modelfusion 0.5.0 → 0.7.0

package/README.md

@@ -10,9 +10,8 @@
 
 [Introduction](#introduction) | [Quick Install](#quick-install) | [Usage](#usage-examples) | [Features](#features) | [Integrations](#integrations) | [Documentation](#documentation) | [Examples](#more-examples) | [modelfusion.dev](https://modelfusion.dev)
 
-## Disclaimer
-
-ModelFusion is in its initial development phase. Until version 1.0 there may be breaking changes.
+> [!NOTE]
+> ModelFusion is in its initial development phase. Until version 1.0 there may be breaking changes, because we are still exploring the API design. We welcome your feedback and suggestions.
 
 ## Introduction
 
@@ -49,7 +48,7 @@ You can use [prompt mappings](https://modelfusion.dev/guide/function/generate-te
 #### generateText
 
 ```ts
-const { text } = await generateText(
+const text = await generateText(
   new OpenAITextGenerationModel({ model: "text-davinci-003" }),
   "Write a short story about a robot learning to love:\n\n"
 );
@@ -58,7 +57,7 @@ const { text } = await generateText(
 #### streamText
 
 ```ts
-const { textStream } = await streamText(
+const textStream = await streamText(
   new OpenAIChatModel({ model: "gpt-3.5-turbo", maxTokens: 1000 }),
   [
     OpenAIChatMessage.system("You are a story writer."),
@@ -76,7 +75,7 @@ for await (const textFragment of textStream) {
 [Prompt mapping](https://modelfusion.dev/guide/function/generate-text/prompt-mapping) lets you use higher level prompt structures (such as instruction or chat prompts) for different models.
 
 ```ts
-const { text } = await generateText(
+const text = await generateText(
   new LlamaCppTextGenerationModel({
     contextWindowSize: 4096, // Llama 2 context window size
     nPredict: 1000,
@@ -89,7 +88,7 @@ const { text } = await generateText(
 ```
 
 ```ts
-const { textStream } = await streamText(
+const textStream = await streamText(
   new OpenAIChatModel({
     model: "gpt-3.5-turbo",
   }).mapPrompt(ChatToOpenAIChatPromptMapping()),
@@ -104,14 +103,15 @@ const { textStream } = await streamText(
 
 #### Metadata and original responses
 
-Most ModelFusion model functions return rich results that include the original response and metadata.
+ModelFusion model functions return rich results that include the original response and metadata when you set the `fullResponse` option to `true`.
 
 ```ts
 const { text, response, metadata } = await generateText(
   new OpenAITextGenerationModel({
     model: "text-davinci-003",
   }),
-  "Write a short story about a robot learning to love:\n\n"
+  "Write a short story about a robot learning to love:\n\n",
+  { fullResponse: true }
 );
 ```
 
@@ -120,7 +120,7 @@ const { text, response, metadata } = await generateText(
 Generate JSON value that matches a schema.
 
 ```ts
-const { value } = await generateJson(
+const value = await generateJson(
   new OpenAIChatModel({
     model: "gpt-3.5-turbo",
     temperature: 0,
@@ -249,7 +249,7 @@ const { tool, parameters, result, text } = await useToolOrGenerateText(
 Turn audio (voice) into text.
 
 ```ts
-const { transcription } = await transcribe(
+const transcription = await transcribe(
   new OpenAITranscriptionModel({ model: "whisper-1" }),
   {
     type: "mp3",
@@ -263,7 +263,7 @@ const { transcription } = await transcribe(
 Generate a base64-encoded image from a prompt.
 
 ```ts
-const { image } = await generateImage(
+const image = await generateImage(
   new OpenAIImageGenerationModel({ size: "512x512" }),
   "the wicked witch of the west in the style of early 19th century painting"
 );
@@ -274,7 +274,7 @@ const { image } = await generateImage(
 Create embeddings for text. Embeddings are vectors that represent the meaning of the text.
