dist/index.js

@@ -5,8 +5,11 @@ import {
   ELEVENLABS_CATALOG,
   ELEVENLABS_MODELS,
   ELEVENLABS_OUTPUT_FORMATS,
+  GOOGLE_AUDIO_MODELS,
   GOOGLE_CATALOG,
   GOOGLE_IMAGE_MODELS,
+  GOOGLE_TEXT_MODELS,
+  GOOGLE_TTS_VOICES,
   GOOGLE_VIDEO_MODELS,
   KLING_AUDIO_MODEL,
   KLING_CAPABILITIES,
@@ -15,6 +18,7 @@ import {
   KLING_VOICES,
   OPENAI_CATALOG,
   OPENAI_IMAGE_MODELS,
+  OPENAI_TEXT_MODELS,
   QWEN_AUDIO_MODELS,
   QWEN_CATALOG,
   QWEN_IMAGE_MODELS,
@@ -31,7 +35,7 @@ import {
   resetCatalog,
   resolveProvider,
   videoPathFor
-} from "./chunk-GKLTQ55S.js";
+} from "./chunk-PL534BFN.js";
 
 // src/lib/jobs.ts
 import { AsyncLocalStorage } from "async_hooks";
@@ -582,7 +586,12 @@ var POLL_INTERVAL_MS2 = 1e4;
 var DEFAULT_MAX_POLL_ATTEMPTS2 = 120;
 var GoogleAdapter = class {
   providerName = "google";
-  supportedTypes = ["image", "video"];
+  supportedTypes = [
+    "image",
+    "video",
+    "text",
+    "audio"
+  ];
   opts;
   constructor(opts) {
     this.opts = opts;
@@ -904,17 +913,143 @@ var GoogleAdapter = class {
   generateVideo(params) {
     return this.runVideo(params);
   }
-  generateAudio(_params) {
-    throw new Error("Google has no speech synthesis wired up here.");
+  textBody(modelId, params) {
+    const input = [
+      { type: "text", text: params.prompt }
+    ];
+    for (const image of params.referenceImages ?? []) {
+      const [meta, data] = image.split(",");
+      input.push({
+        type: "image",
+        mime_type: meta?.match(/data:([^;]+)/)?.[1] ?? "image/png",
+        data: data ?? image
+      });
+    }
+    return {
+      model: modelId,
+      input,
+      ...params.systemPrompt ? { instructions: params.systemPrompt } : {},
+      generation_config: {
+        ...params.maxTokens ? { max_output_tokens: params.maxTokens } : {},
+        ...params.temperature !== void 0 ? { temperature: params.temperature } : {},
+        ...params.topP !== void 0 ? { top_p: params.topP } : {}
+      },
+      ...params.providerOptions?.google ?? {}
+    };
   }
-  generateText(_params) {
-    throw new Error("The Google adapter here covers image and video only.");
+  speechBody(modelId, params, voice) {
+    return {
+      model: modelId,
+      input: params.prompt,
+      response_format: { type: "audio" },
+      generation_config: {
+        speech_config: [{ voice }]
+      },
+      ...params.providerOptions?.google ?? {}
+    };
+  }
+  async generateText(params) {
+    const startedAt = Date.now();
+    const modelId = params.model ?? "";
+    const failure = (error) => ({
+      success: false,
+      outputs: [],
+      creditsUsed: 0,
+      provider: this.providerName,
+      model: modelId,
+      processingTimeMs: Date.now() - startedAt,
+      error
+    });
+    if (!(modelId in GOOGLE_TEXT_MODELS)) {
+      return failure(
+        `"${modelId}" is not a Gemini text model. Known: ${Object.keys(GOOGLE_TEXT_MODELS).join(", ")}.`
+      );
+    }
+    try {
+      const payload = await this.request(
+        INTERACTIONS_PATH,
+        {
+          method: "POST",
+          body: this.textBody(modelId, params)
+        }
+      );
+      const text = payload.output_text ?? payload.steps?.flatMap((step) => step.content ?? []).find((part) => part.type === "text" || part.text)?.text;
+      if (!text) return failure("Google returned no text.");
+      return {
+        success: true,
+        outputs: [
+          {
+            url: `data:text/plain;base64,${Buffer.from(text).toString("base64")}`,
+            mimeType: "text/plain",
+            raw: { text }
+          }
+        ],
+        creditsUsed: 0,
+        provider: this.providerName,
+        model: modelId,
+        processingTimeMs: Date.now() - startedAt
+      };
+    } catch (error) {
+      return failure(error instanceof Error ? error.message : String(error));
+    }
+  }
+  async generateAudio(params) {
+    const startedAt = Date.now();
+    const modelId = params.model ?? "";
+    const failure = (error) => ({
+      success: false,
+      outputs: [],
+      creditsUsed: 0,
+      provider: this.providerName,
+      model: modelId,
+      processingTimeMs: Date.now() - startedAt,
+      error
+    });
+    const binding = GOOGLE_AUDIO_MODELS[modelId];
+    if (!binding) {
+      return failure(
+        `"${modelId}" is not a Gemini speech model. Known: ${Object.keys(GOOGLE_AUDIO_MODELS).join(", ")}.`
+      );
+    }
+    const voice = params.voice ?? "Kore";
+    if (!binding.voices.includes(voice)) {
+      return failure(
+        `Gemini TTS offers ${GOOGLE_TTS_VOICES.join(", ")}, not "${voice}".`
+      );
+    }
+    try {
+      const payload = await this.request(
+        INTERACTIONS_PATH,
+        {
+          method: "POST",
+          body: this.speechBody(modelId, params, voice)
+        }
+      );
+      const data = payload.output_audio?.data;
+      if (!data) return failure("Google returned no audio.");
+      return {
+        success: true,
+        outputs: [
+          {
+            url: `data:audio/wav;base64,${data}`,
+            mimeType: "audio/wav",
