modelfusion 0.49.0 → 0.51.0

package/model-function/generate-text/streamText.cjs

@@ -7,7 +7,7 @@ function streamText(model, prompt, options) {
     let accumulatedText = "";
     let lastFullDelta;
     return new AsyncIterableResultPromise_js_1.AsyncIterableResultPromise((0, executeStreamCall_js_1.executeStreamCall)({
-        functionType: "text-streaming",
+        functionType: "stream-text",
         input: prompt,
         model,
         options,

package/model-function/generate-text/streamText.js

@@ -4,7 +4,7 @@ export function streamText(model, prompt, options) {
     let accumulatedText = "";
     let lastFullDelta;
     return new AsyncIterableResultPromise(executeStreamCall({
-        functionType: "text-streaming",
+        functionType: "stream-text",
         input: prompt,
         model,
         options,

package/model-function/generate-transcription/TranscriptionEvent.d.ts

@@ -1,6 +1,6 @@
 import { BaseModelCallFinishedEvent, BaseModelCallStartedEvent } from "../ModelCallEvent.js";
 export interface TranscriptionStartedEvent extends BaseModelCallStartedEvent {
-    functionType: "transcription";
+    functionType: "generate-transcription";
 }
 export type TranscriptionFinishedEventResult = {
     status: "success";
@@ -13,6 +13,6 @@ export type TranscriptionFinishedEventResult = {
     status: "abort";
 };
 export interface TranscriptionFinishedEvent extends BaseModelCallFinishedEvent {
-    functionType: "transcription";
+    functionType: "generate-transcription";
     result: TranscriptionFinishedEventResult;
 }

package/model-function/generate-transcription/generateTranscription.cjs

@@ -19,7 +19,7 @@ const ModelFunctionPromise_js_1 = require("../ModelFunctionPromise.cjs");
  */
 function generateTranscription(model, data, options) {
     return new ModelFunctionPromise_js_1.ModelFunctionPromise((0, executeStandardCall_js_1.executeStandardCall)({
-        functionType: "transcription",
+        functionType: "generate-transcription",
         input: data,
         model,
         options,

package/model-function/generate-transcription/generateTranscription.js

@@ -16,7 +16,7 @@ import { ModelFunctionPromise } from "../ModelFunctionPromise.js";
  */
 export function generateTranscription(model, data, options) {
     return new ModelFunctionPromise(executeStandardCall({
-        functionType: "transcription",
+        functionType: "generate-transcription",
         input: data,
         model,
         options,

package/model-provider/elevenlabs/ElevenLabsError.cjs

@@ -5,7 +5,6 @@ const ApiCallError_js_1 = require("../../core/api/ApiCallError.cjs");
 const failedElevenLabsCallResponseHandler = async ({ response, url, requestBodyValues }) => {
     const responseBody = await response.text();
     try {
-        // TODO implement ElevenLabsError
         return new ApiCallError_js_1.ApiCallError({
             message: responseBody,
             statusCode: response.status,

package/model-provider/elevenlabs/ElevenLabsError.js

@@ -2,7 +2,6 @@ import { ApiCallError } from "../../core/api/ApiCallError.js";
 export const failedElevenLabsCallResponseHandler = async ({ response, url, requestBodyValues }) => {
     const responseBody = await response.text();
     try {
-        // TODO implement ElevenLabsError
         return new ApiCallError({
             message: responseBody,
             statusCode: response.status,

