@huggingface/inference 3.13.2 → 3.14.0

package/dist/esm/snippets/getInferenceSnippets.js

@@ -0,0 +1,309 @@
+import { Template } from "@huggingface/jinja";
+import { getModelInputSnippet, inferenceSnippetLanguages, } from "@huggingface/tasks";
+import { getProviderHelper } from "../lib/getProviderHelper.js";
+import { makeRequestOptionsFromResolvedModel } from "../lib/makeRequestOptions.js";
+import { templates } from "./templates.exported.js";
+const PYTHON_CLIENTS = ["huggingface_hub", "fal_client", "requests", "openai"];
+const JS_CLIENTS = ["fetch", "huggingface.js", "openai"];
+const SH_CLIENTS = ["curl"];
+const CLIENTS = {
+    js: [...JS_CLIENTS],
+    python: [...PYTHON_CLIENTS],
+    sh: [...SH_CLIENTS],
+};
+const CLIENTS_AUTO_POLICY = {
+    js: ["huggingface.js"],
+    python: ["huggingface_hub"],
+};
+// Helpers to find + load templates
+const hasTemplate = (language, client, templateName) => templates[language]?.[client]?.[templateName] !== undefined;
+const loadTemplate = (language, client, templateName) => {
+    const template = templates[language]?.[client]?.[templateName];
+    if (!template) {
+        throw new Error(`Template not found: ${language}/${client}/${templateName}`);
+    }
+    return (data) => new Template(template).render({ ...data });
+};
+const snippetImportPythonInferenceClient = loadTemplate("python", "huggingface_hub", "importInferenceClient");
+const snippetImportRequests = loadTemplate("python", "requests", "importRequests");
+// Needed for huggingface_hub basic snippets
+const HF_PYTHON_METHODS = {
+    "audio-classification": "audio_classification",
+    "audio-to-audio": "audio_to_audio",
+    "automatic-speech-recognition": "automatic_speech_recognition",
+    "document-question-answering": "document_question_answering",
+    "feature-extraction": "feature_extraction",
+    "fill-mask": "fill_mask",
+    "image-classification": "image_classification",
+    "image-segmentation": "image_segmentation",
+    "image-to-image": "image_to_image",
+    "image-to-text": "image_to_text",
+    "object-detection": "object_detection",
+    "question-answering": "question_answering",
+    "sentence-similarity": "sentence_similarity",
+    summarization: "summarization",
+    "table-question-answering": "table_question_answering",
+    "tabular-classification": "tabular_classification",
+    "tabular-regression": "tabular_regression",
+    "text-classification": "text_classification",
+    "text-generation": "text_generation",
+    "text-to-image": "text_to_image",
+    "text-to-speech": "text_to_speech",
+    "text-to-video": "text_to_video",
+    "token-classification": "token_classification",
+    translation: "translation",
+    "visual-question-answering": "visual_question_answering",
+    "zero-shot-classification": "zero_shot_classification",
+    "zero-shot-image-classification": "zero_shot_image_classification",
+};
+// Needed for huggingface.js basic snippets
+const HF_JS_METHODS = {
+    "automatic-speech-recognition": "automaticSpeechRecognition",
+    "feature-extraction": "featureExtraction",
+    "fill-mask": "fillMask",
+    "image-classification": "imageClassification",
+    "question-answering": "questionAnswering",
+    "sentence-similarity": "sentenceSimilarity",
+    summarization: "summarization",
+    "table-question-answering": "tableQuestionAnswering",
+    "text-classification": "textClassification",
+    "text-generation": "textGeneration",
+    "text2text-generation": "textGeneration",
+    "token-classification": "tokenClassification",
+    "text-to-speech": "textToSpeech",
+    translation: "translation",
+};
+// Snippet generators
+const snippetGenerator = (templateName, inputPreparationFn) => {
+    return (model, accessToken, provider, inferenceProviderMapping, opts) => {
+        const providerModelId = inferenceProviderMapping?.providerId ?? model.id;
+        /// Hacky: hard-code conversational templates here
+        let task = model.pipeline_tag;
+        if (model.pipeline_tag &&
+            ["text-generation", "image-text-to-text"].includes(model.pipeline_tag) &&
+            model.tags.includes("conversational")) {
+            templateName = opts?.streaming ? "conversationalStream" : "conversational";
+            inputPreparationFn = prepareConversationalInput;
+            task = "conversational";
+        }
+        let providerHelper;
+        try {
+            providerHelper = getProviderHelper(provider, task);
+        }
+        catch (e) {
+            console.error(`Failed to get provider helper for ${provider} (${task})`, e);
+            return [];
+        }
+        /// Prepare inputs + make request
+        const inputs = inputPreparationFn ? inputPreparationFn(model, opts) : { inputs: getModelInputSnippet(model) };
