add speech job support

Author: wangzuo

src/index.ts

@@ -6,6 +6,7 @@ export * from "./jobs/embedding";
 export * from "./jobs/models";
 
 export * from "./providers/anthropic";
+export * from "./providers/elevenlabs";
 export * from "./providers/deepseek";
 export * from "./providers/fal";
 export * from "./providers/fireworks";

src/jobs/load.ts

@@ -1,6 +1,7 @@
 import { z } from "zod";
 import { anthropic, AnthropicJobSchema } from "~/providers/anthropic";
 import { deepseek, DeepseekJobSchema } from "~/providers/deepseek";
+import { elevenlabs, ElevenlabsJobSchema } from "~/providers/elevenlabs";
 import { fal, FalJobSchema } from "~/providers/fal";
 import { GoogleJobSchema } from "~/providers/google";
 import { LumaJobSchema } from "~/providers/luma";
@@ -11,6 +12,7 @@ import { voyage, VoyageJobSchema } from "~/providers/voyage";
 export const JobSchema = z.union([
   AnthropicJobSchema,
   DeepseekJobSchema,
+  ElevenlabsJobSchema,
   FalJobSchema,
   GoogleJobSchema,
   LumaJobSchema,
@@ -31,6 +33,8 @@ export function load(obj: Job) {
     provider = anthropic(obj.options);
   } else if (obj.provider === "deepseek") {
     provider = deepseek(obj.options);
+  } else if (obj.provider === "elevenlabs") {
+    provider = elevenlabs(obj.options);
   } else if (obj.provider === "fal") {
     provider = fal(obj.options);
   } else if (obj.provider === "ollama") {
@@ -59,6 +63,9 @@ export function load(obj: Job) {
   if (obj.type === "models" && "models" in provider) {
     builder = provider.models();
   }
+  if (obj.type === "speech" && "speech" in provider) {
+    builder = provider.speech(obj.input.model);
+  }
 
   if (!builder) {
     throw new Error("Failed to load job");

src/jobs/schema.ts

@@ -2,6 +2,7 @@ import { z } from "zod";
 
 export const JobProviderSchema = z.enum([
   "anthropic",
+  "elevenlabs",
   "deepseek",
   "fal",
   "google",
@@ -11,7 +12,13 @@ export const JobProviderSchema = z.enum([
   "voyage",
 ]);
 
-export const JobTypeSchema = z.enum(["chat", "image", "models", "embedding"]);
+export const JobTypeSchema = z.enum([
+  "chat",
+  "image",
+  "models",
+  "embedding",
+  "speech",
+]);
 
 export const JobOptionsSchema = z.object({
   apiKey: z.string().optional(),

src/jobs/speech/builder.ts

@@ -0,0 +1,14 @@
+import { JobBuilder } from "~/jobs/builder";
+import type { SpeechInput, SpeechOutput } from "./schema";
+
+export class SpeechJobBuilder extends JobBuilder<SpeechInput, SpeechOutput> {
+  input: SpeechInput;
+
+  constructor(model: string) {
+    super();
+    this.type = "speech";
+    this.input = {
+      model: model,
+    };
+  }
+}

src/jobs/speech/index.ts

@@ -0,0 +1,2 @@
+export * from "./schema";
+export * from "./builder";

src/jobs/speech/schema.ts

@@ -0,0 +1,16 @@
+import { z } from "zod";
+
+export const SpeechInputSchema = z.object({
+  model: z.string(),
+});
+
+export const SpeechOutputSchema = z.object({});
+
+export const SpeechJobSchema = z.object({
+  type: z.literal("speech"),
+  input: SpeechInputSchema,
+  output: SpeechOutputSchema.optional(),
+});
+
+export type SpeechInput = z.infer<typeof SpeechInputSchema>;
+export type SpeechOutput = z.infer<typeof SpeechOutputSchema>;

src/providers/elevenlabs/index.ts

@@ -0,0 +1,34 @@
+import type { JobOptions } from "~/jobs/schema";
+import { z } from "zod";
+import { SpeechJobBuilder, SpeechJobSchema } from "~/jobs/speech";
+
+export function elevenlabs(options?: JobOptions) {
+  options = options || {};
+  options.apiKey = options.apiKey || process.env.ELEVENLABS_API_KEY;
+
+  return {
+    speech(model: string) {
+      return new ElevenlabsSpeechJobBuilder(options, model);
+    },
+  };
+}
+
+class ElevenlabsSpeechJobBuilder extends SpeechJobBuilder {
+  constructor(options: JobOptions, model: string) {
+    super(model);
+    this.provider = "elevenlabs";
+    this.options = options;
+  }
+}
+
+export const ElevenlabsBaseJobSchema = z.object({
+  provider: z.literal("fal"),
+});
+
+export const ElevenlabsImageJobSchema = SpeechJobSchema.extend(
+  ElevenlabsBaseJobSchema
+);
+
+export const ElevenlabsJobSchema = z.discriminatedUnion("type", [
+  ElevenlabsImageJobSchema,
+]);

src/providers/fal/image.ts

@@ -1,5 +1,5 @@
-import { ImageJobBuilder } from "~/jobs/image";
 import type { JobOptions } from "~/jobs/schema";
+import { ImageJobBuilder } from "~/jobs/image";
 
 export class FalImageJobBuilder extends ImageJobBuilder {
   constructor(options: JobOptions, model: string) {

src/providers/luma/index.ts

@@ -15,10 +15,6 @@ export function luma(options?: JobOptions) {
   options = options || {};
   options.apiKey = options.apiKey || process.env.LUMA_API_KEY;
 
-  if (!options.apiKey) {
-    throw new Error("Luma API key is required");
-  }
-
   return {
     image(model: string) {
       return new LumaImageJobBuilder(options, model);

src/providers/openai/index.ts

@@ -3,15 +3,12 @@ import { OpenAIChatJobBuilder } from "~/providers/openai/chat";
 import { OpenAIImageJobBuilder } from "~/providers/openai/image";
 import { OpenAIEmbeddingJobBuilder } from "~/providers/openai/embedding";
 import { OpenAIModelsJobBuilder } from "~/providers/openai/models";
+import { OpenAISpeechJobBuilder } from "~/providers/openai/speech";
 
 export function openai(options?: JobOptions) {
   options = options || {};
   options.apiKey = options.apiKey || process.env.OPENAI_API_KEY;
 
-  if (!options.apiKey) {
-    throw new Error("OpenAI API key is required");
-  }
-
   return {
     chat(model: string) {
       return new OpenAIChatJobBuilder(options, model);
@@ -25,6 +22,9 @@ export function openai(options?: JobOptions) {
     models() {
       return new OpenAIModelsJobBuilder(options);
     },
+    speech(model: string) {
+      return new OpenAISpeechJobBuilder(options, model);
+    },
   };
 }