package/model-provider/elevenlabs/ElevenLabsSpeechModel.cjs

@@ -15,11 +15,14 @@ const elevenLabsModels = [
     "eleven_multilingual_v1",
     "eleven_monolingual_v1",
 ];
-const defaultModel = "eleven_multilingual_v2";
+const defaultModel = "eleven_monolingual_v1";
 /**
  * Synthesize speech using the ElevenLabs Text to Speech API.
  *
- * @see https://api.elevenlabs.io/docs#/text-to-speech/Text_to_speech_v1_text_to_speech__voice_id__post
+ * Both regular text-to-speech and full duplex text-to-speech streaming are supported.
+ *
+ * @see https://docs.elevenlabs.io/api-reference/text-to-speech
+ * @see https://docs.elevenlabs.io/api-reference/text-to-speech-websockets
  */
 class ElevenLabsSpeechModel extends AbstractModel_js_1.AbstractModel {
     constructor(settings) {
@@ -84,7 +87,11 @@ class ElevenLabsSpeechModel extends AbstractModel_js_1.AbstractModel {
         ]);
         const queue = new AsyncQueue_js_1.AsyncQueue();
         const model = this.settings.model ?? defaultModel;
-        const socket = await (0, SimpleWebSocket_js_1.createSimpleWebSocket)(`wss://api.elevenlabs.io/v1/text-to-speech/${this.settings.voice}/stream-input?model_id=${model}`);
+        const socket = await (0, SimpleWebSocket_js_1.createSimpleWebSocket)(`wss://api.elevenlabs.io/v1/text-to-speech/${this.settings.voice}/stream-input${assembleQuery({
+            model_id: model,
+            optimize_streaming_latency: this.settings.optimizeStreamingLatency,
+            output_format: this.settings.outputFormat,
+        })}`);
         socket.onopen = async () => {
             const api = this.settings.api ?? new ElevenLabsApiConfiguration_js_1.ElevenLabsApiConfiguration();
             // send begin-of-stream (BOS) message:
@@ -158,9 +165,12 @@ class ElevenLabsSpeechModel extends AbstractModel_js_1.AbstractModel {
     }
 }
 exports.ElevenLabsSpeechModel = ElevenLabsSpeechModel;
-async function callElevenLabsTextToSpeechAPI({ api = new ElevenLabsApiConfiguration_js_1.ElevenLabsApiConfiguration(), abortSignal, text, voiceId, modelId, voiceSettings, }) {
+async function callElevenLabsTextToSpeechAPI({ api = new ElevenLabsApiConfiguration_js_1.ElevenLabsApiConfiguration(), abortSignal, text, voiceId, modelId, optimizeStreamingLatency, outputFormat, voiceSettings, }) {
     return (0, postToApi_js_1.postJsonToApi)({
-        url: api.assembleUrl(`/text-to-speech/${voiceId}`),
+        url: api.assembleUrl(`/text-to-speech/${voiceId}${assembleQuery({
+            optimize_streaming_latency: optimizeStreamingLatency,
+            output_format: outputFormat,
+        })}`),
         headers: api.headers,
         body: {
             text,
@@ -172,6 +182,24 @@ async function callElevenLabsTextToSpeechAPI({ api = new ElevenLabsApiConfigurat
         abortSignal,
     });
 }
+function assembleQuery(parameters) {
+    let query = "";
+    let hasQuestionMark = false;
+    for (const [key, value] of Object.entries(parameters)) {
+        if (value == null) {
+            continue;
+        }
+        if (!hasQuestionMark) {
+            query += "?";
+            hasQuestionMark = true;
+        }
+        else {
+            query += "&";
+        }
+        query += `${key}=${value}`;
+    }
+    return query;
+}
 function toApiVoiceSettings(voiceSettings) {
     return voiceSettings != null
         ? {

package/model-provider/elevenlabs/ElevenLabsSpeechModel.d.ts

@@ -11,6 +11,8 @@ export interface ElevenLabsSpeechModelSettings extends SpeechGenerationModelSett
     };
     voice: string;
     model?: (typeof elevenLabsModels)[number] | (string & {});
+    optimizeStreamingLatency?: 0 | 1 | 2 | 3 | 4;
+    outputFormat?: "mp3_44100" | "pcm_16000" | "pcm_22050" | "pcm_24000" | "pcm_44100";
     voiceSettings?: {
         stability: number;
         similarityBoost: number;
@@ -24,7 +26,10 @@ export interface ElevenLabsSpeechModelSettings extends SpeechGenerationModelSett
 /**
  * Synthesize speech using the ElevenLabs Text to Speech API.
  *
- * @see https://api.elevenlabs.io/docs#/text-to-speech/Text_to_speech_v1_text_to_speech__voice_id__post
+ * Both regular text-to-speech and full duplex text-to-speech streaming are supported.
+ *
+ * @see https://docs.elevenlabs.io/api-reference/text-to-speech
+ * @see https://docs.elevenlabs.io/api-reference/text-to-speech-websockets
  */
 export declare class ElevenLabsSpeechModel extends AbstractModel<ElevenLabsSpeechModelSettings> implements StreamingSpeechGenerationModel<ElevenLabsSpeechModelSettings> {
     constructor(settings: ElevenLabsSpeechModelSettings);