 
 ```ts
-const { embeddings } = await embedTexts(
+const embeddings = await embedTexts(
   new OpenAITextEmbeddingModel({ model: "text-embedding-ada-002" }),
   [
     "At first, Nox didn't know what to do with the pup.",
@@ -343,9 +343,9 @@ const { chunks } = await retrieveTextChunks(
   - [Transcribe Audio](https://modelfusion.dev/guide/function/transcribe-audio)
   - [Generate images](https://modelfusion.dev/guide/function/generate-image)
 - Summarize text
-- Split text
 - [Tools](https://modelfusion.dev/guide/tools)
-- [Text Chunks](https://modelfusion.dev/guide/text-chunks)
+- [Text Chunks](https://modelfusion.dev/guide/text-chunk/)
+  - [Split Text](https://modelfusion.dev/guide/text-chunk/split)
 - [Run abstraction](https://modelfusion.dev/guide/run/)
   - [Abort signals](https://modelfusion.dev/guide/run/abort)
   - [Cost calculation](https://modelfusion.dev/guide/run/cost-calculation)
@@ -399,6 +399,7 @@ Use higher level prompts that are mapped into model specific prompt formats.
 - [Examples & Tutorials](https://modelfusion.dev/tutorial)
 - [Integrations](https://modelfusion.dev/integration/model-provider)
 - [API Reference](https://modelfusion.dev/api/modules)
+- [Blog](https://modelfusion.dev/api/blog)
 
 ## More Examples
 
@@ -416,6 +417,12 @@ Examples for the individual functions and objects.
 
 A web chat with an AI assistant, implemented as a Next.js app.
 
+### [Chat with PDF](https://github.com/lgrammel/modelfusion/tree/main/examples/pdf-chat-terminal)
+
+> _terminal app_, _PDF parsing_, _in memory vector indices_, _retrieval augmented generation_, _hypothetical document embedding_
+
+Ask questions about a PDF document and get answers from the document.
+
 ### [Image generator (Next.js)](https://github.com/lgrammel/modelfusion/tree/main/examples/image-generator-next-js)
 
 > _Next.js app_, _Stability AI image generation_
@@ -440,12 +447,6 @@ TypeScript implementation of the classic [BabyAGI](https://github.com/yoheinakaj
 
 Small agent that solves middle school math problems. It uses a calculator tool to solve the problems.
 
-### [Chat with PDF](https://github.com/lgrammel/modelfusion/tree/main/examples/pdf-chat-terminal)
-
-> _terminal app_, _PDF parsing_, _in memory vector indices_, _retrieval augmented generation_, _hypothetical document embedding_
-
-Ask questions about a PDF document and get answers from the document.
-
 ### [PDF to Tweet](https://github.com/lgrammel/modelfusion/tree/main/examples/pdf-to-tweet)
 
 > _terminal app_, _PDF parsing_, _recursive information extraction_, _in memory vector index, \_style example retrieval_, _OpenAI GPT-4_, _cost calculation_

package/composed-function/summarize/summarizeRecursivelyWithTextGenerationAndTokenSplitting.cjs

@@ -13,12 +13,12 @@ async function summarizeRecursivelyWithTextGenerationAndTokenSplitting({ text, m
     (model.maxCompletionTokens ?? model.contextWindowSize / 4), join, }, options) {
     const emptyPromptTokens = await model.countPromptTokens(await prompt({ text: "" }));
     return (0, summarizeRecursively_js_1.summarizeRecursively)({
-        split: (0, splitRecursively_js_1.splitRecursivelyAtToken)({
+        split: (0, splitRecursively_js_1.splitAtToken)({
             tokenizer: model.tokenizer,
-            maxChunkSize: tokenLimit - emptyPromptTokens,
+            maxTokensPerChunk: tokenLimit - emptyPromptTokens,
         }),
         summarize: async (input) => {
-            const { text } = await (0, generateText_js_1.generateText)(model, await prompt(input), options);
+            const text = await (0, generateText_js_1.generateText)(model, await prompt(input), options);
             return text;
         },
         join,

package/composed-function/summarize/summarizeRecursivelyWithTextGenerationAndTokenSplitting.js

@@ -1,5 +1,5 @@
 import { generateText } from "../../model-function/generate-text/generateText.js";
-import { splitRecursivelyAtToken } from "../../text-chunk/split/splitRecursively.js";
+import { splitAtToken } from "../../text-chunk/split/splitRecursively.js";
 import { summarizeRecursively } from "./summarizeRecursively.js";
 /**
  * Recursively summarizes a text using a text generation model, e.g. for summarization or text extraction.