+        const request = makeRequestOptionsFromResolvedModel(providerModelId, providerHelper, {
+            accessToken,
+            provider,
+            ...inputs,
+        }, inferenceProviderMapping, {
+            task,
+            billTo: opts?.billTo,
+        });
+        /// Parse request.info.body if not a binary.
+        /// This is the body sent to the provider. Important for snippets with raw payload (e.g curl, requests, etc.)
+        let providerInputs = inputs;
+        const bodyAsObj = request.info.body;
+        if (typeof bodyAsObj === "string") {
+            try {
+                providerInputs = JSON.parse(bodyAsObj);
+            }
+            catch (e) {
+                console.error("Failed to parse body as JSON", e);
+            }
+        }
+        /// Prepare template injection data
+        const params = {
+            accessToken,
+            authorizationHeader: request.info.headers?.Authorization,
+            baseUrl: removeSuffix(request.url, "/chat/completions"),
+            fullUrl: request.url,
+            inputs: {
+                asObj: inputs,
+                asCurlString: formatBody(inputs, "curl"),
+                asJsonString: formatBody(inputs, "json"),
+                asPythonString: formatBody(inputs, "python"),
+                asTsString: formatBody(inputs, "ts"),
+            },
+            providerInputs: {
+                asObj: providerInputs,
+                asCurlString: formatBody(providerInputs, "curl"),
+                asJsonString: formatBody(providerInputs, "json"),
+                asPythonString: formatBody(providerInputs, "python"),
+                asTsString: formatBody(providerInputs, "ts"),
+            },
+            model,
+            provider,
+            providerModelId: providerModelId ?? model.id,
+            billTo: opts?.billTo,
+        };
+        /// Iterate over clients => check if a snippet exists => generate
+        const clients = provider === "auto" ? CLIENTS_AUTO_POLICY : CLIENTS;
+        return inferenceSnippetLanguages
+            .map((language) => {
+            const langClients = clients[language] ?? [];
+            return langClients
+                .map((client) => {
+                if (!hasTemplate(language, client, templateName)) {
+                    return;
+                }
+                const template = loadTemplate(language, client, templateName);
+                if (client === "huggingface_hub" && templateName.includes("basic")) {
+                    if (!(model.pipeline_tag && model.pipeline_tag in HF_PYTHON_METHODS)) {
+                        return;
+                    }
+                    params["methodName"] = HF_PYTHON_METHODS[model.pipeline_tag];
+                }
+                if (client === "huggingface.js" && templateName.includes("basic")) {
+                    if (!(model.pipeline_tag && model.pipeline_tag in HF_JS_METHODS)) {
+                        return;
+                    }
+                    params["methodName"] = HF_JS_METHODS[model.pipeline_tag];
+                }
+                /// Generate snippet
+                let snippet = template(params).trim();
+                if (!snippet) {
+                    return;
+                }
+                /// Add import section separately
+                if (client === "huggingface_hub") {
+                    const importSection = snippetImportPythonInferenceClient({ ...params });
+                    snippet = `${importSection}\n\n${snippet}`;
+                }
+                else if (client === "requests") {
+                    const importSection = snippetImportRequests({
+                        ...params,
+                        importBase64: snippet.includes("base64"),
+                        importJson: snippet.includes("json."),
+                    });
+                    snippet = `${importSection}\n\n${snippet}`;
+                }
+                /// Snippet is ready!
+                return { language, client: client, content: snippet };
+            })
+                .filter((snippet) => snippet !== undefined);
+        })
+            .flat();
+    };
+};
+const prepareDocumentQuestionAnsweringInput = (model) => {
+    return JSON.parse(getModelInputSnippet(model));
+};
+const prepareImageToImageInput = (model) => {
+    const data = JSON.parse(getModelInputSnippet(model));
+    return { inputs: data.image, parameters: { prompt: data.prompt } };
+};
+const prepareConversationalInput = (model, opts) => {
+    return {
+        messages: opts?.messages ?? getModelInputSnippet(model),
+        ...(opts?.temperature ? { temperature: opts?.temperature } : undefined),
+        ...(opts?.max_tokens ? { max_tokens: opts?.max_tokens } : undefined),