package/model-provider/elevenlabs/ElevenLabsSpeechModel.js

@@ -12,11 +12,14 @@ const elevenLabsModels = [
     "eleven_multilingual_v1",
     "eleven_monolingual_v1",
 ];
-const defaultModel = "eleven_multilingual_v2";
+const defaultModel = "eleven_monolingual_v1";
 /**
  * Synthesize speech using the ElevenLabs Text to Speech API.
  *
- * @see https://api.elevenlabs.io/docs#/text-to-speech/Text_to_speech_v1_text_to_speech__voice_id__post
+ * Both regular text-to-speech and full duplex text-to-speech streaming are supported.
+ *
+ * @see https://docs.elevenlabs.io/api-reference/text-to-speech
+ * @see https://docs.elevenlabs.io/api-reference/text-to-speech-websockets
  */
 export class ElevenLabsSpeechModel extends AbstractModel {
     constructor(settings) {
@@ -81,7 +84,11 @@ export class ElevenLabsSpeechModel extends AbstractModel {
         ]);
         const queue = new AsyncQueue();
         const model = this.settings.model ?? defaultModel;
-        const socket = await createSimpleWebSocket(`wss://api.elevenlabs.io/v1/text-to-speech/${this.settings.voice}/stream-input?model_id=${model}`);
+        const socket = await createSimpleWebSocket(`wss://api.elevenlabs.io/v1/text-to-speech/${this.settings.voice}/stream-input${assembleQuery({
+            model_id: model,
+            optimize_streaming_latency: this.settings.optimizeStreamingLatency,
+            output_format: this.settings.outputFormat,
+        })}`);
         socket.onopen = async () => {
             const api = this.settings.api ?? new ElevenLabsApiConfiguration();
             // send begin-of-stream (BOS) message:
@@ -154,9 +161,12 @@ export class ElevenLabsSpeechModel extends AbstractModel {
         });
     }
 }
-async function callElevenLabsTextToSpeechAPI({ api = new ElevenLabsApiConfiguration(), abortSignal, text, voiceId, modelId, voiceSettings, }) {
+async function callElevenLabsTextToSpeechAPI({ api = new ElevenLabsApiConfiguration(), abortSignal, text, voiceId, modelId, optimizeStreamingLatency, outputFormat, voiceSettings, }) {
     return postJsonToApi({
-        url: api.assembleUrl(`/text-to-speech/${voiceId}`),
+        url: api.assembleUrl(`/text-to-speech/${voiceId}${assembleQuery({
+            optimize_streaming_latency: optimizeStreamingLatency,
+            output_format: outputFormat,
+        })}`),
         headers: api.headers,
         body: {
             text,
@@ -168,6 +178,24 @@ async function callElevenLabsTextToSpeechAPI({ api = new ElevenLabsApiConfigurat
         abortSignal,
     });
 }
+function assembleQuery(parameters) {
+    let query = "";
+    let hasQuestionMark = false;
+    for (const [key, value] of Object.entries(parameters)) {
+        if (value == null) {
+            continue;
+        }
+        if (!hasQuestionMark) {
+            query += "?";
+            hasQuestionMark = true;
+        }
+        else {
+            query += "&";
+        }
+        query += `${key}=${value}`;
+    }
+    return query;
+}
 function toApiVoiceSettings(voiceSettings) {
     return voiceSettings != null
         ? {

