vidspotai-shared 1.0.82-dev.0 → 1.0.83

package/lib/services/aiGen/aiGenFactory.service.js

@@ -13,6 +13,7 @@ const alibaba_1 = require("./providers/alibaba/alibaba");
 const pixverse_service_1 = require("./providers/pixverse/pixverse.service");
 const bytedance_service_1 = require("./providers/bytedance/bytedance.service");
 const elevenlabs_service_1 = require("./providers/elevenlabs/elevenlabs.service");
+const anthropic_service_1 = require("./providers/anthropic/anthropic.service");
 exports.videoServicesMap = {
     // --- Pixverse ---
     "pixverse-v6": pixverse_service_1.PixVerseService,
@@ -134,12 +135,26 @@ exports.sfxServicesMap = {
     [aiModels_1.ESfxGenModels.ELEVENLABS_SOUND_EFFECTS]: elevenlabs_service_1.ElevenLabsAiGenService,
 };
 exports.textServicesMap = {
-    "openai-gpt-4o-mini": openai_service_1.OpenaiService,
+    // --- OpenAI ---
+    "openai-gpt-5.5": openai_service_1.OpenaiService,
+    "openai-gpt-5.4-mini": openai_service_1.OpenaiService,
     "openai-gpt-5-mini": openai_service_1.OpenaiService,
+    "openai-gpt-4o-mini": openai_service_1.OpenaiService,
+    // Native Anthropic / Gemini text-gen (LLM_NATIVE_MIGRATION_PLAN Phase 6).
+    // GoogleService gained a generateText() alongside its video/image/music
+    // methods; AnthropicService is text-only.
+    // --- Anthropic ---
+    "anthropic-claude-fable-5": anthropic_service_1.AnthropicService,
+    "anthropic-claude-opus-4-8": anthropic_service_1.AnthropicService,
+    "anthropic-claude-sonnet-4-6": anthropic_service_1.AnthropicService,
+    "anthropic-claude-haiku-4-5": anthropic_service_1.AnthropicService,
+    // --- Google ---
+    "google-gemini-3.5-flash": google_service_1.GoogleService,
+    "google-gemini-3.1-pro": google_service_1.GoogleService,
+    "google-gemini-2.5-flash": google_service_1.GoogleService,
+    "google-gemini-2.5-pro": google_service_1.GoogleService,
     // "openai-gpt-4o": OpenaiService,
     // "openai-gpt-3.5-turbo": OpenaiService,
-    // "anthropic-claude-3": ClaudeService,
-    // "anthropic-claude-instant-100k": ClaudeService,
 };
 function getAiGenProviderService(modelKey) {
     if (modelKey in exports.videoServicesMap) {

package/lib/services/aiGen/helpers.d.ts

@@ -14,5 +14,13 @@ export declare function modelSupportsLastFrame(modelKey: string): boolean;
  * a model that can't do it. Derived from the configs so it never drifts.
  */
 export declare function getModelsSupportingLastFrame(): string[];
+/**
+ * Generic capability probe: does the model's config DECLARE this field? Source
+ * of truth for "should we send this optional param to the provider", so a model
+ * that doesn't list e.g. `negative_prompt` never has it forwarded (and never
+ * eats an opaque provider 400 like veo-3.1-lite's "negativePrompt isn't
+ * supported by this model"). Same pattern as modelSupportsLastFrame, generalized.
+ */
+export declare function modelDeclaresField(modelKey: string, field: string): boolean;
 export declare function validateParams(params: VideoGenerationParams): boolean;
 //# sourceMappingURL=helpers.d.ts.map

