modelfusion 0.2.0 → 0.4.0

package/README.md

@@ -366,7 +366,7 @@ const { chunks } = await retrieveTextChunks(
 | [Stream text](https://modelfusion.dev/guide/function/generate-text)                   | ✅                                                                  | ✅                                                                  | ✅                                                                       |                                                                                |                                                                              |                                                                                   |
 | [Generate JSON](https://modelfusion.dev/guide/function/generate-json)                 | chat models                                                         |                                                                     |                                                                          |                                                                                |                                                                              |                                                                                   |
 | [Generate JSON or Text](https://modelfusion.dev/guide/function/generate-json-or-text) | chat models                                                         |                                                                     |                                                                          |                                                                                |                                                                              |                                                                                   |
-| [Embed text](https://modelfusion.dev/guide/function/embed-text)                       | ✅                                                                  | ✅                                                                  | ✅                                                                       |                                                                                |                                                                              |                                                                                   |
+| [Embed text](https://modelfusion.dev/guide/function/embed-text)                       | ✅                                                                  | ✅                                                                  | ✅                                                                       | ✅                                                                             |                                                                              |                                                                                   |
 | [Tokenize text](https://modelfusion.dev/guide/function/tokenize-text)                 | full                                                                | full                                                                | basic                                                                    |                                                                                |                                                                              |                                                                                   |
 | [Generate image](https://modelfusion.dev/guide/function/generate-image)               | ✅                                                                  |                                                                     |                                                                          |                                                                                | ✅                                                                           | ✅                                                                                |
 | [Transcribe audio](https://modelfusion.dev/guide/function/transcribe-audio)           | ✅                                                                  |                                                                     |                                                                          |                                                                                |                                                                              |                                                                                   |
@@ -377,17 +377,21 @@ const { chunks } = await retrieveTextChunks(
 - [Memory](https://modelfusion.dev/integration/vector-index/memory)
 - [Pinecone](https://modelfusion.dev/integration/vector-index/pinecone)
 
-### Prompt Mappings
+### Observability
 
-Use higher level prompts that are mapped into model specific prompts.
+- [Helicone](https://modelfusion.dev/integration/observability/helicone)
 
-| Model        | Instruction Prompt | Chat Prompt |
-| ------------ | ------------------ | ----------- |
-| OpenAI       | ✅                 | ✅          |
-| Llama 2      | ✅                 | ✅          |
-| Alpaca       | ✅                 | ❌          |
-| Vicuna       | ❌                 | ✅          |
-| Generic Text | ✅                 | ✅          |
+### Prompt Formats
+
+Use higher level prompts that are mapped into model specific prompt formats.
+
+| Prompt Format | Instruction Prompt | Chat Prompt |
+| ------------- | ------------------ | ----------- |
+| OpenAI Chat   | ✅                 | ✅          |
+| Llama 2       | ✅                 | ✅          |
+| Alpaca        | ✅                 | ❌          |
+| Vicuna        | ❌                 | ✅          |
+| Generic Text  | ✅                 | ✅          |
 
 ## Documentation
 
@@ -402,17 +406,15 @@ Use higher level prompts that are mapped into model specific prompts.
 
 Examples for the individual functions and objects.
 
-### [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_
+### [Chatbot (Terminal)](https://github.com/lgrammel/modelfusion/tree/main/examples/chatbot-terminal)
 
-Extracts information about a topic from a PDF and writes a tweet in your own style about it.
+> _Terminal app_, _chat_, _llama.cpp_
 
-### [AI Chat (Next.JS)](https://github.com/lgrammel/modelfusion/tree/main/examples/ai-chat-next-js)
+### [Chatbot (Next.JS)](https://github.com/lgrammel/modelfusion/tree/main/examples/chatbot-next-js)
 
 > _Next.js app_, _OpenAI GPT-3.5-turbo_, _streaming_, _abort handling_
 
-A basic web chat with an AI assistant, implemented as a Next.js app.
+A web chat with an AI assistant, implemented as a Next.js app.
 
 ### [Image generator (Next.js)](https://github.com/lgrammel/modelfusion/tree/main/examples/image-generator-next-js)
 
@@ -426,20 +428,20 @@ Create an 19th century painting image for your input.
 
 Record audio with push-to-talk and transcribe it using Whisper, implemented as a Next.js app. The app shows a list of the transcriptions.
 
-### [BabyAGI Classic](https://github.com/lgrammel/modelfusion/tree/main/examples/baby-agi)
+### [BabyAGI Agent](https://github.com/lgrammel/modelfusion/tree/main/examples/babyagi-agent)
 
 > _terminal app_, _agent_, _BabyAGI_, _OpenAI text-davinci-003_
 
 TypeScript implementation of the classic [BabyAGI](https://github.com/yoheinakajima/babyagi/blob/main/classic/babyagi.py) by [@yoheinakajima](https://twitter.com/yoheinakajima) without embeddings.
 
-### [Middle school math](https://github.com/lgrammel/modelfusion/tree/main/examples/middle-school-math)
+### [Middle school math agent](https://github.com/lgrammel/modelfusion/tree/main/examples/middle-school-math-agent)
 
 > _terminal app_, _agent_, _tools_, _GPT-4_
 
 Small agent that solves middle school math problems. It uses a calculator tool to solve the problems.
 