+        ...(opts?.top_p ? { top_p: opts?.top_p } : undefined),
+    };
+};
+const prepareQuestionAnsweringInput = (model) => {
+    const data = JSON.parse(getModelInputSnippet(model));
+    return { question: data.question, context: data.context };
+};
+const prepareTableQuestionAnsweringInput = (model) => {
+    const data = JSON.parse(getModelInputSnippet(model));
+    return { query: data.query, table: JSON.stringify(data.table) };
+};
+const snippets = {
+    "audio-classification": snippetGenerator("basicAudio"),
+    "audio-to-audio": snippetGenerator("basicAudio"),
+    "automatic-speech-recognition": snippetGenerator("basicAudio"),
+    "document-question-answering": snippetGenerator("documentQuestionAnswering", prepareDocumentQuestionAnsweringInput),
+    "feature-extraction": snippetGenerator("basic"),
+    "fill-mask": snippetGenerator("basic"),
+    "image-classification": snippetGenerator("basicImage"),
+    "image-segmentation": snippetGenerator("basicImage"),
+    "image-text-to-text": snippetGenerator("conversational"),
+    "image-to-image": snippetGenerator("imageToImage", prepareImageToImageInput),
+    "image-to-text": snippetGenerator("basicImage"),
+    "object-detection": snippetGenerator("basicImage"),
+    "question-answering": snippetGenerator("questionAnswering", prepareQuestionAnsweringInput),
+    "sentence-similarity": snippetGenerator("basic"),
+    summarization: snippetGenerator("basic"),
+    "tabular-classification": snippetGenerator("tabular"),
+    "tabular-regression": snippetGenerator("tabular"),
+    "table-question-answering": snippetGenerator("tableQuestionAnswering", prepareTableQuestionAnsweringInput),
+    "text-classification": snippetGenerator("basic"),
+    "text-generation": snippetGenerator("basic"),
+    "text-to-audio": snippetGenerator("textToAudio"),
+    "text-to-image": snippetGenerator("textToImage"),
+    "text-to-speech": snippetGenerator("textToSpeech"),
+    "text-to-video": snippetGenerator("textToVideo"),
+    "text2text-generation": snippetGenerator("basic"),
+    "token-classification": snippetGenerator("basic"),
+    translation: snippetGenerator("basic"),
+    "zero-shot-classification": snippetGenerator("zeroShotClassification"),
+    "zero-shot-image-classification": snippetGenerator("zeroShotImageClassification"),
+};
+export function getInferenceSnippets(model, accessToken, provider, inferenceProviderMapping, opts) {
+    return model.pipeline_tag && model.pipeline_tag in snippets
+        ? snippets[model.pipeline_tag]?.(model, accessToken, provider, inferenceProviderMapping, opts) ?? []
+        : [];
+}
+// String manipulation helpers
+function formatBody(obj, format) {
+    switch (format) {
+        case "curl":
+            return indentString(formatBody(obj, "json"));
+        case "json":
+            /// Hacky: remove outer brackets to make is extendable in templates
+            return JSON.stringify(obj, null, 4).split("\n").slice(1, -1).join("\n");
+        case "python":
+            return indentString(Object.entries(obj)
+                .map(([key, value]) => {
+                const formattedValue = JSON.stringify(value, null, 4).replace(/"/g, '"');
+                return `${key}=${formattedValue},`;
+            })
+                .join("\n"));
+        case "ts":
+            /// Hacky: remove outer brackets to make is extendable in templates
+            return formatTsObject(obj).split("\n").slice(1, -1).join("\n");
+        default:
+            throw new Error(`Unsupported format: ${format}`);
+    }
+}
+function formatTsObject(obj, depth) {
+    depth = depth ?? 0;
+    /// Case int, boolean, string, etc.
+    if (typeof obj !== "object" || obj === null) {
+        return JSON.stringify(obj);
+    }
+    /// Case array
+    if (Array.isArray(obj)) {
+        const items = obj
+            .map((item) => {
+            const formatted = formatTsObject(item, depth + 1);
+            return `${" ".repeat(4 * (depth + 1))}${formatted},`;
+        })
+            .join("\n");
+        return `[\n${items}\n${" ".repeat(4 * depth)}]`;
+    }
+    /// Case mapping
+    const entries = Object.entries(obj);
+    const lines = entries
+        .map(([key, value]) => {
+        const formattedValue = formatTsObject(value, depth + 1);
+        const keyStr = /^[a-zA-Z_$][a-zA-Z0-9_$]*$/.test(key) ? key : `"${key}"`;
+        return `${" ".repeat(4 * (depth + 1))}${keyStr}: ${formattedValue},`;
+    })
+        .join("\n");
+    return `{\n${lines}\n${" ".repeat(4 * depth)}}`;
+}
+function indentString(str) {
+    return str
+        .split("\n")
+        .map((line) => " ".repeat(4) + line)
+        .join("\n");
+}
+function removeSuffix(str, suffix) {
+    return str.endsWith(suffix) ? str.slice(0, -suffix.length) : str;
+}