package/lib/services/aiGen/helpers.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"helpers.d.ts","sourceRoot":"","sources":["../../../src/services/aiGen/helpers.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,qBAAqB,EAAE,MAAM,mBAAmB,CAAC;AAE1D;;;;;;;GAOG;AACH,wBAAgB,sBAAsB,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAGhE;AAED;;;;GAIG;AACH,wBAAgB,4BAA4B,IAAI,MAAM,EAAE,CAEvD;AAED,wBAAgB,cAAc,CAAC,MAAM,EAAE,qBAAqB,GAAG,OAAO,CAqJrE"}
+{"version":3,"file":"helpers.d.ts","sourceRoot":"","sources":["../../../src/services/aiGen/helpers.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,qBAAqB,EAAE,MAAM,mBAAmB,CAAC;AAE1D;;;;;;;GAOG;AACH,wBAAgB,sBAAsB,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAGhE;AAED;;;;GAIG;AACH,wBAAgB,4BAA4B,IAAI,MAAM,EAAE,CAEvD;AAED;;;;;;GAMG;AACH,wBAAgB,kBAAkB,CAAC,QAAQ,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,OAAO,CAG3E;AAED,wBAAgB,cAAc,CAAC,MAAM,EAAE,qBAAqB,GAAG,OAAO,CAqJrE"}

package/lib/services/aiGen/helpers.js

@@ -2,6 +2,7 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.modelSupportsLastFrame = modelSupportsLastFrame;
 exports.getModelsSupportingLastFrame = getModelsSupportingLastFrame;
+exports.modelDeclaresField = modelDeclaresField;
 exports.validateParams = validateParams;
 const aiModels_1 = require("../../globals/aiModels");
 /**
@@ -24,6 +25,17 @@ function modelSupportsLastFrame(modelKey) {
 function getModelsSupportingLastFrame() {
     return Object.keys(aiModels_1.aiModelConfigs).filter((key) => modelSupportsLastFrame(key));
 }
+/**
+ * Generic capability probe: does the model's config DECLARE this field? Source
+ * of truth for "should we send this optional param to the provider", so a model
+ * that doesn't list e.g. `negative_prompt` never has it forwarded (and never
+ * eats an opaque provider 400 like veo-3.1-lite's "negativePrompt isn't
+ * supported by this model"). Same pattern as modelSupportsLastFrame, generalized.
+ */
+function modelDeclaresField(modelKey, field) {
+    const cfg = aiModels_1.aiModelConfigs[modelKey];
+    return !!cfg?.fields && field in cfg.fields;
+}
 function validateParams(params) {
     const modelConfig = aiModels_1.aiModelConfigs[params.modelKey];
     if (!modelConfig) {

package/lib/services/aiGen/providers/anthropic/anthropic.service.d.ts

@@ -0,0 +1,26 @@
+import { BaseAiGenProviderService } from "../baseAiGenProvider.service";
+import { TextGenerationParams, TextGenerationResult } from "../types";
+/**
+ * Native Anthropic text-gen service for the aiGen text factory
+ * (LLM_NATIVE_MIGRATION_PLAN Phase 6). Anthropic is text-only in our stack, so
+ * this service implements `generateText` only — every other capability falls
+ * through to BaseAiGenProviderService's throw.
+ *
+ * Mirrors OpenaiService.generateText's contract: accepts a bare prompt string
+ * or an OpenAI-style message array (role + string content). System messages
+ * fold into Anthropic's top-level `system`; the rest map to user/assistant
+ * turns. Reads ANTHROPIC_API_KEY (same key the agent AnthropicLlmCaller uses).
+ */
+export declare class AnthropicService extends BaseAiGenProviderService {
+    private client;
+    constructor();
+    generateText(params: TextGenerationParams): Promise<TextGenerationResult>;
+    /**
+     * Map an OpenAI-style `input` (string | {role, content}[]) onto Anthropic's
+     * shape: a top-level `system` string + user/assistant MessageParam[]. Array
+     * `content` (vision parts) is flattened to its text segments — the text
+     * factory is text-only; image inputs go through the agent AnthropicLlmCaller.
+     */
+    private encodeTextInput;
+}
+//# sourceMappingURL=anthropic.service.d.ts.map