-### [Terminal Chat (llama.cpp)](https://github.com/lgrammel/modelfusion/tree/main/examples/terminal-chat-llamacpp)
+### [PDF to Tweet](https://github.com/lgrammel/modelfusion/tree/main/examples/pdf-to-tweet)
 
-> _Terminal app_, _chat_, _llama.cpp_
+> _terminal app_, _PDF parsing_, _recursive information extraction_, _in memory vector index, \_style example retrieval_, _OpenAI GPT-4_, _cost calculation_
 
-A terminal chat with a Llama.cpp server backend.
+Extracts information about a topic from a PDF and writes a tweet in your own style about it.

package/model-provider/cohere/CohereError.cjs

@@ -25,6 +25,17 @@ class CohereError extends ApiCallError_js_1.ApiCallError {
 exports.CohereError = CohereError;
 const failedCohereCallResponseHandler = async ({ response, url, requestBodyValues }) => {
     const responseBody = await response.text();
+    // For some errors, the body of Cohere responses is empty:
+    if (responseBody.trim() === "") {
+        return new CohereError({
+            url,
+            requestBodyValues,
+            statusCode: response.status,
+            data: {
+                message: response.statusText,
+            },
+        });
+    }
     const parsedError = exports.cohereErrorDataSchema.parse(secure_json_parse_1.default.parse(responseBody));
     return new CohereError({
         url,

package/model-provider/cohere/CohereError.js

@@ -18,6 +18,17 @@ export class CohereError extends ApiCallError {
 }
 export const failedCohereCallResponseHandler = async ({ response, url, requestBodyValues }) => {
     const responseBody = await response.text();
+    // For some errors, the body of Cohere responses is empty:
+    if (responseBody.trim() === "") {
+        return new CohereError({
+            url,
+            requestBodyValues,
+            statusCode: response.status,
+            data: {
+                message: response.statusText,
+            },
+        });
+    }
     const parsedError = cohereErrorDataSchema.parse(SecureJSON.parse(responseBody));
     return new CohereError({
         url,

package/model-provider/cohere/CohereTextEmbeddingModel.cjs

@@ -159,7 +159,9 @@ const cohereTextEmbeddingResponseSchema = zod_1.default.object({
 async function callCohereEmbeddingAPI({ baseUrl = "https://api.cohere.ai/v1", abortSignal, apiKey, model, texts, truncate, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/embed`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             model,
             texts,

package/model-provider/cohere/CohereTextEmbeddingModel.js

@@ -152,7 +152,9 @@ const cohereTextEmbeddingResponseSchema = z.object({
 async function callCohereEmbeddingAPI({ baseUrl = "https://api.cohere.ai/v1", abortSignal, apiKey, model, texts, truncate, }) {
     return postJsonToApi({
         url: `${baseUrl}/embed`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             model,
             texts,

package/model-provider/cohere/CohereTextGenerationModel.cjs

@@ -179,7 +179,9 @@ const cohereTextGenerationResponseSchema = zod_1.z.object({
 async function callCohereTextGenerationAPI({ baseUrl = "https://api.cohere.ai/v1", abortSignal, responseFormat, apiKey, model, prompt, numGenerations, maxTokens, temperature, k, p, frequencyPenalty, presencePenalty, endSequences, stopSequences, returnLikelihoods, logitBias, truncate, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/generate`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             stream: responseFormat.stream,
             model,

package/model-provider/cohere/CohereTextGenerationModel.js

@@ -172,7 +172,9 @@ const cohereTextGenerationResponseSchema = z.object({
 async function callCohereTextGenerationAPI({ baseUrl = "https://api.cohere.ai/v1", abortSignal, responseFormat, apiKey, model, prompt, numGenerations, maxTokens, temperature, k, p, frequencyPenalty, presencePenalty, endSequences, stopSequences, returnLikelihoods, logitBias, truncate, }) {
     return postJsonToApi({
         url: `${baseUrl}/generate`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             stream: responseFormat.stream,
             model,

package/model-provider/cohere/CohereTokenizer.cjs

@@ -97,7 +97,9 @@ const cohereDetokenizationResponseSchema = zod_1.default.object({
 async function callCohereDetokenizeAPI({ baseUrl = "https://api.cohere.ai/v1", abortSignal, apiKey, model, tokens, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/detokenize`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             model,
             tokens,
@@ -124,7 +126,9 @@ const cohereTokenizationResponseSchema = zod_1.default.object({
 async function callCohereTokenizeAPI({ baseUrl = "https://api.cohere.ai/v1", abortSignal, apiKey, model, text, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/tokenize`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             model,
             text,

package/model-provider/cohere/CohereTokenizer.js

@@ -90,7 +90,9 @@ const cohereDetokenizationResponseSchema = z.object({
 async function callCohereDetokenizeAPI({ baseUrl = "https://api.cohere.ai/v1", abortSignal, apiKey, model, tokens, }) {
     return postJsonToApi({
         url: `${baseUrl}/detokenize`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             model,
             tokens,
@@ -117,7 +119,9 @@ const cohereTokenizationResponseSchema = z.object({
 async function callCohereTokenizeAPI({ baseUrl = "https://api.cohere.ai/v1", abortSignal, apiKey, model, text, }) {
     return postJsonToApi({
         url: `${baseUrl}/tokenize`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             model,
             text,