package/dist/esm/snippets/index.d.ts

@@ -0,0 +1,2 @@
+export { getInferenceSnippets, type InferenceSnippetOptions } from "./getInferenceSnippets.js";
+//# sourceMappingURL=index.d.ts.map

package/dist/esm/snippets/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/snippets/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,oBAAoB,EAAE,KAAK,uBAAuB,EAAE,MAAM,2BAA2B,CAAC"}

package/dist/esm/snippets/index.js

@@ -0,0 +1 @@
+export { getInferenceSnippets } from "./getInferenceSnippets.js";

package/dist/esm/snippets/templates.exported.d.ts

@@ -0,0 +1,2 @@
+export declare const templates: Record<string, Record<string, Record<string, string>>>;
+//# sourceMappingURL=templates.exported.d.ts.map

package/dist/esm/snippets/templates.exported.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"templates.exported.d.ts","sourceRoot":"","sources":["../../../src/snippets/templates.exported.ts"],"names":[],"mappings":"AACA,eAAO,MAAM,SAAS,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CA4EnE,CAAC"}

package/dist/esm/snippets/templates.exported.js

@@ -0,0 +1,78 @@
+// Generated file - do not edit directly
+export const templates = {
+    "js": {
+        "fetch": {
+            "basic": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
+            "basicAudio": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"audio/flac\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
+            "basicImage": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"image/jpeg\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.json();\n\treturn result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});",
+            "textToAudio": "{% if model.library_name == \"transformers\" %}\nasync function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.blob();\n    return result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    // Returns a byte object of the Audio wavform. Use it directly!\n});\n{% else %}\nasync function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n    const result = await response.json();\n    return result;\n}\n\nquery({ inputs: {{ providerInputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});\n{% endif %} ",
+            "textToImage": "async function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.blob();\n\treturn result;\n}\n\n\nquery({ {{ providerInputs.asTsString }} }).then((response) => {\n    // Use image\n});",
+            "textToSpeech": "{% if model.library_name == \"transformers\" %}\nasync function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n{% if billTo %}\n\t\t\t\t\"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n\tconst result = await response.blob();\n    return result;\n}\n\nquery({ text: {{ inputs.asObj.inputs }} }).then((response) => {\n    // Returns a byte object of the Audio wavform. Use it directly!\n});\n{% else %}\nasync function query(data) {\n\tconst response = await fetch(\n\t\t\"{{ fullUrl }}\",\n\t\t{\n\t\t\theaders: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t},\n\t\t\tmethod: \"POST\",\n\t\t\tbody: JSON.stringify(data),\n\t\t}\n\t);\n    const result = await response.json();\n    return result;\n}\n\nquery({ text: {{ inputs.asObj.inputs }} }).then((response) => {\n    console.log(JSON.stringify(response));\n});\n{% endif %} ",
+            "zeroShotClassification": "async function query(data) {\n    const response = await fetch(\n\t\t\"{{ fullUrl }}\",\n        {\n            headers: {\n\t\t\t\tAuthorization: \"{{ authorizationHeader }}\",\n                \"Content-Type\": \"application/json\",\n{% if billTo %}\n                \"X-HF-Bill-To\": \"{{ billTo }}\",\n{% endif %}         },\n            method: \"POST\",\n            body: JSON.stringify(data),\n        }\n    );\n    const result = await response.json();\n    return result;\n}\n\nquery({\n    inputs: {{ providerInputs.asObj.inputs }},\n    parameters: { candidate_labels: [\"refund\", \"legal\", \"faq\"] }\n}).then((response) => {\n    console.log(JSON.stringify(response));\n});"
+        },
+        "huggingface.js": {
+            "basic": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst output = await client.{{ methodName }}({\n\tmodel: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
+            "basicAudio": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync({{inputs.asObj.inputs}});\n\nconst output = await client.{{ methodName }}({\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
+            "basicImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync({{inputs.asObj.inputs}});\n\nconst output = await client.{{ methodName }}({\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
+            "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n{{ inputs.asTsString }}\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(chatCompletion.choices[0].message);",
+            "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = client.chatCompletionStream({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n{{ inputs.asTsString }}\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n\nfor await (const chunk of stream) {\n\tif (chunk.choices && chunk.choices.length > 0) {\n\t\tconst newContent = chunk.choices[0].delta.content;\n\t\tout += newContent;\n\t\tconsole.log(newContent);\n\t}\n}",
+            "textToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst image = await client.textToImage({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tparameters: { num_inference_steps: 5 },\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n/// Use the generated image (it's a Blob)",
+            "textToSpeech": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst audio = await client.textToSpeech({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n// Use the generated audio (it's a Blob)",