package/lib/services/aiGen/providers/anthropic/anthropic.service.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"anthropic.service.d.ts","sourceRoot":"","sources":["../../../../../src/services/aiGen/providers/anthropic/anthropic.service.ts"],"names":[],"mappings":"AAKA,OAAO,EAAE,wBAAwB,EAAE,MAAM,8BAA8B,CAAC;AACxE,OAAO,EAAE,oBAAoB,EAAE,oBAAoB,EAAE,MAAM,UAAU,CAAC;AAEtE;;;;;;;;;;GAUG;AACH,qBAAa,gBAAiB,SAAQ,wBAAwB;IAC5D,OAAO,CAAC,MAAM,CAAY;;IAOpB,YAAY,CAChB,MAAM,EAAE,oBAAoB,GAC3B,OAAO,CAAC,oBAAoB,CAAC;IAuChC;;;;;OAKG;IACH,OAAO,CAAC,eAAe;CAuCxB"}

package/lib/services/aiGen/providers/anthropic/anthropic.service.js

@@ -0,0 +1,95 @@
+"use strict";
+// anthropic.service.ts
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.AnthropicService = void 0;
+const sdk_1 = __importDefault(require("@anthropic-ai/sdk"));
+const aiModels_1 = require("../../../../globals/aiModels");
+const baseAiGenProvider_service_1 = require("../baseAiGenProvider.service");
+/**
+ * Native Anthropic text-gen service for the aiGen text factory
+ * (LLM_NATIVE_MIGRATION_PLAN Phase 6). Anthropic is text-only in our stack, so
+ * this service implements `generateText` only — every other capability falls
+ * through to BaseAiGenProviderService's throw.
+ *
+ * Mirrors OpenaiService.generateText's contract: accepts a bare prompt string
+ * or an OpenAI-style message array (role + string content). System messages
+ * fold into Anthropic's top-level `system`; the rest map to user/assistant
+ * turns. Reads ANTHROPIC_API_KEY (same key the agent AnthropicLlmCaller uses).
+ */
+class AnthropicService extends baseAiGenProvider_service_1.BaseAiGenProviderService {
+    constructor() {
+        super();
+        this.client = new sdk_1.default({ apiKey: process.env.ANTHROPIC_API_KEY });
+    }
+    async generateText(params) {
+        const { input, modelKey, options } = params;
+        const model = aiModels_1.aiModelConfigs[modelKey];
+        const modelId = model?.modelId || "claude-haiku-4-5";
+        const { system, messages } = this.encodeTextInput(input);
+        // Anthropic requires max_tokens; fall back to the config default.
+        const maxTokens = (typeof options?.maxOutputTokens === "number"
+            ? options.maxOutputTokens
+            : undefined) ??
+            (typeof model?.fields?.maxOutputTokens?.default === "number"
+                ? model.fields.maxOutputTokens.default
+                : 4096);
+        const temperature = model?.fields?.temperature?.default
+            ? options?.temperature ?? model.fields.temperature.default
+            : undefined;
+        const msg = await this.client.messages.create({
+            model: modelId,
+            max_tokens: maxTokens,
+            messages,
+            ...(system ? { system } : {}),
+            ...(typeof temperature === "number" ? { temperature } : {}),
+        });
+        const text = msg.content
+            .filter((b) => b.type === "text")
+            .map((b) => b.text)
+            .join("")
+            .trim() || "No response";
+        return { text };
+    }
+    /**
+     * Map an OpenAI-style `input` (string | {role, content}[]) onto Anthropic's
+     * shape: a top-level `system` string + user/assistant MessageParam[]. Array
+     * `content` (vision parts) is flattened to its text segments — the text
+     * factory is text-only; image inputs go through the agent AnthropicLlmCaller.
+     */
+    encodeTextInput(input) {
+        if (typeof input === "string") {
+            return { messages: [{ role: "user", content: input }] };
+        }
+        const raw = Array.isArray(input)
+            ? input
+            : [];
+        let system;
+        const messages = [];
+        for (const m of raw) {
+            const text = typeof m.content === "string"
+                ? m.content
+                : Array.isArray(m.content)
+                    ? m.content
+                        .map((p) => p?.text ?? "")
+                        .join("")
+                    : String(m.content ?? "");
+            if (m.role === "system") {
+                system = system ? `${system}\n\n${text}` : text;
+                continue;
+            }
+            const role = m.role === "assistant" ? "assistant" : "user";
+            messages.push({ role, content: text });
+        }
+        // Anthropic rejects an empty messages array — guarantee at least one turn.
+        if (messages.length === 0) {
+            messages.push({ role: "user", content: system ?? "" });
+            if (system)
+                system = undefined;
+        }
+        return { system, messages };
+    }
+}
+exports.AnthropicService = AnthropicService;