package/model-provider/huggingface/HuggingFaceTextEmbeddingModel.cjs

@@ -0,0 +1,138 @@
+"use strict";
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.HuggingFaceTextEmbeddingModel = void 0;
+const zod_1 = __importDefault(require("zod"));
+const AbstractModel_js_1 = require("../../model-function/AbstractModel.cjs");
+const callWithRetryAndThrottle_js_1 = require("../../util/api/callWithRetryAndThrottle.cjs");
+const postToApi_js_1 = require("../../util/api/postToApi.cjs");
+const HuggingFaceError_js_1 = require("./HuggingFaceError.cjs");
+/**
+ * Create a text embeddinng model that calls a Hugging Face Inference API Feature Extraction Task.
+ *
+ * @see https://huggingface.co/docs/api-inference/detailed_parameters#feature-extraction-task
+ *
+ * @example
+ * const model = new HuggingFaceTextGenerationModel({
+ *   model: "intfloat/e5-base-v2",
+ *   maxTexstsPerCall: 5,
+ *   retry: retryWithExponentialBackoff({ maxTries: 5 }),
+ * });
+ *
+ * const { embeddings } = await embedTexts(
+ *   model,
+ *   [
+ *     "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.",
+ *   ]
+ * );
+ */
+class HuggingFaceTextEmbeddingModel extends AbstractModel_js_1.AbstractModel {
+    constructor(settings) {
+        super({ settings });
+        Object.defineProperty(this, "provider", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: "huggingface"
+        });
+        Object.defineProperty(this, "maxTextsPerCall", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "contextWindowSize", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: undefined
+        });
+        Object.defineProperty(this, "embeddingDimensions", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "tokenizer", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: undefined
+        });
+        Object.defineProperty(this, "countPromptTokens", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: undefined
+        });
+        // There is no limit documented in the HuggingFace API. Use 1024 as a reasonable default.
+        this.maxTextsPerCall = settings.maxTextsPerCall ?? 1024;
+        this.embeddingDimensions = settings.embeddingDimensions;
+    }
+    get modelName() {
+        return this.settings.model;
+    }
+    get apiKey() {
+        const apiKey = this.settings.apiKey ?? process.env.HUGGINGFACE_API_KEY;
+        if (apiKey == null) {
+            throw new Error("No Hugging Face API key provided. Pass it in the constructor or set the HUGGINGFACE_API_KEY environment variable.");
+        }
+        return apiKey;
+    }
+    async callAPI(texts, options) {
+        if (texts.length > this.maxTextsPerCall) {
+            throw new Error(`The HuggingFace feature extraction API is configured to only support ${this.maxTextsPerCall} texts per API call.`);
+        }
+        const run = options?.run;
+        const settings = options?.settings;
+        const callSettings = Object.assign({
+            apiKey: this.apiKey,
+            options: {
+                useCache: true,
+                waitForModel: true,
+            },
+        }, this.settings, settings, {
+            abortSignal: run?.abortSignal,
+            inputs: texts,
+        });
+        return (0, callWithRetryAndThrottle_js_1.callWithRetryAndThrottle)({
+            retry: this.settings.retry,
+            throttle: this.settings.throttle,
+            call: async () => callHuggingFaceTextGenerationAPI(callSettings),
+        });
+    }
+    generateEmbeddingResponse(texts, options) {
+        return this.callAPI(texts, options);
+    }
+    extractEmbeddings(response) {
+        return response;
+    }
+    withSettings(additionalSettings) {
+        return new HuggingFaceTextEmbeddingModel(Object.assign({}, this.settings, additionalSettings));
+    }
+}
+exports.HuggingFaceTextEmbeddingModel = HuggingFaceTextEmbeddingModel;
+const huggingFaceTextEmbeddingResponseSchema = zod_1.default.array(zod_1.default.array(zod_1.default.number()));
+async function callHuggingFaceTextGenerationAPI({ baseUrl = "https://api-inference.huggingface.co/pipeline/feature-extraction", abortSignal, apiKey, model, inputs, options, }) {
+    return (0, postToApi_js_1.postJsonToApi)({
+        url: `${baseUrl}/${model}`,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
+        body: {
+            inputs,
+            options: options
+                ? {
+                    use_cache: options?.useCache,
+                    wait_for_model: options?.waitForModel,
+                }
+                : undefined,
+        },
+        failedResponseHandler: HuggingFaceError_js_1.failedHuggingFaceCallResponseHandler,
+        successfulResponseHandler: (0, postToApi_js_1.createJsonResponseHandler)(huggingFaceTextEmbeddingResponseSchema),
+        abortSignal,
+    });
+}

package/model-provider/huggingface/HuggingFaceTextEmbeddingModel.d.ts

@@ -0,0 +1,57 @@
+import z from "zod";
+import { AbstractModel } from "../../model-function/AbstractModel.js";
+import { FunctionOptions } from "../../model-function/FunctionOptions.js";
+import { TextEmbeddingModel, TextEmbeddingModelSettings } from "../../model-function/embed-text/TextEmbeddingModel.js";
+import { RetryFunction } from "../../util/api/RetryFunction.js";
+import { ThrottleFunction } from "../../util/api/ThrottleFunction.js";
+export interface HuggingFaceTextEmbeddingModelSettings extends TextEmbeddingModelSettings {
+    model: string;
+    baseUrl?: string;
+    apiKey?: string;
+    maxTextsPerCall?: number;
+    embeddingDimensions?: number;
+    retry?: RetryFunction;
+    throttle?: ThrottleFunction;
+    options?: {
+        useCache?: boolean;
+        waitForModel?: boolean;
+    };
+}
+/**
+ * Create a text embeddinng model that calls a Hugging Face Inference API Feature Extraction Task.
+ *
+ * @see https://huggingface.co/docs/api-inference/detailed_parameters#feature-extraction-task
+ *
+ * @example
+ * const model = new HuggingFaceTextGenerationModel({
+ *   model: "intfloat/e5-base-v2",
+ *   maxTexstsPerCall: 5,
+ *   retry: retryWithExponentialBackoff({ maxTries: 5 }),
+ * });
+ *
+ * const { embeddings } = await embedTexts(
+ *   model,
+ *   [
+ *     "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 declare class HuggingFaceTextEmbeddingModel extends AbstractModel<HuggingFaceTextEmbeddingModelSettings> implements TextEmbeddingModel<HuggingFaceTextEmbeddingResponse, HuggingFaceTextEmbeddingModelSettings> {
+    constructor(settings: HuggingFaceTextEmbeddingModelSettings);
+    readonly provider = "huggingface";
+    get modelName(): string;
+    readonly maxTextsPerCall: number;
+    readonly contextWindowSize: undefined;
+    readonly embeddingDimensions: number | undefined;
+    readonly tokenizer: undefined;
+    private get apiKey();
+    callAPI(texts: Array<string>, options?: FunctionOptions<HuggingFaceTextEmbeddingModelSettings>): Promise<HuggingFaceTextEmbeddingResponse>;
+    readonly countPromptTokens: undefined;
+    generateEmbeddingResponse(texts: string[], options?: FunctionOptions<HuggingFaceTextEmbeddingModelSettings>): Promise<number[][]>;
+    extractEmbeddings(response: HuggingFaceTextEmbeddingResponse): number[][];
+    withSettings(additionalSettings: Partial<HuggingFaceTextEmbeddingModelSettings>): this;
+}
+declare const huggingFaceTextEmbeddingResponseSchema: z.ZodArray<z.ZodArray<z.ZodNumber, "many">, "many">;
+export type HuggingFaceTextEmbeddingResponse = z.infer<typeof huggingFaceTextEmbeddingResponseSchema>;
+export {};