+            raw: { inline: true, voice }
+          }
+        ],
+        creditsUsed: 0,
+        provider: this.providerName,
+        model: modelId,
+        processingTimeMs: Date.now() - startedAt
+      };
+    } catch (error) {
+      return failure(error instanceof Error ? error.message : String(error));
+    }
   }
   async estimateCost(type, params) {
     return estimateFor(this.providerName, type, params);
   }
   supportsModel(model) {
-    return model in GOOGLE_IMAGE_MODELS || model in GOOGLE_VIDEO_MODELS;
+    return model in GOOGLE_IMAGE_MODELS || model in GOOGLE_VIDEO_MODELS || model in GOOGLE_TEXT_MODELS || model in GOOGLE_AUDIO_MODELS;
   }
   getAvailableModels() {
     return [
@@ -927,6 +1062,16 @@ var GoogleAdapter = class {
         id,
         name: id,
         type: "video"
+      })),
+      ...Object.keys(GOOGLE_TEXT_MODELS).map((id) => ({
+        id,
+        name: id,
+        type: "text"
+      })),
+      ...Object.keys(GOOGLE_AUDIO_MODELS).map((id) => ({
+        id,
+        name: id,
+        type: "audio"
       }))
     ];
   }
@@ -1283,7 +1428,7 @@ var IMAGES_PATH2 = "/v1/images/generations";
 var OpenAIAdapter = class {
   providerName = "openai";
   // No video: the Videos API shuts down on 24 September 2026.
-  supportedTypes = ["image"];
+  supportedTypes = ["image", "text"];
   opts;
   constructor(opts) {
     this.opts = opts;
@@ -1398,8 +1543,79 @@ var OpenAIAdapter = class {
   generateAudio(_params) {
     throw new Error("OpenAI has no speech synthesis wired up here.");
   }
-  generateText(_params) {
-    throw new Error("The OpenAI adapter here covers images only.");
+  textBody(params) {
+    const content = params.referenceImages?.length ? [
+      { type: "input_text", text: params.prompt },
+      ...params.referenceImages.map((url) => ({
+        type: "input_image",
+        image_url: url
+      }))
+    ] : params.prompt;
+    const input = [];
+    if (params.systemPrompt) {
+      input.push({ role: "system", content: params.systemPrompt });
+    }
+    input.push({ role: "user", content });
+    return {
+      model: params.model,
+      input,
+      ...params.maxTokens ? { max_output_tokens: params.maxTokens } : {},
+      ...params.temperature !== void 0 ? { temperature: params.temperature } : {},
+      ...params.topP !== void 0 ? { top_p: params.topP } : {},
+      ...params.providerOptions?.openai ?? {}
+    };
+  }
+  async generateText(params) {
+    const startedAt = Date.now();
+    const modelId = params.model ?? "";
+    const failure = (error) => ({
+      success: false,
+      outputs: [],
+      creditsUsed: 0,
+      provider: this.providerName,
+      model: modelId,
+      processingTimeMs: Date.now() - startedAt,
+      error
+    });
+    if (!(modelId in OPENAI_TEXT_MODELS)) {
+      return failure(
+        `"${modelId}" is not an OpenAI text model. Known: ${Object.keys(OPENAI_TEXT_MODELS).join(", ")}.`
+      );
+    }
+    try {
+      const response = await fetch(`${this.baseUrl}/v1/responses`, {
+        method: "POST",
+        headers: {
+          Authorization: `Bearer ${this.opts.apiKey}`,
+          "Content-Type": "application/json",
+          ...this.opts.organization ? { "OpenAI-Organization": this.opts.organization } : {}
+        },
+        body: JSON.stringify(this.textBody(params))
+      });
+      const payload = await response.json();
+      if (payload.error) throw new Error(payload.error.message);
+      if (!response.ok) {
+        throw new Error(`OpenAI returned ${response.status}.`);
+      }
+      const text = payload.output_text ?? payload.output?.flatMap((item) => item.content ?? []).find((part) => part.type === "output_text" || part.text)?.text;
+      if (!text) return failure("OpenAI returned no text.");
+      return {
+        success: true,
+        outputs: [
+          {
+            url: `data:text/plain;base64,${Buffer.from(text).toString("base64")}`,
+            mimeType: "text/plain",
+            raw: { text, tokens: payload.usage?.output_tokens }
+          }
+        ],
+        creditsUsed: 0,
+        provider: this.providerName,
+        model: modelId,
+        processingTimeMs: Date.now() - startedAt
+      };
+    } catch (error) {
+      return failure(error instanceof Error ? error.message : String(error));
+    }
   }
   /** Synchronous provider: there is never a queued job to come back to. */
   async completeJob(job) {
@@ -1420,14 +1636,21 @@ var OpenAIAdapter = class {
     };
   }
   supportsModel(model) {
-    return model in OPENAI_IMAGE_MODELS;
+    return model in OPENAI_IMAGE_MODELS || model in OPENAI_TEXT_MODELS;
   }
   getAvailableModels() {
-    return Object.keys(OPENAI_IMAGE_MODELS).map((id) => ({
-      id,
-      name: id,
-      type: "image"
-    }));
+    return [
+      ...Object.keys(OPENAI_IMAGE_MODELS).map((id) => ({
+        id,
+        name: id,
+        type: "image"
+      })),
+      ...Object.keys(OPENAI_TEXT_MODELS).map((id) => ({
+        id,
+        name: id,
+        type: "text"
+      }))
+    ];
   }
 };
 