@@ -10,12 +10,12 @@ export async function summarizeRecursivelyWithTextGenerationAndTokenSplitting({
     (model.maxCompletionTokens ?? model.contextWindowSize / 4), join, }, options) {
     const emptyPromptTokens = await model.countPromptTokens(await prompt({ text: "" }));
     return summarizeRecursively({
-        split: splitRecursivelyAtToken({
+        split: splitAtToken({
             tokenizer: model.tokenizer,
-            maxChunkSize: tokenLimit - emptyPromptTokens,
+            maxTokensPerChunk: tokenLimit - emptyPromptTokens,
         }),
         summarize: async (input) => {
-            const { text } = await generateText(model, await prompt(input), options);
+            const text = await generateText(model, await prompt(input), options);
             return text;
         },
         join,

package/composed-function/use-tool/useTool.cjs

@@ -19,7 +19,10 @@ async function useTool(model, tool, prompt, options) {
         name: tool.name,
         description: tool.description,
         schema: tool.inputSchema,
-    }, () => prompt(tool), options);
+    }, () => prompt(tool), {
+        ...(options ?? {}),
+        fullResponse: true,
+    });
     return {
         tool: tool.name,
         parameters: value,

package/composed-function/use-tool/useTool.js

@@ -16,7 +16,10 @@ export async function useTool(model, tool, prompt, options) {
         name: tool.name,
         description: tool.description,
         schema: tool.inputSchema,
-    }, () => prompt(tool), options);
+    }, () => prompt(tool), {
+        ...(options ?? {}),
+        fullResponse: true,
+    });
     return {
         tool: tool.name,
         parameters: value,

package/model-function/embed-text/embedText.cjs

@@ -2,18 +2,6 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.embedText = exports.embedTexts = void 0;
 const executeCall_js_1 = require("../executeCall.cjs");
-/**
- * Generate embeddings for multiple texts.
- *
- * @example
- * const { embeddings } = await embedTexts(
- *   new OpenAITextEmbeddingModel(...),
- *   [
- *     "At first, Nox didn't know what to do with the pup.",
- *     "He keenly observed and absorbed everything around him, from the birds in the sky to the trees in the forest.",
- *   ]
- * );
- */
 async function embedTexts(model, texts, options) {
     const result = await (0, executeCall_js_1.executeCall)({
         model,
@@ -65,26 +53,24 @@ async function embedTexts(model, texts, options) {
             generatedEmbeddings: output,
         }),
     });
-    return {
-        embeddings: result.output,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            embeddings: result.output,
+            metadata: result.metadata,
+        }
+        : result.output;
 }
 exports.embedTexts = embedTexts;
-/**
- * Generate an embedding for a single text.
- *
- * @example
- * const { embedding } = await embedText(
- *   new OpenAITextEmbeddingModel(...),
- *   "At first, Nox didn't know what to do with the pup."
- * );
- */
 async function embedText(model, text, options) {
-    const result = await embedTexts(model, [text], options);
-    return {
-        embedding: result.embeddings[0],
-        metadata: result.metadata,
-    };
+    const result = await embedTexts(model, [text], {
+        ...(options ?? {}),
+        fullResponse: true,
+    });
+    return options?.fullResponse === true
+        ? {
+            embedding: result.embeddings[0],
+            metadata: result.metadata,
+        }
+        : result.embeddings[0];
 }
 exports.embedText = embedText;

package/model-function/embed-text/embedText.d.ts

@@ -6,7 +6,7 @@ import { TextEmbeddingModel, TextEmbeddingModelSettings } from "./TextEmbeddingM
  * Generate embeddings for multiple texts.