package/model-provider/huggingface/HuggingFaceTextEmbeddingModel.js

@@ -0,0 +1,131 @@
+import z from "zod";
+import { AbstractModel } from "../../model-function/AbstractModel.js";
+import { callWithRetryAndThrottle } from "../../util/api/callWithRetryAndThrottle.js";
+import { createJsonResponseHandler, postJsonToApi, } from "../../util/api/postToApi.js";
+import { failedHuggingFaceCallResponseHandler } from "./HuggingFaceError.js";
+/**
+ * Create a text embeddinng model that calls a Hugging Face Inference API Feature Extraction Task.
+ *
+ * @see https://huggingface.co/docs/api-inference/detailed_parameters#feature-extraction-task
+ *
+ * @example
+ * const model = new HuggingFaceTextGenerationModel({
+ *   model: "intfloat/e5-base-v2",
+ *   maxTexstsPerCall: 5,
+ *   retry: retryWithExponentialBackoff({ maxTries: 5 }),
+ * });
+ *
+ * const { embeddings } = await embedTexts(
+ *   model,
+ *   [
+ *     "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 class HuggingFaceTextEmbeddingModel extends AbstractModel {
+    constructor(settings) {
+        super({ settings });
+        Object.defineProperty(this, "provider", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: "huggingface"
+        });
+        Object.defineProperty(this, "maxTextsPerCall", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "contextWindowSize", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: undefined
+        });
+        Object.defineProperty(this, "embeddingDimensions", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "tokenizer", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: undefined
+        });
+        Object.defineProperty(this, "countPromptTokens", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: undefined
+        });
+        // There is no limit documented in the HuggingFace API. Use 1024 as a reasonable default.
+        this.maxTextsPerCall = settings.maxTextsPerCall ?? 1024;
+        this.embeddingDimensions = settings.embeddingDimensions;
+    }
+    get modelName() {
+        return this.settings.model;
+    }
+    get apiKey() {
+        const apiKey = this.settings.apiKey ?? process.env.HUGGINGFACE_API_KEY;
+        if (apiKey == null) {
+            throw new Error("No Hugging Face API key provided. Pass it in the constructor or set the HUGGINGFACE_API_KEY environment variable.");
+        }
+        return apiKey;
+    }
+    async callAPI(texts, options) {
+        if (texts.length > this.maxTextsPerCall) {
+            throw new Error(`The HuggingFace feature extraction API is configured to only support ${this.maxTextsPerCall} texts per API call.`);
+        }
+        const run = options?.run;
+        const settings = options?.settings;
+        const callSettings = Object.assign({
+            apiKey: this.apiKey,
+            options: {
+                useCache: true,
+                waitForModel: true,
+            },
+        }, this.settings, settings, {
+            abortSignal: run?.abortSignal,
+            inputs: texts,
+        });
+        return callWithRetryAndThrottle({
+            retry: this.settings.retry,
+            throttle: this.settings.throttle,
+            call: async () => callHuggingFaceTextGenerationAPI(callSettings),
+        });
+    }
+    generateEmbeddingResponse(texts, options) {
+        return this.callAPI(texts, options);
+    }
+    extractEmbeddings(response) {
+        return response;
+    }
+    withSettings(additionalSettings) {
+        return new HuggingFaceTextEmbeddingModel(Object.assign({}, this.settings, additionalSettings));
+    }
+}
+const huggingFaceTextEmbeddingResponseSchema = z.array(z.array(z.number()));
+async function callHuggingFaceTextGenerationAPI({ baseUrl = "https://api-inference.huggingface.co/pipeline/feature-extraction", abortSignal, apiKey, model, inputs, options, }) {
+    return postJsonToApi({
+        url: `${baseUrl}/${model}`,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
+        body: {
+            inputs,
+            options: options
+                ? {
+                    use_cache: options?.useCache,
+                    wait_for_model: options?.waitForModel,
+                }
+                : undefined,
+        },
+        failedResponseHandler: failedHuggingFaceCallResponseHandler,
+        successfulResponseHandler: createJsonResponseHandler(huggingFaceTextEmbeddingResponseSchema),
+        abortSignal,
+    });
+}

package/model-provider/huggingface/HuggingFaceTextGenerationModel.cjs

@@ -149,7 +149,9 @@ const huggingFaceTextGenerationResponseSchema = zod_1.default.array(zod_1.defaul
 async function callHuggingFaceTextGenerationAPI({ baseUrl = "https://api-inference.huggingface.co/models", abortSignal, apiKey, model, inputs, topK, topP, temperature, repetitionPenalty, maxNewTokens, maxTime, numReturnSequences, doSample, options, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/${model}`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             inputs,
             top_k: topK,

package/model-provider/huggingface/HuggingFaceTextGenerationModel.js

@@ -142,7 +142,9 @@ const huggingFaceTextGenerationResponseSchema = z.array(z.object({
 async function callHuggingFaceTextGenerationAPI({ baseUrl = "https://api-inference.huggingface.co/models", abortSignal, apiKey, model, inputs, topK, topP, temperature, repetitionPenalty, maxNewTokens, maxTime, numReturnSequences, doSample, options, }) {
     return postJsonToApi({
         url: `${baseUrl}/${model}`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             inputs,
             top_k: topK,

package/model-provider/huggingface/index.cjs

@@ -17,4 +17,5 @@ Object.defineProperty(exports, "__esModule", { value: true });
 exports.HuggingFaceError = void 0;
 var HuggingFaceError_js_1 = require("./HuggingFaceError.cjs");
 Object.defineProperty(exports, "HuggingFaceError", { enumerable: true, get: function () { return HuggingFaceError_js_1.HuggingFaceError; } });
+__exportStar(require("./HuggingFaceTextEmbeddingModel.cjs"), exports);
 __exportStar(require("./HuggingFaceTextGenerationModel.cjs"), exports);

package/model-provider/huggingface/index.d.ts

@@ -1,2 +1,3 @@
 export { HuggingFaceError, HuggingFaceErrorData } from "./HuggingFaceError.js";
+export * from "./HuggingFaceTextEmbeddingModel.js";
 export * from "./HuggingFaceTextGenerationModel.js";

package/model-provider/huggingface/index.js

@@ -1,2 +1,3 @@
 export { HuggingFaceError } from "./HuggingFaceError.js";
+export * from "./HuggingFaceTextEmbeddingModel.js";
 export * from "./HuggingFaceTextGenerationModel.js";