+            "textToVideo": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst video = await client.textToVideo({\n    provider: \"{{ provider }}\",\n    model: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n}{% if billTo %}, {\n    billTo: \"{{ billTo }}\",\n}{% endif %});\n// Use the generated video (it's a Blob)"
+        },
+        "openai": {
+            "conversational": "import { OpenAI } from \"openai\";\n\nconst client = new OpenAI({\n\tbaseURL: \"{{ baseUrl }}\",\n\tapiKey: \"{{ accessToken }}\",\n{% if billTo %}\n\tdefaultHeaders: {\n\t\t\"X-HF-Bill-To\": \"{{ billTo }}\" \n\t}\n{% endif %}\n});\n\nconst chatCompletion = await client.chat.completions.create({\n\tmodel: \"{{ providerModelId }}\",\n{{ inputs.asTsString }}\n});\n\nconsole.log(chatCompletion.choices[0].message);",
+            "conversationalStream": "import { OpenAI } from \"openai\";\n\nconst client = new OpenAI({\n\tbaseURL: \"{{ baseUrl }}\",\n\tapiKey: \"{{ accessToken }}\",\n{% if billTo %}\n    defaultHeaders: {\n\t\t\"X-HF-Bill-To\": \"{{ billTo }}\" \n\t}\n{% endif %}\n});\n\nconst stream = await client.chat.completions.create({\n    model: \"{{ providerModelId }}\",\n{{ inputs.asTsString }}\n    stream: true,\n});\n\nfor await (const chunk of stream) {\n    process.stdout.write(chunk.choices[0]?.delta?.content || \"\");\n}"
+        }
+    },
+    "python": {
+        "fal_client": {
+            "textToImage": "{% if provider == \"fal-ai\" %}\nimport fal_client\n\n{% if providerInputs.asObj.loras is defined and providerInputs.asObj.loras != none %}\nresult = fal_client.subscribe(\n    \"{{ providerModelId }}\",\n    arguments={\n        \"prompt\": {{ inputs.asObj.inputs }},\n        \"loras\":{{ providerInputs.asObj.loras | tojson }},\n    },\n)\n{% else %}\nresult = fal_client.subscribe(\n    \"{{ providerModelId }}\",\n    arguments={\n        \"prompt\": {{ inputs.asObj.inputs }},\n    },\n)\n{% endif %} \nprint(result)\n{% endif %} "
+        },
+        "huggingface_hub": {
+            "basic": "result = client.{{ methodName }}(\n    {{ inputs.asObj.inputs }},\n    model=\"{{ model.id }}\",\n)",
+            "basicAudio": "output = client.{{ methodName }}({{ inputs.asObj.inputs }}, model=\"{{ model.id }}\")",
+            "basicImage": "output = client.{{ methodName }}({{ inputs.asObj.inputs }}, model=\"{{ model.id }}\")",
+            "conversational": "completion = client.chat.completions.create(\n    model=\"{{ model.id }}\",\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
+            "conversationalStream": "stream = client.chat.completions.create(\n    model=\"{{ model.id }}\",\n{{ inputs.asPythonString }}\n    stream=True,\n)\n\nfor chunk in stream:\n    print(chunk.choices[0].delta.content, end=\"\") ",
+            "documentQuestionAnswering": "output = client.document_question_answering(\n    \"{{ inputs.asObj.image }}\",\n    question=\"{{ inputs.asObj.question }}\",\n    model=\"{{ model.id }}\",\n) ",
+            "imageToImage": "# output is a PIL.Image object\nimage = client.image_to_image(\n    \"{{ inputs.asObj.inputs }}\",\n    prompt=\"{{ inputs.asObj.parameters.prompt }}\",\n    model=\"{{ model.id }}\",\n) ",
+            "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n    provider=\"{{ provider }}\",\n    api_key=\"{{ accessToken }}\",\n{% if billTo %}\n    bill_to=\"{{ billTo }}\",\n{% endif %}\n)",
+            "questionAnswering": "answer = client.question_answering(\n    question=\"{{ inputs.asObj.question }}\",\n    context=\"{{ inputs.asObj.context }}\",\n    model=\"{{ model.id }}\",\n) ",
+            "tableQuestionAnswering": "answer = client.question_answering(\n    query=\"{{ inputs.asObj.query }}\",\n    table={{ inputs.asObj.table }},\n    model=\"{{ model.id }}\",\n) ",
+            "textToImage": "# output is a PIL.Image object\nimage = client.text_to_image(\n    {{ inputs.asObj.inputs }},\n    model=\"{{ model.id }}\",\n) ",
+            "textToSpeech": "# audio is returned as bytes\naudio = client.text_to_speech(\n    {{ inputs.asObj.inputs }},\n    model=\"{{ model.id }}\",\n) \n",
+            "textToVideo": "video = client.text_to_video(\n    {{ inputs.asObj.inputs }},\n    model=\"{{ model.id }}\",\n) "
+        },
+        "openai": {
+            "conversational": "from openai import OpenAI\n\nclient = OpenAI(\n    base_url=\"{{ baseUrl }}\",\n    api_key=\"{{ accessToken }}\",\n{% if billTo %}\n    default_headers={\n        \"X-HF-Bill-To\": \"{{ billTo }}\"\n    }\n{% endif %}\n)\n\ncompletion = client.chat.completions.create(\n    model=\"{{ providerModelId }}\",\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
+            "conversationalStream": "from openai import OpenAI\n\nclient = OpenAI(\n    base_url=\"{{ baseUrl }}\",\n    api_key=\"{{ accessToken }}\",\n{% if billTo %}\n    default_headers={\n        \"X-HF-Bill-To\": \"{{ billTo }}\"\n    }\n{% endif %}\n)\n\nstream = client.chat.completions.create(\n    model=\"{{ providerModelId }}\",\n{{ inputs.asPythonString }}\n    stream=True,\n)\n\nfor chunk in stream:\n    print(chunk.choices[0].delta.content, end=\"\")"
+        },
+        "requests": {
+            "basic": "def query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\noutput = query({\n    \"inputs\": {{ providerInputs.asObj.inputs }},\n}) ",
+            "basicAudio": "def query(filename):\n    with open(filename, \"rb\") as f:\n        data = f.read()\n    response = requests.post(API_URL, headers={\"Content-Type\": \"audio/flac\", **headers}, data=data)\n    return response.json()\n\noutput = query({{ providerInputs.asObj.inputs }})",
+            "basicImage": "def query(filename):\n    with open(filename, \"rb\") as f:\n        data = f.read()\n    response = requests.post(API_URL, headers={\"Content-Type\": \"image/jpeg\", **headers}, data=data)\n    return response.json()\n\noutput = query({{ providerInputs.asObj.inputs }})",
+            "conversational": "def query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\nresponse = query({\n{{ providerInputs.asJsonString }}\n})\n\nprint(response[\"choices\"][0][\"message\"])",