  *
  * @example
- * const { embeddings } = await embedTexts(
+ * const embeddings = await embedTexts(
  *   new OpenAITextEmbeddingModel(...),
  *   [
  *     "At first, Nox didn't know what to do with the pup.",
@@ -14,20 +14,30 @@ import { TextEmbeddingModel, TextEmbeddingModelSettings } from "./TextEmbeddingM
  *   ]
  * );
  */
-export declare function embedTexts<RESPONSE, SETTINGS extends TextEmbeddingModelSettings>(model: TextEmbeddingModel<RESPONSE, SETTINGS>, texts: string[], options?: FunctionOptions<SETTINGS>): Promise<{
+export declare function embedTexts<RESPONSE, SETTINGS extends TextEmbeddingModelSettings>(model: TextEmbeddingModel<RESPONSE, SETTINGS>, texts: string[], options: FunctionOptions<SETTINGS> & {
+    fullResponse: true;
+}): Promise<{
     embeddings: Array<Vector>;
     metadata: CallMetadata<TextEmbeddingModel<RESPONSE, SETTINGS>>;
 }>;
+export declare function embedTexts<RESPONSE, SETTINGS extends TextEmbeddingModelSettings>(model: TextEmbeddingModel<RESPONSE, SETTINGS>, texts: string[], options?: FunctionOptions<SETTINGS> & {
+    fullResponse?: false;
+}): Promise<Array<Vector>>;
 /**
  * Generate an embedding for a single text.
  *
  * @example
- * const { embedding } = await embedText(
+ * const embedding = await embedText(
  *   new OpenAITextEmbeddingModel(...),
  *   "At first, Nox didn't know what to do with the pup."
  * );
  */
-export declare function embedText<RESPONSE, SETTINGS extends TextEmbeddingModelSettings>(model: TextEmbeddingModel<RESPONSE, SETTINGS>, text: string, options?: FunctionOptions<SETTINGS>): Promise<{
+export declare function embedText<RESPONSE, SETTINGS extends TextEmbeddingModelSettings>(model: TextEmbeddingModel<RESPONSE, SETTINGS>, text: string, options: FunctionOptions<SETTINGS> & {
+    fullResponse: true;
+}): Promise<{
     embedding: Vector;
     metadata: CallMetadata<TextEmbeddingModel<RESPONSE, SETTINGS>>;
 }>;
+export declare function embedText<RESPONSE, SETTINGS extends TextEmbeddingModelSettings>(model: TextEmbeddingModel<RESPONSE, SETTINGS>, text: string, options?: FunctionOptions<SETTINGS> & {
+    fullResponse?: false;
+}): Promise<Vector>;

package/model-function/embed-text/embedText.js

@@ -1,16 +1,4 @@
 import { executeCall } from "../executeCall.js";
-/**
- * Generate embeddings for multiple texts.
- *
- * @example
- * const { embeddings } = await embedTexts(
- *   new OpenAITextEmbeddingModel(...),
- *   [
- *     "At first, Nox didn't know what to do with the pup.",
- *     "He keenly observed and absorbed everything around him, from the birds in the sky to the trees in the forest.",
- *   ]
- * );
- */
 export async function embedTexts(model, texts, options) {
     const result = await executeCall({
         model,
@@ -62,24 +50,22 @@ export async function embedTexts(model, texts, options) {
             generatedEmbeddings: output,
         }),
     });
-    return {
-        embeddings: result.output,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            embeddings: result.output,
+            metadata: result.metadata,
+        }
+        : result.output;
 }
-/**
- * Generate an embedding for a single text.
- *
- * @example
- * const { embedding } = await embedText(
- *   new OpenAITextEmbeddingModel(...),
- *   "At first, Nox didn't know what to do with the pup."
- * );
- */
 export async function embedText(model, text, options) {
-    const result = await embedTexts(model, [text], options);
-    return {
-        embedding: result.embeddings[0],
-        metadata: result.metadata,
-    };
+    const result = await embedTexts(model, [text], {
+        ...(options ?? {}),
+        fullResponse: true,
+    });
+    return options?.fullResponse === true
+        ? {
+            embedding: result.embeddings[0],
+            metadata: result.metadata,
+        }
+        : result.embeddings[0];
 }

package/model-function/generate-image/generateImage.cjs

@@ -2,21 +2,6 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.generateImage = void 0;
 const executeCall_js_1 = require("../executeCall.cjs");
-/**
- * Generates a base64-encoded image using a prompt.
- * The prompt format depends on the model.