package/model-provider/llamacpp/LlamaCppTextEmbeddingModel.cjs

@@ -35,7 +35,7 @@ class LlamaCppTextEmbeddingModel extends AbstractModel_js_1.AbstractModel {
             enumerable: true,
             configurable: true,
             writable: true,
-            value: undefined
+            value: void 0
         });
         Object.defineProperty(this, "tokenizer", {
             enumerable: true,
@@ -48,6 +48,7 @@ class LlamaCppTextEmbeddingModel extends AbstractModel_js_1.AbstractModel {
             retry: this.settings.tokenizerSettings?.retry,
             throttle: this.settings.tokenizerSettings?.throttle,
         });
+        this.embeddingDimensions = this.settings.embeddingDimensions;
     }
     get modelName() {
         return null;

package/model-provider/llamacpp/LlamaCppTextEmbeddingModel.d.ts

@@ -4,8 +4,9 @@ import { FunctionOptions } from "../../model-function/FunctionOptions.js";
 import { TextEmbeddingModel, TextEmbeddingModelSettings } from "../../model-function/embed-text/TextEmbeddingModel.js";
 import { RetryFunction } from "../../util/api/RetryFunction.js";
 import { ThrottleFunction } from "../../util/api/ThrottleFunction.js";
-export interface LlamaCppEmbeddingModelSettings extends TextEmbeddingModelSettings {
+export interface LlamaCppTextEmbeddingModelSettings extends TextEmbeddingModelSettings {
     baseUrl?: string;
+    embeddingDimensions?: number;
     retry?: RetryFunction;
     throttle?: ThrottleFunction;
     tokenizerSettings?: {
@@ -13,21 +14,21 @@ export interface LlamaCppEmbeddingModelSettings extends TextEmbeddingModelSettin
         throttle?: ThrottleFunction;
     };
 }
-export declare class LlamaCppTextEmbeddingModel extends AbstractModel<LlamaCppEmbeddingModelSettings> implements TextEmbeddingModel<LlamaCppTextEmbeddingResponse, LlamaCppEmbeddingModelSettings> {
-    constructor(settings?: LlamaCppEmbeddingModelSettings);
+export declare class LlamaCppTextEmbeddingModel extends AbstractModel<LlamaCppTextEmbeddingModelSettings> implements TextEmbeddingModel<LlamaCppTextEmbeddingResponse, LlamaCppTextEmbeddingModelSettings> {
+    constructor(settings?: LlamaCppTextEmbeddingModelSettings);
     readonly provider: "llamacpp";
     get modelName(): null;
     readonly maxTextsPerCall = 1;
     readonly contextWindowSize: undefined;
-    readonly embeddingDimensions: undefined;
+    readonly embeddingDimensions: number | undefined;
     private readonly tokenizer;
     tokenize(text: string): Promise<number[]>;
-    callAPI(texts: Array<string>, options?: FunctionOptions<LlamaCppEmbeddingModelSettings>): Promise<LlamaCppTextEmbeddingResponse>;
-    generateEmbeddingResponse(texts: string[], options?: FunctionOptions<LlamaCppEmbeddingModelSettings>): Promise<{
+    callAPI(texts: Array<string>, options?: FunctionOptions<LlamaCppTextEmbeddingModelSettings>): Promise<LlamaCppTextEmbeddingResponse>;
+    generateEmbeddingResponse(texts: string[], options?: FunctionOptions<LlamaCppTextEmbeddingModelSettings>): Promise<{
         embedding: number[];
     }>;
     extractEmbeddings(response: LlamaCppTextEmbeddingResponse): number[][];
-    withSettings(additionalSettings: Partial<LlamaCppEmbeddingModelSettings>): this;
+    withSettings(additionalSettings: Partial<LlamaCppTextEmbeddingModelSettings>): this;
 }
 declare const llamaCppTextEmbeddingResponseSchema: z.ZodObject<{
     embedding: z.ZodArray<z.ZodNumber, "many">;

package/model-provider/llamacpp/LlamaCppTextEmbeddingModel.js

@@ -29,7 +29,7 @@ export class LlamaCppTextEmbeddingModel extends AbstractModel {
             enumerable: true,
             configurable: true,
             writable: true,
-            value: undefined
+            value: void 0
         });
         Object.defineProperty(this, "tokenizer", {
             enumerable: true,
@@ -42,6 +42,7 @@ export class LlamaCppTextEmbeddingModel extends AbstractModel {
             retry: this.settings.tokenizerSettings?.retry,
             throttle: this.settings.tokenizerSettings?.throttle,
         });
+        this.embeddingDimensions = this.settings.embeddingDimensions;
     }
     get modelName() {
         return null;

package/model-provider/openai/OpenAIError.cjs

@@ -16,7 +16,7 @@ exports.openAIErrorDataSchema = zod_1.z.object({
     }),
 });
 class OpenAIError extends ApiCallError_js_1.ApiCallError {
-    constructor({ data, statusCode, url, requestBodyValues, message = data.error.message, }) {
+    constructor({ data, statusCode, url, requestBodyValues, message, }) {
         super({
             message,
             statusCode,
@@ -24,7 +24,7 @@ class OpenAIError extends ApiCallError_js_1.ApiCallError {
             url,
             isRetryable: (statusCode === 429 &&
                 // insufficient_quota is also reported as a 429, but it's not retryable:
-                data.error.type !== "insufficient_quota") ||
+                data?.error.type !== "insufficient_quota") ||
                 statusCode >= 500,
         });
         Object.defineProperty(this, "data", {
@@ -39,12 +39,24 @@ class OpenAIError extends ApiCallError_js_1.ApiCallError {
 exports.OpenAIError = OpenAIError;
 const failedOpenAICallResponseHandler = async ({ response, url, requestBodyValues }) => {
     const responseBody = await response.text();
-    const parsedError = exports.openAIErrorDataSchema.parse(secure_json_parse_1.default.parse(responseBody));
-    return new OpenAIError({
-        url,
-        requestBodyValues,
-        statusCode: response.status,
-        data: parsedError,
-    });
+    // resilient parsing in case the response is not JSON or does not match the schema:
+    try {
+        const parsedError = exports.openAIErrorDataSchema.parse(secure_json_parse_1.default.parse(responseBody));
+        return new OpenAIError({
+            url,
+            requestBodyValues,
+            statusCode: response.status,
+            message: parsedError.error.message,
+            data: parsedError,
+        });
+    }
+    catch (parseError) {
+        return new OpenAIError({
+            url,
+            requestBodyValues,
+            statusCode: response.status,
+            message: responseBody.trim() !== "" ? responseBody : response.statusText,
+        });
+    }
 };
 exports.failedOpenAICallResponseHandler = failedOpenAICallResponseHandler;