+            "conversationalStream": "def query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload, stream=True)\n    for line in response.iter_lines():\n        if not line.startswith(b\"data:\"):\n            continue\n        if line.strip() == b\"data: [DONE]\":\n            return\n        yield json.loads(line.decode(\"utf-8\").lstrip(\"data:\").rstrip(\"/n\"))\n\nchunks = query({\n{{ providerInputs.asJsonString }},\n    \"stream\": True,\n})\n\nfor chunk in chunks:\n    print(chunk[\"choices\"][0][\"delta\"][\"content\"], end=\"\")",
+            "documentQuestionAnswering": "def query(payload):\n    with open(payload[\"image\"], \"rb\") as f:\n        img = f.read()\n        payload[\"image\"] = base64.b64encode(img).decode(\"utf-8\")\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\noutput = query({\n    \"inputs\": {\n        \"image\": \"{{ inputs.asObj.image }}\",\n        \"question\": \"{{ inputs.asObj.question }}\",\n    },\n}) ",
+            "imageToImage": "def query(payload):\n    with open(payload[\"inputs\"], \"rb\") as f:\n        img = f.read()\n        payload[\"inputs\"] = base64.b64encode(img).decode(\"utf-8\")\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\nimage_bytes = query({\n{{ providerInputs.asJsonString }}\n})\n\n# You can access the image with PIL.Image for example\nimport io\nfrom PIL import Image\nimage = Image.open(io.BytesIO(image_bytes)) ",
+            "importRequests": "{% if importBase64 %}\nimport base64\n{% endif %}\n{% if importJson %}\nimport json\n{% endif %}\nimport requests\n\nAPI_URL = \"{{ fullUrl }}\"\nheaders = {\n    \"Authorization\": \"{{ authorizationHeader }}\",\n{% if billTo %}\n    \"X-HF-Bill-To\": \"{{ billTo }}\"\n{% endif %}\n}",
+            "tabular": "def query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\nresponse = query({\n    \"inputs\": {\n        \"data\": {{ providerInputs.asObj.inputs }}\n    },\n}) ",
+            "textToAudio": "{% if model.library_name == \"transformers\" %}\ndef query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\naudio_bytes = query({\n    \"inputs\": {{ inputs.asObj.inputs }},\n})\n# You can access the audio with IPython.display for example\nfrom IPython.display import Audio\nAudio(audio_bytes)\n{% else %}\ndef query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\naudio, sampling_rate = query({\n    \"inputs\": {{ inputs.asObj.inputs }},\n})\n# You can access the audio with IPython.display for example\nfrom IPython.display import Audio\nAudio(audio, rate=sampling_rate)\n{% endif %} ",
+            "textToImage": "{% if provider == \"hf-inference\" %}\ndef query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\nimage_bytes = query({\n    \"inputs\": {{ providerInputs.asObj.inputs }},\n})\n\n# You can access the image with PIL.Image for example\nimport io\nfrom PIL import Image\nimage = Image.open(io.BytesIO(image_bytes))\n{% endif %}",
+            "textToSpeech": "{% if model.library_name == \"transformers\" %}\ndef query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.content\n\naudio_bytes = query({\n    \"text\": {{ inputs.asObj.inputs }},\n})\n# You can access the audio with IPython.display for example\nfrom IPython.display import Audio\nAudio(audio_bytes)\n{% else %}\ndef query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\naudio, sampling_rate = query({\n    \"text\": {{ inputs.asObj.inputs }},\n})\n# You can access the audio with IPython.display for example\nfrom IPython.display import Audio\nAudio(audio, rate=sampling_rate)\n{% endif %} ",
+            "zeroShotClassification": "def query(payload):\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\noutput = query({\n    \"inputs\": {{ providerInputs.asObj.inputs }},\n    \"parameters\": {\"candidate_labels\": [\"refund\", \"legal\", \"faq\"]},\n}) ",
+            "zeroShotImageClassification": "def query(data):\n    with open(data[\"image_path\"], \"rb\") as f:\n        img = f.read()\n    payload={\n        \"parameters\": data[\"parameters\"],\n        \"inputs\": base64.b64encode(img).decode(\"utf-8\")\n    }\n    response = requests.post(API_URL, headers=headers, json=payload)\n    return response.json()\n\noutput = query({\n    \"image_path\": {{ providerInputs.asObj.inputs }},\n    \"parameters\": {\"candidate_labels\": [\"cat\", \"dog\", \"llama\"]},\n}) "
+        }
+    },
+    "sh": {
+        "curl": {
+            "basic": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    -d '{\n{{ providerInputs.asCurlString }}\n    }'",
+            "basicAudio": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: audio/flac' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    --data-binary @{{ providerInputs.asObj.inputs }}",
+            "basicImage": "curl {{ fullUrl }} \\\n    -X POST \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: image/jpeg' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    --data-binary @{{ providerInputs.asObj.inputs }}",
+            "conversational": "curl {{ fullUrl }} \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    -d '{\n{{ providerInputs.asCurlString }},\n        \"stream\": false\n    }'",
+            "conversationalStream": "curl {{ fullUrl }} \\\n    -H 'Authorization: {{ authorizationHeader }}' \\\n    -H 'Content-Type: application/json' \\\n{% if billTo %}\n    -H 'X-HF-Bill-To: {{ billTo }}' \\\n{% endif %}\n    -d '{\n{{ providerInputs.asCurlString }},\n        \"stream\": true\n    }'",
+            "zeroShotClassification": "curl {{ fullUrl }} \\\n    -X POST \\\n    -d '{\"inputs\": {{ providerInputs.asObj.inputs }}, \"parameters\": {\"candidate_labels\": [\"refund\", \"legal\", \"faq\"]}}' \\\n    -H 'Content-Type: application/json' \\\n    -H 'Authorization: {{ authorizationHeader }}'\n{% if billTo %} \\\n    -H 'X-HF-Bill-To: {{ billTo }}'\n{% endif %}"
+        }
+    }
+};