package/lib/services/aiGen/providers/google/google.service.d.ts

@@ -1,9 +1,28 @@
 import { BaseAiGenProviderService } from "../baseAiGenProvider.service";
-import { CreditUsageParams, ImageGenerationParams, ImageGenerationResult, MusicGenerationParams, MusicGenerationResult, VideoGenerationParams, VideoGenerationResult, VideoStatusParams, VideoStatusResult } from "../types";
+import { CreditUsageParams, ImageGenerationParams, ImageGenerationResult, MusicGenerationParams, MusicGenerationResult, TextGenerationParams, TextGenerationResult, VideoGenerationParams, VideoGenerationResult, VideoStatusParams, VideoStatusResult } from "../types";
 export declare class GoogleService extends BaseAiGenProviderService {
     private ai;
+    private keyPool;
     private static readonly MAX_RETRY_ATTEMPTS;
     constructor();
+    /**
+     * Plain text generation via Gemini (`@google/genai` generateContent) for the
+     * aiGen text factory (LLM_NATIVE_MIGRATION_PLAN Phase 6). Accepts the same
+     * `input` shape as OpenaiService.generateText — a bare prompt string or an
+     * OpenAI-style message array (role + string content). System messages fold
+     * into Gemini's `systemInstruction`; assistant turns map to the "model" role.
+     * `options.responseFormat` of type json_object/json_schema flips on Gemini's
+     * native JSON mode (responseMimeType) so callers asking OpenAI for JSON get
+     * the same contract here.
+     */
+    generateText(params: TextGenerationParams): Promise<TextGenerationResult>;
+    /**
+     * Map an OpenAI-style `input` (string | {role, content}[]) onto Gemini's
+     * shape: merged systemInstruction + Content[] (assistant→"model"). Array
+     * `content` (vision parts) is flattened to its text segments — the text
+     * factory is text-only; image inputs go through the agent GeminiLlmCaller.
+     */
+    private encodeTextInput;
     /**
      * Retries `fn` on transient network errors (undici "fetch failed", ECONNRESET, ETIMEDOUT,
      * DNS hiccups, etc.) with exponential backoff: 1s → 2s → 4s. Non-network errors
@@ -21,6 +40,11 @@ export declare class GoogleService extends BaseAiGenProviderService {
      *
      * Required env: GCP_PROJECT_ID. Optional: GCP_LOCATION (default us-central1).
      */
+    /**
+     * Lyria 2 (Vertex AI) — delegates to the standalone generateGoogleMusic
+     * (Vertex predict endpoint via ADC, not the GoogleGenAI client). See
+     * ./googleMusic.
+     */
     generateMusic(params: MusicGenerationParams): Promise<MusicGenerationResult>;
     getCreditUsed({ modelKey, duration, resolution, multiClip, numImages, imageSize }: CreditUsageParams): number;
 }