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

@@ -35,13 +35,13 @@ export declare const openAIErrorDataSchema: z.ZodObject<{
 }>;
 export type OpenAIErrorData = z.infer<typeof openAIErrorDataSchema>;
 export declare class OpenAIError extends ApiCallError {
-    readonly data: OpenAIErrorData;
+    readonly data?: OpenAIErrorData;
     constructor({ data, statusCode, url, requestBodyValues, message, }: {
-        message?: string;
+        message: string;
         statusCode: number;
         url: string;
         requestBodyValues: unknown;
-        data: OpenAIErrorData;
+        data?: OpenAIErrorData;
     });
 }
 export declare const failedOpenAICallResponseHandler: ResponseHandler<ApiCallError>;

package/model-provider/openai/OpenAIError.js

@@ -10,7 +10,7 @@ export const openAIErrorDataSchema = z.object({
     }),
 });
 export class OpenAIError extends ApiCallError {
-    constructor({ data, statusCode, url, requestBodyValues, message = data.error.message, }) {
+    constructor({ data, statusCode, url, requestBodyValues, message, }) {
         super({
             message,
             statusCode,
@@ -18,7 +18,7 @@ export class OpenAIError extends ApiCallError {
             url,
             isRetryable: (statusCode === 429 &&
                 // insufficient_quota is also reported as a 429, but it's not retryable:
-                data.error.type !== "insufficient_quota") ||
+                data?.error.type !== "insufficient_quota") ||
                 statusCode >= 500,
         });
         Object.defineProperty(this, "data", {
@@ -32,11 +32,23 @@ export class OpenAIError extends ApiCallError {
 }
 export const failedOpenAICallResponseHandler = async ({ response, url, requestBodyValues }) => {
     const responseBody = await response.text();
-    const parsedError = openAIErrorDataSchema.parse(SecureJSON.parse(responseBody));
-    return new OpenAIError({
-        url,
-        requestBodyValues,
-        statusCode: response.status,
-        data: parsedError,
-    });
+    // resilient parsing in case the response is not JSON or does not match the schema:
+    try {
+        const parsedError = openAIErrorDataSchema.parse(SecureJSON.parse(responseBody));
+        return new OpenAIError({
+            url,
+            requestBodyValues,
+            statusCode: response.status,
+            message: parsedError.error.message,
+            data: parsedError,
+        });
+    }
+    catch (parseError) {
+        return new OpenAIError({
+            url,
+            requestBodyValues,
+            statusCode: response.status,
+            message: responseBody.trim() !== "" ? responseBody : response.statusText,
+        });
+    }
 };

package/model-provider/openai/OpenAIImageGenerationModel.cjs

@@ -109,7 +109,9 @@ exports.OpenAIImageGenerationResponseFormat = {
 async function callOpenAIImageGenerationAPI({ baseUrl = "https://api.openai.com/v1", abortSignal, apiKey, prompt, n, size, responseFormat, user, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/images/generations`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             prompt,
             n,

package/model-provider/openai/OpenAIImageGenerationModel.js

@@ -104,7 +104,9 @@ export const OpenAIImageGenerationResponseFormat = {
 async function callOpenAIImageGenerationAPI({ baseUrl = "https://api.openai.com/v1", abortSignal, apiKey, prompt, n, size, responseFormat, user, }) {
     return postJsonToApi({
         url: `${baseUrl}/images/generations`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             prompt,
             n,

package/model-provider/openai/OpenAITextEmbeddingModel.cjs

@@ -158,7 +158,9 @@ const openAITextEmbeddingResponseSchema = zod_1.default.object({
 async function callOpenAITextEmbeddingAPI({ baseUrl = "https://api.openai.com/v1", abortSignal, apiKey, model, input, user, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/embeddings`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             model,
             input,

package/model-provider/openai/OpenAITextEmbeddingModel.js

@@ -149,7 +149,9 @@ const openAITextEmbeddingResponseSchema = z.object({
 async function callOpenAITextEmbeddingAPI({ baseUrl = "https://api.openai.com/v1", abortSignal, apiKey, model, input, user, }) {
     return postJsonToApi({
         url: `${baseUrl}/embeddings`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             model,
             input,

package/model-provider/openai/OpenAITextGenerationModel.cjs

@@ -209,10 +209,13 @@ const openAITextGenerationResponseSchema = zod_1.default.object({
  *
  * console.log(response.choices[0].text);
  */
-async function callOpenAITextGenerationAPI({ baseUrl = "https://api.openai.com/v1", abortSignal, responseFormat, apiKey, model, prompt, suffix, maxTokens, temperature, topP, n, logprobs, echo, stop, presencePenalty, frequencyPenalty, bestOf, user, }) {
+async function callOpenAITextGenerationAPI({ baseUrl = "https://api.openai.com/v1", headers, abortSignal, responseFormat, apiKey, model, prompt, suffix, maxTokens, temperature, topP, n, logprobs, echo, stop, presencePenalty, frequencyPenalty, bestOf, user, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/completions`,
-        apiKey,
+        headers: {
+            ...headers,
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             stream: responseFormat.stream,
             model,