- * For example, OpenAI image models expect a string prompt,
- * and Stability AI models expect an array of text prompts with optional weights.
- *
- * @example
- * const { image } = await generateImage(
- *   new StabilityImageGenerationModel(...),
- *   [
- *     { text: "the wicked witch of the west" },
- *     { text: "style of early 19th century painting", weight: 0.5 },
- *   ]
- * );
- */
 async function generateImage(model, prompt, options) {
     const result = await (0, executeCall_js_1.executeCall)({
         model,
@@ -54,10 +39,12 @@ async function generateImage(model, prompt, options) {
             generatedImage: output,
         }),
     });
-    return {
-        image: result.output,
-        response: result.response,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            image: result.output,
+            response: result.response,
+            metadata: result.metadata,
+        }
+        : result.output;
 }
 exports.generateImage = generateImage;

package/model-function/generate-image/generateImage.d.ts

@@ -8,7 +8,7 @@ import { ImageGenerationModel, ImageGenerationModelSettings } from "./ImageGener
  * and Stability AI models expect an array of text prompts with optional weights.
  *
  * @example
- * const { image } = await generateImage(
+ * const image = await generateImage(
  *   new StabilityImageGenerationModel(...),
  *   [
  *     { text: "the wicked witch of the west" },
@@ -16,8 +16,13 @@ import { ImageGenerationModel, ImageGenerationModelSettings } from "./ImageGener
  *   ]
  * );
  */
-export declare function generateImage<PROMPT, RESPONSE, SETTINGS extends ImageGenerationModelSettings>(model: ImageGenerationModel<PROMPT, RESPONSE, SETTINGS>, prompt: PROMPT, options?: FunctionOptions<SETTINGS>): Promise<{
+export declare function generateImage<PROMPT, RESPONSE, SETTINGS extends ImageGenerationModelSettings>(model: ImageGenerationModel<PROMPT, RESPONSE, SETTINGS>, prompt: PROMPT, options: FunctionOptions<SETTINGS> & {
+    fullResponse: true;
+}): Promise<{
     image: string;
     response: RESPONSE;
     metadata: CallMetadata<ImageGenerationModel<PROMPT, RESPONSE, SETTINGS>>;
 }>;
+export declare function generateImage<PROMPT, RESPONSE, SETTINGS extends ImageGenerationModelSettings>(model: ImageGenerationModel<PROMPT, RESPONSE, SETTINGS>, prompt: PROMPT, options?: FunctionOptions<SETTINGS> & {
+    fullResponse?: false;
+}): Promise<string>;

package/model-function/generate-image/generateImage.js

@@ -1,19 +1,4 @@
 import { executeCall } from "../executeCall.js";
-/**
- * Generates a base64-encoded image using a prompt.
- * The prompt format depends on the model.
- * For example, OpenAI image models expect a string prompt,
- * and Stability AI models expect an array of text prompts with optional weights.
- *
- * @example
- * const { image } = await generateImage(
- *   new StabilityImageGenerationModel(...),
- *   [
- *     { text: "the wicked witch of the west" },
- *     { text: "style of early 19th century painting", weight: 0.5 },
- *   ]
- * );
- */
 export async function generateImage(model, prompt, options) {
     const result = await executeCall({
         model,
@@ -51,9 +36,11 @@ export async function generateImage(model, prompt, options) {
             generatedImage: output,
         }),
     });
-    return {
-        image: result.output,
-        response: result.response,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            image: result.output,
+            response: result.response,
+            metadata: result.metadata,
+        }
+        : result.output;
 }

package/model-function/generate-json/generateJson.cjs

@@ -52,10 +52,12 @@ async function generateJson(model, schemaDefinition, prompt, options) {
             generatedJson: output,
         }),
     });
-    return {
-        value: result.output,