@@ -2351,8 +2574,11 @@ export {
   ELEVENLABS_MODELS,
   ELEVENLABS_OUTPUT_FORMATS,
   ElevenLabsAdapter,
+  GOOGLE_AUDIO_MODELS,
   GOOGLE_CATALOG,
   GOOGLE_IMAGE_MODELS,
+  GOOGLE_TEXT_MODELS,
+  GOOGLE_TTS_VOICES,
   GOOGLE_VIDEO_MODELS,
   GoogleAdapter,
   KLING_CAPABILITIES,
@@ -2361,6 +2587,7 @@ export {
   KlingAdapter,
   OPENAI_CATALOG,
   OPENAI_IMAGE_MODELS,
+  OPENAI_TEXT_MODELS,
   OpenAIAdapter,
   PendingJob,
   QWEN_AUDIO_MODELS,

package/dist/providers/google.models.d.ts

@@ -38,5 +38,28 @@ export interface GoogleVideoBinding {
     usdByResolution?: Record<string, number>;
 }
 export declare const GOOGLE_VIDEO_MODELS: Record<string, GoogleVideoBinding>;
+/**
+ * Gemini text, through the same Interactions API as images. Rates are USD per
+ * million output tokens, list price on the Gemini Developer API paid tier.
+ * Promotional intro rates (3.6 / 3.7 through 31 Dec 2026) are what the
+ * pricing page shows today; a guess at the 2027 step-up would go stale.
+ */
+export declare const GOOGLE_TEXT_MODELS: Record<string, {
+    usdPerMillionOutput?: number;
+}>;
+/**
+ * Official prebuilt voices for Gemini TTS. The Interactions API takes one of
+ * these in `generation_config.speech_config[].voice`.
+ */
+export declare const GOOGLE_TTS_VOICES: readonly ["Zephyr", "Puck", "Charon", "Kore", "Fenrir", "Leda", "Orus", "Aoede", "Callirrhoe", "Autonoe", "Enceladus", "Iapetus", "Umbriel", "Algieba", "Despina", "Erinome", "Algenib", "Rasalgethi", "Laomedeia", "Achernar", "Alnilam", "Schedar", "Gacrux", "Pulcherrima", "Achird", "Zubenelgenubi", "Vindemiatrix", "Sadachbia", "Sadaltager", "Sulafat"];
+export type GoogleTtsVoice = (typeof GOOGLE_TTS_VOICES)[number];
+/**
+ * Gemini TTS. Audio-only in, audio-only out, on the Interactions API.
+ * Token rates are not listed next to the model cards the way Veo is, so
+ * the catalog leaves them unset rather than guessing.
+ */
+export declare const GOOGLE_AUDIO_MODELS: Record<string, {
+    voices: readonly string[];
+}>;
 export declare const GOOGLE_CATALOG: AIModelConfig[];
 //# sourceMappingURL=google.models.d.ts.map