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

@@ -65,6 +65,7 @@ export declare const calculateOpenAITextGenerationCostInMillicents: ({ model, re
 }) => number;
 export interface OpenAITextGenerationModelSettings extends TextGenerationModelSettings {
     model: OpenAITextGenerationModelType;
+    headers?: Record<string, string>;
     baseUrl?: string;
     apiKey?: string;
     retry?: RetryFunction;

package/model-provider/openai/OpenAITextGenerationModel.js

@@ -200,10 +200,13 @@ const openAITextGenerationResponseSchema = z.object({
  *
  * console.log(response.choices[0].text);
  */
-async function callOpenAITextGenerationAPI({ baseUrl = "https://api.openai.com/v1", abortSignal, responseFormat, apiKey, model, prompt, suffix, maxTokens, temperature, topP, n, logprobs, echo, stop, presencePenalty, frequencyPenalty, bestOf, user, }) {
+async function callOpenAITextGenerationAPI({ baseUrl = "https://api.openai.com/v1", headers, abortSignal, responseFormat, apiKey, model, prompt, suffix, maxTokens, temperature, topP, n, logprobs, echo, stop, presencePenalty, frequencyPenalty, bestOf, user, }) {
     return postJsonToApi({
         url: `${baseUrl}/completions`,
-        apiKey,
+        headers: {
+            ...headers,
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             stream: responseFormat.stream,
             model,

package/model-provider/openai/OpenAITranscriptionModel.cjs

@@ -132,8 +132,9 @@ async function callOpenAITranscriptionAPI({ baseUrl = "https://api.openai.com/v1
     }
     return (0, postToApi_js_1.postToApi)({
         url: `${baseUrl}/audio/transcriptions`,
-        apiKey,
-        contentType: null,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             content: formData,
             values: {

package/model-provider/openai/OpenAITranscriptionModel.js

@@ -124,8 +124,9 @@ async function callOpenAITranscriptionAPI({ baseUrl = "https://api.openai.com/v1
     }
     return postToApi({
         url: `${baseUrl}/audio/transcriptions`,
-        apiKey,
-        contentType: null,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             content: formData,
             values: {

package/model-provider/openai/chat/OpenAIChatModel.cjs

@@ -246,10 +246,13 @@ const openAIChatResponseSchema = zod_1.default.object({
         total_tokens: zod_1.default.number(),
     }),
 });
-async function callOpenAIChatCompletionAPI({ baseUrl = "https://api.openai.com/v1", abortSignal, responseFormat, apiKey, model, messages, functions, functionCall, temperature, topP, n, stop, maxTokens, presencePenalty, frequencyPenalty, user, }) {
+async function callOpenAIChatCompletionAPI({ baseUrl = "https://api.openai.com/v1", headers, abortSignal, responseFormat, apiKey, model, messages, functions, functionCall, temperature, topP, n, stop, maxTokens, presencePenalty, frequencyPenalty, user, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/chat/completions`,
-        apiKey,
+        headers: {
+            ...headers,
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             stream: responseFormat.stream,
             model,

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

@@ -77,6 +77,7 @@ export declare const calculateOpenAIChatCostInMillicents: ({ model, response, }:
 }) => number;
 export interface OpenAIChatCallSettings {
     model: OpenAIChatModelType;
+    headers?: Record<string, string>;
     functions?: Array<{
         name: string;
         description?: string;

package/model-provider/openai/chat/OpenAIChatModel.js

@@ -237,10 +237,13 @@ const openAIChatResponseSchema = z.object({
         total_tokens: z.number(),
     }),
 });
-async function callOpenAIChatCompletionAPI({ baseUrl = "https://api.openai.com/v1", abortSignal, responseFormat, apiKey, model, messages, functions, functionCall, temperature, topP, n, stop, maxTokens, presencePenalty, frequencyPenalty, user, }) {
+async function callOpenAIChatCompletionAPI({ baseUrl = "https://api.openai.com/v1", headers, abortSignal, responseFormat, apiKey, model, messages, functions, functionCall, temperature, topP, n, stop, maxTokens, presencePenalty, frequencyPenalty, user, }) {
     return postJsonToApi({
         url: `${baseUrl}/chat/completions`,
-        apiKey,
+        headers: {
+            ...headers,
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             stream: responseFormat.stream,
             model,

package/model-provider/stability/StabilityImageGenerationModel.cjs

@@ -113,7 +113,9 @@ const stabilityImageGenerationResponseSchema = zod_1.z.object({
 async function callStabilityImageGenerationAPI({ baseUrl = "https://api.stability.ai/v1", abortSignal, apiKey, engineId, height, width, textPrompts, cfgScale, clipGuidancePreset, sampler, samples, seed, steps, stylePreset, }) {
     return (0, postToApi_js_1.postJsonToApi)({
         url: `${baseUrl}/generation/${engineId}/text-to-image`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             height,
             width,

package/model-provider/stability/StabilityImageGenerationModel.js

@@ -109,7 +109,9 @@ const stabilityImageGenerationResponseSchema = z.object({
 async function callStabilityImageGenerationAPI({ baseUrl = "https://api.stability.ai/v1", abortSignal, apiKey, engineId, height, width, textPrompts, cfgScale, clipGuidancePreset, sampler, samples, seed, steps, stylePreset, }) {
     return postJsonToApi({
         url: `${baseUrl}/generation/${engineId}/text-to-image`,
-        apiKey,
+        headers: {
+            Authorization: `Bearer ${apiKey}`,
+        },
         body: {
             height,
             width,

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "modelfusion",
   "description": "Build AI applications, chatbots, and agents with JavaScript and TypeScript.",
-  "version": "0.2.0",
+  "version": "0.4.0",
   "author": "Lars Grammel",
   "license": "MIT",
   "keywords": [

package/prompt/chat/trimChatPrompt.cjs

@@ -9,6 +9,8 @@ const validateChatPrompt_js_1 = require("./validateChatPrompt.cjs");
  *
  * When the minimal chat prompt (system message + last user message) is already too long, it will only
  * return this minimal chat prompt.
+ *
+ * @see https://modelfusion.dev/guide/function/generate-text/prompt-mapping#limiting-the-chat-length
  */
 async function trimChatPrompt({ prompt, model, tokenLimit = model.contextWindowSize - model.maxCompletionTokens, }) {
     (0, validateChatPrompt_js_1.validateChatPrompt)(prompt);

package/prompt/chat/trimChatPrompt.d.ts

@@ -7,6 +7,8 @@ import { ChatPrompt } from "./ChatPrompt.js";
  *
  * When the minimal chat prompt (system message + last user message) is already too long, it will only
  * return this minimal chat prompt.
+ *
+ * @see https://modelfusion.dev/guide/function/generate-text/prompt-mapping#limiting-the-chat-length
  */
 export declare function trimChatPrompt({ prompt, model, tokenLimit, }: {
     prompt: ChatPrompt;