package/model-provider/lmnt/LmntError.cjs

@@ -5,7 +5,6 @@ const ApiCallError_js_1 = require("../../core/api/ApiCallError.cjs");
 const failedLmntCallResponseHandler = async ({ response, url, requestBodyValues }) => {
     const responseBody = await response.text();
     try {
-        // TODO implement LmntError
         return new ApiCallError_js_1.ApiCallError({
             message: responseBody,
             statusCode: response.status,

package/model-provider/lmnt/LmntError.js

@@ -2,7 +2,6 @@ import { ApiCallError } from "../../core/api/ApiCallError.js";
 export const failedLmntCallResponseHandler = async ({ response, url, requestBodyValues }) => {
     const responseBody = await response.text();
     try {
-        // TODO implement LmntError
         return new ApiCallError({
             message: responseBody,
             statusCode: response.status,

package/model-provider/openai/OpenAICompletionModel.d.ts

@@ -148,12 +148,12 @@ export declare class OpenAICompletionModel extends AbstractModel<OpenAICompletio
     doGenerateText(prompt: string, options?: FunctionOptions): Promise<{
         response: {
             object: "text_completion";
-            model: string;
             usage: {
                 prompt_tokens: number;
                 completion_tokens: number;
                 total_tokens: number;
             };
+            model: string;
             id: string;
             created: number;
             choices: {
@@ -221,12 +221,12 @@ declare const OpenAICompletionResponseSchema: z.ZodObject<{
     }>;
 }, "strip", z.ZodTypeAny, {
     object: "text_completion";
-    model: string;
     usage: {
         prompt_tokens: number;
         completion_tokens: number;
         total_tokens: number;
     };
+    model: string;
     id: string;
     created: number;
     choices: {
@@ -237,12 +237,12 @@ declare const OpenAICompletionResponseSchema: z.ZodObject<{
     }[];
 }, {
     object: "text_completion";
-    model: string;
     usage: {
         prompt_tokens: number;
         completion_tokens: number;
         total_tokens: number;
     };
+    model: string;
     id: string;
     created: number;
     choices: {
@@ -265,12 +265,12 @@ export declare const OpenAITextResponseFormat: {
         stream: false;
         handler: ResponseHandler<{
             object: "text_completion";
-            model: string;
             usage: {
                 prompt_tokens: number;
                 completion_tokens: number;
                 total_tokens: number;
             };
+            model: string;
             id: string;
             created: number;
             choices: {

package/model-provider/openai/OpenAICostCalculator.cjs

@@ -19,12 +19,12 @@ class OpenAICostCalculator {
         const type = call.functionType;
         const model = call.model.modelName;
         switch (type) {
-            case "image-generation": {
+            case "generate-image": {
                 return (0, OpenAIImageGenerationModel_js_1.calculateOpenAIImageGenerationCostInMillicents)({
                     settings: call.settings,
                 });
             }
-            case "embedding": {
+            case "embed": {
                 if (model == null) {
                     return null;
                 }
@@ -39,8 +39,8 @@ class OpenAICostCalculator {
                 }
                 break;
             }
-            case "structure-generation":
-            case "text-generation": {
+            case "generate-structure":
+            case "generate-text": {
                 if (model == null) {
                     return null;
                 }
@@ -58,7 +58,7 @@ class OpenAICostCalculator {
                 }
                 break;
             }
-            case "transcription": {
+            case "generate-transcription": {
                 if (model == null) {
                     return null;
                 }

package/model-provider/openai/OpenAICostCalculator.js

@@ -16,12 +16,12 @@ export class OpenAICostCalculator {
         const type = call.functionType;
         const model = call.model.modelName;
         switch (type) {
-            case "image-generation": {
+            case "generate-image": {
                 return calculateOpenAIImageGenerationCostInMillicents({
                     settings: call.settings,
                 });
             }
-            case "embedding": {
+            case "embed": {
                 if (model == null) {
                     return null;
                 }
@@ -36,8 +36,8 @@ export class OpenAICostCalculator {
                 }
                 break;
             }
-            case "structure-generation":
-            case "text-generation": {
+            case "generate-structure":
+            case "generate-text": {
                 if (model == null) {
                     return null;
                 }
@@ -55,7 +55,7 @@ export class OpenAICostCalculator {
                 }
                 break;
             }
-            case "transcription": {
+            case "generate-transcription": {
                 if (model == null) {
                     return null;
                 }