-        response: result.response,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            value: result.output,
+            response: result.response,
+            metadata: result.metadata,
+        }
+        : result.output;
 }
 exports.generateJson = generateJson;

package/model-function/generate-json/generateJson.d.ts

@@ -2,8 +2,13 @@ import { FunctionOptions } from "../FunctionOptions.js";
 import { CallMetadata } from "../executeCall.js";
 import { GenerateJsonModel, GenerateJsonModelSettings, GenerateJsonPrompt } from "./GenerateJsonModel.js";
 import { SchemaDefinition } from "./SchemaDefinition.js";
-export declare function generateJson<STRUCTURE, PROMPT, RESPONSE, NAME extends string, SETTINGS extends GenerateJsonModelSettings>(model: GenerateJsonModel<PROMPT, RESPONSE, SETTINGS>, schemaDefinition: SchemaDefinition<NAME, STRUCTURE>, prompt: (schemaDefinition: SchemaDefinition<NAME, STRUCTURE>) => PROMPT & GenerateJsonPrompt<RESPONSE>, options?: FunctionOptions<SETTINGS>): Promise<{
+export declare function generateJson<STRUCTURE, PROMPT, RESPONSE, NAME extends string, SETTINGS extends GenerateJsonModelSettings>(model: GenerateJsonModel<PROMPT, RESPONSE, SETTINGS>, schemaDefinition: SchemaDefinition<NAME, STRUCTURE>, prompt: (schemaDefinition: SchemaDefinition<NAME, STRUCTURE>) => PROMPT & GenerateJsonPrompt<RESPONSE>, options: FunctionOptions<SETTINGS> & {
+    fullResponse: true;
+}): Promise<{
     value: STRUCTURE;
     response: RESPONSE;
     metadata: CallMetadata<GenerateJsonModel<PROMPT, RESPONSE, SETTINGS>>;
 }>;
+export declare function generateJson<STRUCTURE, PROMPT, RESPONSE, NAME extends string, SETTINGS extends GenerateJsonModelSettings>(model: GenerateJsonModel<PROMPT, RESPONSE, SETTINGS>, schemaDefinition: SchemaDefinition<NAME, STRUCTURE>, prompt: (schemaDefinition: SchemaDefinition<NAME, STRUCTURE>) => PROMPT & GenerateJsonPrompt<RESPONSE>, options?: FunctionOptions<SETTINGS> & {
+    fullResponse?: false;
+}): Promise<STRUCTURE>;

package/model-function/generate-json/generateJson.js

@@ -49,9 +49,11 @@ export async function generateJson(model, schemaDefinition, prompt, options) {
             generatedJson: output,
         }),
     });
-    return {
-        value: result.output,
-        response: result.response,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            value: result.output,
+            response: result.response,
+            metadata: result.metadata,
+        }
+        : result.output;
 }

package/model-function/generate-json/generateJsonOrText.cjs

@@ -65,10 +65,12 @@ async function generateJsonOrText(model, schemaDefinitions, prompt, options) {
             generatedJson: output,
         }),
     });
-    return {
-        ...result.output,
-        response: result.response,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            ...result.output,
+            response: result.response,
+            metadata: result.metadata,
+        }
+        : result.output;
 }
 exports.generateJsonOrText = generateJsonOrText;

package/model-function/generate-json/generateJsonOrText.d.ts

@@ -14,7 +14,9 @@ type ToSchemaUnion<T> = {
     } : never;
 }[keyof T];
 type ToOutputValue<SCHEMAS extends SchemaDefinitionArray<SchemaDefinition<any, any>[]>> = ToSchemaUnion<ToSchemaDefinitionsMap<SCHEMAS>>;
-export declare function generateJsonOrText<SCHEMAS extends SchemaDefinition<any, any>[], PROMPT, RESPONSE, SETTINGS extends GenerateJsonOrTextModelSettings>(model: GenerateJsonOrTextModel<PROMPT, RESPONSE, SETTINGS>, schemaDefinitions: SCHEMAS, prompt: (schemaDefinitions: SCHEMAS) => PROMPT & GenerateJsonOrTextPrompt<RESPONSE>, options?: FunctionOptions<SETTINGS>): Promise<({
+export declare function generateJsonOrText<SCHEMAS extends SchemaDefinition<any, any>[], PROMPT, RESPONSE, SETTINGS extends GenerateJsonOrTextModelSettings>(model: GenerateJsonOrTextModel<PROMPT, RESPONSE, SETTINGS>, schemaDefinitions: SCHEMAS, prompt: (schemaDefinitions: SCHEMAS) => PROMPT & GenerateJsonOrTextPrompt<RESPONSE>, options: FunctionOptions<SETTINGS> & {