package/prompt/chat/trimChatPrompt.js

@@ -6,6 +6,8 @@ import { validateChatPrompt } from "./validateChatPrompt.js";
  *
  * When the minimal chat prompt (system message + last user message) is already too long, it will only
  * return this minimal chat prompt.
+ *
+ * @see https://modelfusion.dev/guide/function/generate-text/prompt-mapping#limiting-the-chat-length
  */
 export async function trimChatPrompt({ prompt, model, tokenLimit = model.contextWindowSize - model.maxCompletionTokens, }) {
     validateChatPrompt(prompt);

package/util/api/postToApi.cjs

@@ -18,10 +18,12 @@ const createJsonResponseHandler = (responseSchema) => async ({ response, url, re
 exports.createJsonResponseHandler = createJsonResponseHandler;
 const createTextResponseHandler = () => async ({ response }) => response.text();
 exports.createTextResponseHandler = createTextResponseHandler;
-const postJsonToApi = async ({ url, apiKey, body, failedResponseHandler, successfulResponseHandler, abortSignal, }) => (0, exports.postToApi)({
+const postJsonToApi = async ({ url, headers, body, failedResponseHandler, successfulResponseHandler, abortSignal, }) => (0, exports.postToApi)({
     url,
-    apiKey,
-    contentType: "application/json",
+    headers: {
+        ...headers,
+        "Content-Type": "application/json",
+    },
     body: {
         content: JSON.stringify(body),
         values: body,
@@ -31,15 +33,8 @@ const postJsonToApi = async ({ url, apiKey, body, failedResponseHandler, success
     abortSignal,
 });
 exports.postJsonToApi = postJsonToApi;
-const postToApi = async ({ url, apiKey, contentType, body, successfulResponseHandler, failedResponseHandler, abortSignal, }) => {
+const postToApi = async ({ url, headers = {}, body, successfulResponseHandler, failedResponseHandler, abortSignal, }) => {
     try {
-        const headers = {};
-        if (apiKey !== undefined) {
-            headers["Authorization"] = `Bearer ${apiKey}`;
-        }
-        if (contentType !== null) {
-            headers["Content-Type"] = contentType;
-        }
         const response = await fetch(url, {
             method: "POST",
             headers,

package/util/api/postToApi.d.ts

@@ -7,18 +7,17 @@ export type ResponseHandler<T> = (options: {
 }) => PromiseLike<T>;
 export declare const createJsonResponseHandler: <T>(responseSchema: z.ZodType<T, z.ZodTypeDef, T>) => ResponseHandler<T>;
 export declare const createTextResponseHandler: () => ResponseHandler<string>;
-export declare const postJsonToApi: <T>({ url, apiKey, body, failedResponseHandler, successfulResponseHandler, abortSignal, }: {
+export declare const postJsonToApi: <T>({ url, headers, body, failedResponseHandler, successfulResponseHandler, abortSignal, }: {
     url: string;
-    apiKey?: string | undefined;
+    headers?: Record<string, string> | undefined;
     body: unknown;
     failedResponseHandler: ResponseHandler<ApiCallError>;
     successfulResponseHandler: ResponseHandler<T>;
     abortSignal?: AbortSignal | undefined;
 }) => Promise<T>;
-export declare const postToApi: <T>({ url, apiKey, contentType, body, successfulResponseHandler, failedResponseHandler, abortSignal, }: {
+export declare const postToApi: <T>({ url, headers, body, successfulResponseHandler, failedResponseHandler, abortSignal, }: {
     url: string;
-    apiKey?: string | undefined;
-    contentType: string | null;
+    headers?: Record<string, string> | undefined;
     body: {
         content: string | FormData;
         values: unknown;

package/util/api/postToApi.js

@@ -13,10 +13,12 @@ export const createJsonResponseHandler = (responseSchema) => async ({ response,
     return parsedResult.data;
 };
 export const createTextResponseHandler = () => async ({ response }) => response.text();
-export const postJsonToApi = async ({ url, apiKey, body, failedResponseHandler, successfulResponseHandler, abortSignal, }) => postToApi({
+export const postJsonToApi = async ({ url, headers, body, failedResponseHandler, successfulResponseHandler, abortSignal, }) => postToApi({
     url,
-    apiKey,
-    contentType: "application/json",
+    headers: {
+        ...headers,
+        "Content-Type": "application/json",
+    },
     body: {
         content: JSON.stringify(body),
         values: body,
@@ -25,15 +27,8 @@ export const postJsonToApi = async ({ url, apiKey, body, failedResponseHandler,
     successfulResponseHandler,
     abortSignal,
 });
-export const postToApi = async ({ url, apiKey, contentType, body, successfulResponseHandler, failedResponseHandler, abortSignal, }) => {
+export const postToApi = async ({ url, headers = {}, body, successfulResponseHandler, failedResponseHandler, abortSignal, }) => {
     try {
-        const headers = {};
-        if (apiKey !== undefined) {
-            headers["Authorization"] = `Bearer ${apiKey}`;
-        }
-        if (contentType !== null) {
-            headers["Content-Type"] = contentType;
-        }
         const response = await fetch(url, {
             method: "POST",
             headers,

package/util/api/retryWithExponentialBackoff.cjs

@@ -20,7 +20,7 @@ async function _retryWithExponentialBackoff(f, { maxTries, delay, backoffFactor,
     catch (error) {
         const newErrors = [...errors, error];
         const tryNumber = newErrors.length;
-        if (tryNumber > maxTries) {
+        if (tryNumber >= maxTries) {
             throw new RetryError_js_1.RetryError({
                 message: `Failed after ${tryNumber} tries.`,
                 reason: "maxTriesExceeded",

package/util/api/retryWithExponentialBackoff.js

@@ -16,7 +16,7 @@ async function _retryWithExponentialBackoff(f, { maxTries, delay, backoffFactor,
     catch (error) {
         const newErrors = [...errors, error];
         const tryNumber = newErrors.length;
-        if (tryNumber > maxTries) {
+        if (tryNumber >= maxTries) {
             throw new RetryError({
                 message: `Failed after ${tryNumber} tries.`,
                 reason: "maxTriesExceeded",