package/dist/providers/google.models.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"google.models.d.ts","sourceRoot":"","sources":["../../src/providers/google.models.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,0BAA0B,CAAC;AAE9D;;;;;;;;;;;;GAYG;AAEH,MAAM,WAAW,kBAAkB;IAClC,+EAA+E;IAC/E,KAAK,EAAE,MAAM,EAAE,CAAC;IAChB,8EAA8E;IAC9E,OAAO,EAAE,MAAM,CAAC;CAChB;AAED,eAAO,MAAM,mBAAmB,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAWlE,CAAC;AAEF,MAAM,WAAW,kBAAkB;IAClC;;;;OAIG;IACH,GAAG,EAAE,SAAS,GAAG,cAAc,CAAC;IAChC,WAAW,EAAE,MAAM,EAAE,CAAC;IACtB,sCAAsC;IACtC,SAAS,EAAE,MAAM,EAAE,CAAC;IACpB,eAAe,EAAE,OAAO,CAAC;IACzB,wEAAwE;IACxE,qBAAqB,CAAC,EAAE,OAAO,CAAC;IAChC,8EAA8E;IAC9E,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,kDAAkD;IAClD,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;CACzC;AAED,eAAO,MAAM,mBAAmB,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CA0ClE,CAAC;AAEF,eAAO,MAAM,cAAc,EAAE,aAAa,EA0CzC,CAAC"}
+{"version":3,"file":"google.models.d.ts","sourceRoot":"","sources":["../../src/providers/google.models.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,0BAA0B,CAAC;AAE9D;;;;;;;;;;;;GAYG;AAEH,MAAM,WAAW,kBAAkB;IAClC,+EAA+E;IAC/E,KAAK,EAAE,MAAM,EAAE,CAAC;IAChB,8EAA8E;IAC9E,OAAO,EAAE,MAAM,CAAC;CAChB;AAED,eAAO,MAAM,mBAAmB,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAWlE,CAAC;AAEF,MAAM,WAAW,kBAAkB;IAClC;;;;OAIG;IACH,GAAG,EAAE,SAAS,GAAG,cAAc,CAAC;IAChC,WAAW,EAAE,MAAM,EAAE,CAAC;IACtB,sCAAsC;IACtC,SAAS,EAAE,MAAM,EAAE,CAAC;IACpB,eAAe,EAAE,OAAO,CAAC;IACzB,wEAAwE;IACxE,qBAAqB,CAAC,EAAE,OAAO,CAAC;IAChC,8EAA8E;IAC9E,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,kDAAkD;IAClD,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;CACzC;AAED,eAAO,MAAM,mBAAmB,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CA0ClE,CAAC;AAEF;;;;;GAKG;AACH,eAAO,MAAM,kBAAkB,EAAE,MAAM,CACtC,MAAM,EACN;IAAE,mBAAmB,CAAC,EAAE,MAAM,CAAA;CAAE,CAYhC,CAAC;AAEF;;;GAGG;AACH,eAAO,MAAM,iBAAiB,YAC7B,QAAQ,EACR,MAAM,EACN,QAAQ,EACR,MAAM,EACN,QAAQ,EACR,MAAM,EACN,MAAM,EACN,OAAO,EACP,YAAY,EACZ,SAAS,EACT,WAAW,EACX,SAAS,EACT,SAAS,EACT,SAAS,EACT,SAAS,EACT,SAAS,EACT,SAAS,EACT,YAAY,EACZ,WAAW,EACX,UAAU,EACV,SAAS,EACT,SAAS,EACT,QAAQ,EACR,aAAa,EACb,QAAQ,EACR,eAAe,EACf,cAAc,EACd,WAAW,EACX,YAAY,EACZ,SAAS,CACA,CAAC;AAEX,MAAM,MAAM,cAAc,GAAG,CAAC,OAAO,iBAAiB,CAAC,CAAC,MAAM,CAAC,CAAC;AAEhE;;;;GAIG;AACH,eAAO,MAAM,mBAAmB,EAAE,MAAM,CACvC,MAAM,EACN;IAAE,MAAM,EAAE,SAAS,MAAM,EAAE,CAAA;CAAE,CAK7B,CAAC;AAEF,eAAO,MAAM,cAAc,EAAE,aAAa,EA4EzC,CAAC"}