package/lib/services/aiGen/providers/google/google.service.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"google.service.d.ts","sourceRoot":"","sources":["../../../../../src/services/aiGen/providers/google/google.service.ts"],"names":[],"mappings":"AAgBA,OAAO,EAAE,wBAAwB,EAAE,MAAM,8BAA8B,CAAC;AACxE,OAAO,EACL,iBAAiB,EACjB,qBAAqB,EACrB,qBAAqB,EACrB,qBAAqB,EACrB,qBAAqB,EACrB,qBAAqB,EACrB,qBAAqB,EACrB,iBAAiB,EACjB,iBAAiB,EAClB,MAAM,UAAU,CAAC;AA2MlB,qBAAa,aAAc,SAAQ,wBAAwB;IACzD,OAAO,CAAC,EAAE,CAAc;IACxB,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,kBAAkB,CAAK;;IAO/C;;;;OAIG;YACW,kBAAkB;IA+B1B,aAAa,CACjB,MAAM,EAAE,qBAAqB,GAC5B,OAAO,CAAC,qBAAqB,CAAC;IAyG3B,gBAAgB,CAAC,EACrB,IAAI,EACJ,cAAc,EACd,cAAyB,GAC1B,EAAE,iBAAiB,GAAG,OAAO,CAAC,iBAAiB,CAAC;IAoG3C,aAAa,CACjB,MAAM,EAAE,qBAAqB,GAC5B,OAAO,CAAC,qBAAqB,CAAC;YAiBnB,cAAc;IAwG5B;;;;;;OAMG;IACG,aAAa,CACjB,MAAM,EAAE,qBAAqB,GAC5B,OAAO,CAAC,qBAAqB,CAAC;IAqEjC,aAAa,CAAC,EAAE,QAAQ,EAAE,QAAY,EAAE,UAAmB,EAAE,SAAiB,EAAE,SAAa,EAAE,SAAS,EAAE,EAAE,iBAAiB,GAAG,MAAM;CA8BvI"}
+{"version":3,"file":"google.service.d.ts","sourceRoot":"","sources":["../../../../../src/services/aiGen/providers/google/google.service.ts"],"names":[],"mappings":"AAiBA,OAAO,EAAE,wBAAwB,EAAE,MAAM,8BAA8B,CAAC;AACxE,OAAO,EACL,iBAAiB,EACjB,qBAAqB,EACrB,qBAAqB,EACrB,qBAAqB,EACrB,qBAAqB,EACrB,oBAAoB,EACpB,oBAAoB,EACpB,qBAAqB,EACrB,qBAAqB,EACrB,iBAAiB,EACjB,iBAAiB,EAClB,MAAM,UAAU,CAAC;AAelB,qBAAa,aAAc,SAAQ,wBAAwB;IAKzD,OAAO,CAAC,EAAE,CAAc;IACxB,OAAO,CAAC,OAAO,CAAgB;IAC/B,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,kBAAkB,CAAK;;IAQ/C;;;;;;;;;OASG;IACG,YAAY,CAChB,MAAM,EAAE,oBAAoB,GAC3B,OAAO,CAAC,oBAAoB,CAAC;IAoChC;;;;;OAKG;IACH,OAAO,CAAC,eAAe;IAoCvB;;;;OAIG;YACW,kBAAkB;IA+B1B,aAAa,CACjB,MAAM,EAAE,qBAAqB,GAC5B,OAAO,CAAC,qBAAqB,CAAC;IAiJ3B,gBAAgB,CAAC,EACrB,IAAI,EACJ,cAAc,EACd,cAAyB,GAC1B,EAAE,iBAAiB,GAAG,OAAO,CAAC,iBAAiB,CAAC;IAyG3C,aAAa,CACjB,MAAM,EAAE,qBAAqB,GAC5B,OAAO,CAAC,qBAAqB,CAAC;YAiBnB,cAAc;IAwG5B;;;;;;OAMG;IACH;;;;OAIG;IACG,aAAa,CACjB,MAAM,EAAE,qBAAqB,GAC5B,OAAO,CAAC,qBAAqB,CAAC;IAIjC,aAAa,CAAC,EAAE,QAAQ,EAAE,QAAY,EAAE,UAAmB,EAAE,SAAiB,EAAE,SAAa,EAAE,SAAS,EAAE,EAAE,iBAAiB,GAAG,MAAM;CA8BvI"}