package/model-provider/openai/OpenAITextEmbeddingModel.d.ts

@@ -50,11 +50,11 @@ export declare class OpenAITextEmbeddingModel extends AbstractModel<OpenAITextEm
     doEmbedValues(texts: string[], options?: FunctionOptions): Promise<{
         response: {
             object: "list";
-            model: string;
             usage: {
                 prompt_tokens: number;
                 total_tokens: number;
             };
+            model: string;
             data: {
                 object: "embedding";
                 embedding: number[];
@@ -93,11 +93,11 @@ declare const openAITextEmbeddingResponseSchema: z.ZodObject<{
     }>;
 }, "strip", z.ZodTypeAny, {
     object: "list";
-    model: string;
     usage: {
         prompt_tokens: number;
         total_tokens: number;
     };
+    model: string;
     data: {
         object: "embedding";
         embedding: number[];
@@ -105,11 +105,11 @@ declare const openAITextEmbeddingResponseSchema: z.ZodObject<{
     }[];
 }, {
     object: "list";
-    model: string;
     usage: {
         prompt_tokens: number;
         total_tokens: number;
     };
+    model: string;
     data: {
         object: "embedding";
         embedding: number[];

package/model-provider/openai/chat/OpenAIChatModel.d.ts

@@ -158,12 +158,12 @@ export declare class OpenAIChatModel extends AbstractModel<OpenAIChatSettings> i
     doGenerateText(prompt: OpenAIChatMessage[], options?: FunctionOptions): Promise<{
         response: {
             object: "chat.completion";
-            model: string;
             usage: {
                 prompt_tokens: number;
                 completion_tokens: number;
                 total_tokens: number;
             };
+            model: string;
             id: string;
             created: number;
             choices: {
@@ -198,12 +198,12 @@ export declare class OpenAIChatModel extends AbstractModel<OpenAIChatSettings> i
     doGenerateStructure(structureDefinition: StructureDefinition<string, unknown>, prompt: OpenAIChatMessage[], options?: FunctionOptions): Promise<{
         response: {
             object: "chat.completion";
-            model: string;
             usage: {
                 prompt_tokens: number;
                 completion_tokens: number;
                 total_tokens: number;
             };
+            model: string;
             id: string;
             created: number;
             choices: {
@@ -232,12 +232,12 @@ export declare class OpenAIChatModel extends AbstractModel<OpenAIChatSettings> i
     doGenerateStructureOrText(structureDefinitions: Array<StructureDefinition<string, unknown>>, prompt: OpenAIChatMessage[], options?: FunctionOptions): Promise<{
         response: {
             object: "chat.completion";
-            model: string;
             usage: {
                 prompt_tokens: number;
                 completion_tokens: number;
                 total_tokens: number;
             };
+            model: string;
             id: string;
             created: number;
             choices: {
@@ -268,12 +268,12 @@ export declare class OpenAIChatModel extends AbstractModel<OpenAIChatSettings> i
     } | {
         response: {
             object: "chat.completion";
-            model: string;
             usage: {
                 prompt_tokens: number;
                 completion_tokens: number;
                 total_tokens: number;
             };
+            model: string;
             id: string;
             created: number;
             choices: {
@@ -395,12 +395,12 @@ declare const openAIChatResponseSchema: z.ZodObject<{
     }>;
 }, "strip", z.ZodTypeAny, {
     object: "chat.completion";
-    model: string;
     usage: {
         prompt_tokens: number;
         completion_tokens: number;
         total_tokens: number;
     };
+    model: string;
     id: string;
     created: number;
     choices: {
@@ -418,12 +418,12 @@ declare const openAIChatResponseSchema: z.ZodObject<{
     }[];
 }, {
     object: "chat.completion";
-    model: string;
     usage: {
         prompt_tokens: number;
         completion_tokens: number;
         total_tokens: number;
     };
+    model: string;
     id: string;
     created: number;
     choices: {
@@ -453,12 +453,12 @@ export declare const OpenAIChatResponseFormat: {
         stream: false;
         handler: ResponseHandler<{
             object: "chat.completion";
-            model: string;
             usage: {
                 prompt_tokens: number;
                 completion_tokens: number;
                 total_tokens: number;
             };
+            model: string;
             id: string;
             created: number;
             choices: {