package/dist/esm/tasks/audio/audioClassification.d.ts

@@ -0,0 +1,10 @@
+import type { AudioClassificationInput, AudioClassificationOutput } from "@huggingface/tasks";
+import type { BaseArgs, Options } from "../../types.js";
+import type { LegacyAudioInput } from "./utils.js";
+export type AudioClassificationArgs = BaseArgs & (AudioClassificationInput | LegacyAudioInput);
+/**
+ * This task reads some audio input and outputs the likelihood of classes.
+ * Recommended model:  superb/hubert-large-superb-er
+ */
+export declare function audioClassification(args: AudioClassificationArgs, options?: Options): Promise<AudioClassificationOutput>;
+//# sourceMappingURL=audioClassification.d.ts.map

package/dist/esm/tasks/audio/audioClassification.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"audioClassification.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/audioClassification.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,wBAAwB,EAAE,yBAAyB,EAAE,MAAM,oBAAoB,CAAC;AAG9F,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAExD,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,YAAY,CAAC;AAGnD,MAAM,MAAM,uBAAuB,GAAG,QAAQ,GAAG,CAAC,wBAAwB,GAAG,gBAAgB,CAAC,CAAC;AAE/F;;;GAGG;AACH,wBAAsB,mBAAmB,CACxC,IAAI,EAAE,uBAAuB,EAC7B,OAAO,CAAC,EAAE,OAAO,GACf,OAAO,CAAC,yBAAyB,CAAC,CAUpC"}

package/dist/esm/tasks/audio/audioClassification.js

@@ -0,0 +1,18 @@
+import { resolveProvider } from "../../lib/getInferenceProviderMapping.js";
+import { getProviderHelper } from "../../lib/getProviderHelper.js";
+import { innerRequest } from "../../utils/request.js";
+import { preparePayload } from "./utils.js";
+/**
+ * This task reads some audio input and outputs the likelihood of classes.
+ * Recommended model:  superb/hubert-large-superb-er
+ */
+export async function audioClassification(args, options) {
+    const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+    const providerHelper = getProviderHelper(provider, "audio-classification");
+    const payload = preparePayload(args);
+    const { data: res } = await innerRequest(payload, providerHelper, {
+        ...options,
+        task: "audio-classification",
+    });
+    return providerHelper.getResponse(res);
+}

package/dist/esm/tasks/audio/audioToAudio.d.ts

@@ -0,0 +1,29 @@
+import type { BaseArgs, Options } from "../../types.js";
+import type { LegacyAudioInput } from "./utils.js";
+export type AudioToAudioArgs = (BaseArgs & {
+    /**
+     * Binary audio data
+     */
+    inputs: Blob;
+}) | LegacyAudioInput;
+export interface AudioToAudioOutputElem {
+    /**
+     * The label for the audio output (model specific)
+     */
+    label: string;
+    /**
+     * Base64 encoded audio output.
+     */
+    audio: Blob;
+}
+export interface AudioToAudioOutput {
+    blob: string;
+    "content-type": string;
+    label: string;
+}
+/**
+ * This task reads some audio input and outputs one or multiple audio files.
+ * Example model: speechbrain/sepformer-wham does audio source separation.
+ */
+export declare function audioToAudio(args: AudioToAudioArgs, options?: Options): Promise<AudioToAudioOutput[]>;
+//# sourceMappingURL=audioToAudio.d.ts.map

package/dist/esm/tasks/audio/audioToAudio.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"audioToAudio.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/audioToAudio.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAExD,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,YAAY,CAAC;AAGnD,MAAM,MAAM,gBAAgB,GACzB,CAAC,QAAQ,GAAG;IACZ;;OAEG;IACH,MAAM,EAAE,IAAI,CAAC;CACZ,CAAC,GACF,gBAAgB,CAAC;AAEpB,MAAM,WAAW,sBAAsB;IACtC;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IAEd;;OAEG;IACH,KAAK,EAAE,IAAI,CAAC;CACZ;AAED,MAAM,WAAW,kBAAkB;IAClC,IAAI,EAAE,MAAM,CAAC;IACb,cAAc,EAAE,MAAM,CAAC;IACvB,KAAK,EAAE,MAAM,CAAC;CACd;AAED;;;GAGG;AACH,wBAAsB,YAAY,CAAC,IAAI,EAAE,gBAAgB,EAAE,OAAO,CAAC,EAAE,OAAO,GAAG,OAAO,CAAC,kBAAkB,EAAE,CAAC,CAU3G"}

package/dist/esm/tasks/audio/audioToAudio.js

@@ -0,0 +1,19 @@
+import { resolveProvider } from "../../lib/getInferenceProviderMapping.js";
+import { getProviderHelper } from "../../lib/getProviderHelper.js";
+import { innerRequest } from "../../utils/request.js";
+import { preparePayload } from "./utils.js";
+/**
+ * This task reads some audio input and outputs one or multiple audio files.
+ * Example model: speechbrain/sepformer-wham does audio source separation.
+ */
+export async function audioToAudio(args, options) {
+    const model = "inputs" in args ? args.model : undefined;
+    const provider = await resolveProvider(args.provider, model);
+    const providerHelper = getProviderHelper(provider, "audio-to-audio");
+    const payload = preparePayload(args);
+    const { data: res } = await innerRequest(payload, providerHelper, {
+        ...options,
+        task: "audio-to-audio",
+    });
+    return providerHelper.getResponse(res);
+}

package/dist/esm/tasks/audio/automaticSpeechRecognition.d.ts

@@ -0,0 +1,10 @@
+import type { AutomaticSpeechRecognitionInput, AutomaticSpeechRecognitionOutput } from "@huggingface/tasks";
+import type { BaseArgs, Options } from "../../types.js";
+import type { LegacyAudioInput } from "./utils.js";
+export type AutomaticSpeechRecognitionArgs = BaseArgs & (AutomaticSpeechRecognitionInput | LegacyAudioInput);
+/**
+ * This task reads some audio input and outputs the said words within the audio files.
+ * Recommended model (english language): facebook/wav2vec2-large-960h-lv60-self
+ */
+export declare function automaticSpeechRecognition(args: AutomaticSpeechRecognitionArgs, options?: Options): Promise<AutomaticSpeechRecognitionOutput>;
+//# sourceMappingURL=automaticSpeechRecognition.d.ts.map