package/dist/providers/openai.models.d.ts

@@ -23,5 +23,12 @@ export interface OpenAIImageBinding {
     };
 }
 export declare const OPENAI_IMAGE_MODELS: Record<string, OpenAIImageBinding>;
+/**
+ * Current GPT-5.6 text ladder, as published on the Models page. Rates are
+ * USD per million output tokens on the standard paid tier.
+ */
+export declare const OPENAI_TEXT_MODELS: Record<string, {
+    usdPerMillionOutput: number;
+}>;
 export declare const OPENAI_CATALOG: AIModelConfig[];
 //# sourceMappingURL=openai.models.d.ts.map

package/dist/providers/openai.models.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"openai.models.d.ts","sourceRoot":"","sources":["../../src/providers/openai.models.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,0BAA0B,CAAC;AAE9D;;;;;;GAMG;AACH,MAAM,WAAW,kBAAkB;IAClC,4EAA4E;IAC5E,KAAK,EAAE,MAAM,EAAE,GAAG,KAAK,CAAC;IACxB,oDAAoD;IACpD,SAAS,EAAE,KAAK,CAAC,KAAK,GAAG,QAAQ,GAAG,MAAM,CAAC,CAAC;IAC5C,8DAA8D;IAC9D,qBAAqB,EAAE,OAAO,CAAC;IAC/B,wDAAwD;IACxD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,+CAA+C;IAC/C,GAAG,EAAE;QAAE,GAAG,EAAE,MAAM,CAAC;QAAC,MAAM,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;CACnD;AAID,eAAO,MAAM,mBAAmB,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAuBlE,CAAC;AAEF,eAAO,MAAM,cAAc,EAAE,aAAa,EAkBvC,CAAC"}
+{"version":3,"file":"openai.models.d.ts","sourceRoot":"","sources":["../../src/providers/openai.models.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,0BAA0B,CAAC;AAE9D;;;;;;GAMG;AACH,MAAM,WAAW,kBAAkB;IAClC,4EAA4E;IAC5E,KAAK,EAAE,MAAM,EAAE,GAAG,KAAK,CAAC;IACxB,oDAAoD;IACpD,SAAS,EAAE,KAAK,CAAC,KAAK,GAAG,QAAQ,GAAG,MAAM,CAAC,CAAC;IAC5C,8DAA8D;IAC9D,qBAAqB,EAAE,OAAO,CAAC;IAC/B,wDAAwD;IACxD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,+CAA+C;IAC/C,GAAG,EAAE;QAAE,GAAG,EAAE,MAAM,CAAC;QAAC,MAAM,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;CACnD;AAID,eAAO,MAAM,mBAAmB,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAuBlE,CAAC;AAEF;;;GAGG;AACH,eAAO,MAAM,kBAAkB,EAAE,MAAM,CACtC,MAAM,EACN;IAAE,mBAAmB,EAAE,MAAM,CAAA;CAAE,CAK/B,CAAC;AAEF,eAAO,MAAM,cAAc,EAAE,aAAa,EAgCzC,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@brotu/ai",
-	"version": "0.1.0",
+	"version": "0.2.0",
 	"type": "module",
 	"main": "./dist/index.cjs",
 	"exports": {
@@ -66,7 +66,11 @@
 	"module": "./dist/index.js",
 	"repository": {
 		"type": "git",
-		"url": "git+https://github.com/brotu/brotu-sdk.git",
+		"url": "git+https://github.com/Zorbi-Tech/brotu-sdk.git",
 		"directory": "sdks/node"
+	},
+	"homepage": "https://github.com/Zorbi-Tech/brotu-sdk",
+	"bugs": {
+		"url": "https://github.com/Zorbi-Tech/brotu-sdk/issues"
 	}
 }