+    fullResponse: true;
+}): Promise<({
     schema: null;
     value: null;
     text: string;
@@ -22,4 +24,11 @@ export declare function generateJsonOrText<SCHEMAS extends SchemaDefinition<any,
     response: RESPONSE;
     metadata: CallMetadata<GenerateJsonOrTextModel<PROMPT, RESPONSE, SETTINGS>>;
 }>;
+export declare function generateJsonOrText<SCHEMAS extends SchemaDefinition<any, any>[], PROMPT, RESPONSE, SETTINGS extends GenerateJsonOrTextModelSettings>(model: GenerateJsonOrTextModel<PROMPT, RESPONSE, SETTINGS>, schemaDefinitions: SCHEMAS, prompt: (schemaDefinitions: SCHEMAS) => PROMPT & GenerateJsonOrTextPrompt<RESPONSE>, options?: FunctionOptions<SETTINGS> & {
+    fullResponse?: false;
+}): Promise<{
+    schema: null;
+    value: null;
+    text: string;
+} | ToOutputValue<SCHEMAS>>;
 export {};

package/model-function/generate-json/generateJsonOrText.js

@@ -62,9 +62,11 @@ export async function generateJsonOrText(model, schemaDefinitions, prompt, optio
             generatedJson: output,
         }),
     });
-    return {
-        ...result.output,
-        response: result.response,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            ...result.output,
+            response: result.response,
+            metadata: result.metadata,
+        }
+        : result.output;
 }

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

@@ -2,18 +2,6 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.generateText = void 0;
 const executeCall_js_1 = require("../executeCall.cjs");
-/**
- * Generates a text using a prompt.
- * The prompt format depends on the model.
- * For example, OpenAI text models expect a string prompt, and OpenAI chat models expect an array of chat messages.
- *
- * @example
- * const model = new OpenAITextGenerationModel(...);
- *
- * const { text } = await model.generateText(
- *   "Write a short story about a robot learning to love:\n\n"
- * );
- */
 async function generateText(
 // eslint-disable-next-line @typescript-eslint/no-explicit-any
 model, prompt, options) {
@@ -58,10 +46,12 @@ model, prompt, options) {
             generatedText: output,
         }),
     });
-    return {
-        text: result.output,
-        response: result.response,
-        metadata: result.metadata,
-    };
+    return options?.fullResponse === true
+        ? {
+            text: result.output,
+            response: result.response,
+            metadata: result.metadata,
+        }
+        : result.output;
 }
 exports.generateText = generateText;

package/model-function/generate-text/generateText.d.ts

@@ -9,12 +9,17 @@ import { TextGenerationModel, TextGenerationModelSettings } from "./TextGenerati
  * @example
  * const model = new OpenAITextGenerationModel(...);
  *
- * const { text } = await model.generateText(
+ * const text = await model.generateText(
  *   "Write a short story about a robot learning to love:\n\n"
  * );
  */
-export declare function generateText<PROMPT, RESPONSE, SETTINGS extends TextGenerationModelSettings>(model: TextGenerationModel<PROMPT, RESPONSE, any, SETTINGS>, prompt: PROMPT, options?: FunctionOptions<SETTINGS>): Promise<{
+export declare function generateText<PROMPT, RESPONSE, SETTINGS extends TextGenerationModelSettings>(model: TextGenerationModel<PROMPT, RESPONSE, any, SETTINGS>, prompt: PROMPT, options: FunctionOptions<SETTINGS> & {
+    fullResponse: true;
+}): Promise<{
     text: string;
     response: RESPONSE;
     metadata: CallMetadata<TextGenerationModel<PROMPT, RESPONSE, unknown, SETTINGS>>;
 }>;
+export declare function generateText<PROMPT, RESPONSE, SETTINGS extends TextGenerationModelSettings>(model: TextGenerationModel<PROMPT, RESPONSE, any, SETTINGS>, prompt: PROMPT, options?: FunctionOptions<SETTINGS> & {
+    fullResponse?: false;
+}): Promise<string>;