package/dist/esm/tasks/audio/automaticSpeechRecognition.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"automaticSpeechRecognition.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/automaticSpeechRecognition.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,+BAA+B,EAAE,gCAAgC,EAAE,MAAM,oBAAoB,CAAC;AAI5G,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAExD,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,YAAY,CAAC;AAEnD,MAAM,MAAM,8BAA8B,GAAG,QAAQ,GAAG,CAAC,+BAA+B,GAAG,gBAAgB,CAAC,CAAC;AAC7G;;;GAGG;AACH,wBAAsB,0BAA0B,CAC/C,IAAI,EAAE,8BAA8B,EACpC,OAAO,CAAC,EAAE,OAAO,GACf,OAAO,CAAC,gCAAgC,CAAC,CAa3C"}

package/dist/esm/tasks/audio/automaticSpeechRecognition.js

@@ -0,0 +1,22 @@
+import { resolveProvider } from "../../lib/getInferenceProviderMapping.js";
+import { getProviderHelper } from "../../lib/getProviderHelper.js";
+import { InferenceOutputError } from "../../lib/InferenceOutputError.js";
+import { innerRequest } from "../../utils/request.js";
+/**
+ * This task reads some audio input and outputs the said words within the audio files.
+ * Recommended model (english language): facebook/wav2vec2-large-960h-lv60-self
+ */
+export async function automaticSpeechRecognition(args, options) {
+    const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+    const providerHelper = getProviderHelper(provider, "automatic-speech-recognition");
+    const payload = await providerHelper.preparePayloadAsync(args);
+    const { data: res } = await innerRequest(payload, providerHelper, {
+        ...options,
+        task: "automatic-speech-recognition",
+    });
+    const isValidOutput = typeof res?.text === "string";
+    if (!isValidOutput) {
+        throw new InferenceOutputError("Expected {text: string}");
+    }
+    return providerHelper.getResponse(res);
+}

package/dist/esm/tasks/audio/textToSpeech.d.ts

@@ -0,0 +1,10 @@
+import type { TextToSpeechInput } from "@huggingface/tasks";
+import type { BaseArgs, Options } from "../../types.js";
+type TextToSpeechArgs = BaseArgs & TextToSpeechInput;
+/**
+ * This task synthesize an audio of a voice pronouncing a given text.
+ * Recommended model: espnet/kan-bayashi_ljspeech_vits
+ */
+export declare function textToSpeech(args: TextToSpeechArgs, options?: Options): Promise<Blob>;
+export {};
+//# sourceMappingURL=textToSpeech.d.ts.map

package/dist/esm/tasks/audio/textToSpeech.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"textToSpeech.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/textToSpeech.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAG5D,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAExD,KAAK,gBAAgB,GAAG,QAAQ,GAAG,iBAAiB,CAAC;AAKrD;;;GAGG;AACH,wBAAsB,YAAY,CAAC,IAAI,EAAE,gBAAgB,EAAE,OAAO,CAAC,EAAE,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC,CAQ3F"}

package/dist/esm/tasks/audio/textToSpeech.js

@@ -0,0 +1,16 @@
+import { resolveProvider } from "../../lib/getInferenceProviderMapping.js";
+import { getProviderHelper } from "../../lib/getProviderHelper.js";
+import { innerRequest } from "../../utils/request.js";
+/**
+ * This task synthesize an audio of a voice pronouncing a given text.
+ * Recommended model: espnet/kan-bayashi_ljspeech_vits
+ */
+export async function textToSpeech(args, options) {
+    const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
+    const providerHelper = getProviderHelper(provider, "text-to-speech");
+    const { data: res } = await innerRequest(args, providerHelper, {
+        ...options,
+        task: "text-to-speech",
+    });
+    return providerHelper.getResponse(res);
+}

package/dist/esm/tasks/audio/utils.d.ts

@@ -0,0 +1,12 @@
+import type { BaseArgs, InferenceProvider, RequestArgs } from "../../types.js";
+/**
+ * @deprecated
+ */
+export interface LegacyAudioInput {
+    data: Blob | ArrayBuffer;
+    provider?: InferenceProvider;
+}
+export declare function preparePayload(args: BaseArgs & ({
+    inputs: Blob;
+} | LegacyAudioInput)): RequestArgs;
+//# sourceMappingURL=utils.d.ts.map

package/dist/esm/tasks/audio/utils.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../../../../src/tasks/audio/utils.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,iBAAiB,EAAE,WAAW,EAAE,MAAM,gBAAgB,CAAC;AAG/E;;GAEG;AACH,MAAM,WAAW,gBAAgB;IAChC,IAAI,EAAE,IAAI,GAAG,WAAW,CAAC;IACzB,QAAQ,CAAC,EAAE,iBAAiB,CAAC;CAC7B;AAED,wBAAgB,cAAc,CAAC,IAAI,EAAE,QAAQ,GAAG,CAAC;IAAE,MAAM,EAAE,IAAI,CAAA;CAAE,GAAG,gBAAgB,CAAC,GAAG,WAAW,CAOlG"}

package/dist/esm/tasks/audio/utils.js

@@ -0,0 +1,9 @@
+import { omit } from "../../utils/omit.js";
+export function preparePayload(args) {
+    return "data" in args
+        ? args
+        : {
+            ...omit(args, "inputs"),
+            data: args.inputs,
+        };
+}