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "modelfusion",
   "description": "Build multimodal applications, chatbots, and agents with JavaScript and TypeScript.",
-  "version": "0.49.0",
+  "version": "0.51.0",
   "author": "Lars Grammel",
   "license": "MIT",
   "keywords": [

package/retriever/retrieve.cjs

@@ -1,83 +1,15 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.retrieve = void 0;
-const nanoid_1 = require("nanoid");
-const FunctionEventSource_js_1 = require("../core/FunctionEventSource.cjs");
-const GlobalFunctionLogging_js_1 = require("../core/GlobalFunctionLogging.cjs");
-const GlobalFunctionObservers_js_1 = require("../core/GlobalFunctionObservers.cjs");
-const AbortError_js_1 = require("../core/api/AbortError.cjs");
-const getFunctionCallLogger_js_1 = require("../core/getFunctionCallLogger.cjs");
-const getRun_js_1 = require("../core/getRun.cjs");
-const DurationMeasurement_js_1 = require("../util/DurationMeasurement.cjs");
-const runSafe_js_1 = require("../util/runSafe.cjs");
+const executeFunctionCall_js_1 = require("../core/executeFunctionCall.cjs");
 async function retrieve(retriever, query, options) {
-    const run = await (0, getRun_js_1.getRun)(options?.run);
-    const eventSource = new FunctionEventSource_js_1.FunctionEventSource({
-        observers: [
-            ...(0, getFunctionCallLogger_js_1.getFunctionCallLogger)(options?.logging ?? (0, GlobalFunctionLogging_js_1.getGlobalFunctionLogging)()),
-            ...(0, GlobalFunctionObservers_js_1.getGlobalFunctionObservers)(),
-            ...(run?.functionObserver != null ? [run.functionObserver] : []),
-            ...(options?.observers ?? []),
-        ],
-        errorHandler: run?.errorHandler,
-    });
-    const durationMeasurement = (0, DurationMeasurement_js_1.startDurationMeasurement)();
-    const startMetadata = {
+    return (0, executeFunctionCall_js_1.executeFunctionCall)({
+        options,
+        input: query,
         functionType: "retrieve",
-        callId: `call-${(0, nanoid_1.nanoid)()}`,
-        runId: run?.runId,
-        sessionId: run?.sessionId,
-        userId: run?.userId,
-        functionId: options?.functionId,
-        query,
-        timestamp: durationMeasurement.startDate,
-        startTimestamp: durationMeasurement.startDate,
-    };
-    eventSource.notify({
-        eventType: "started",
-        ...startMetadata,
-    });
-    const result = await (0, runSafe_js_1.runSafe)(() => retriever.retrieve(query, {
-        functionId: options?.functionId,
-        logging: options?.logging,
-        observers: options?.observers,
-        run,
-    }));
-    const finishMetadata = {
-        eventType: "finished",
-        ...startMetadata,
-        finishTimestamp: new Date(),
-        durationInMs: durationMeasurement.durationInMs,
-    };
-    if (!result.ok) {
-        if (result.isAborted) {
-            eventSource.notify({
-                ...finishMetadata,
-                eventType: "finished",
-                result: {
-                    status: "abort",
-                },
-            });
-            throw new AbortError_js_1.AbortError();
-        }
-        eventSource.notify({
-            ...finishMetadata,
-            eventType: "finished",
-            result: {
-                status: "error",
-                error: result.error,
-            },
-        });
-        throw result.error;
-    }
-    eventSource.notify({
-        ...finishMetadata,
-        eventType: "finished",
-        result: {
-            status: "success",
-            value: result.value,
-        },
+        execute: (options) => retriever.retrieve(query, options),
+        inputPropertyName: "query",
+        outputPropertyName: "results",
     });
-    return result.value;
 }
 exports.retrieve = retrieve;