@huggingface/inference 4.13.0 → 4.13.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (24) hide show
  1. package/dist/commonjs/package.d.ts +1 -1
  2. package/dist/commonjs/package.js +1 -1
  3. package/dist/commonjs/providers/replicate.d.ts.map +1 -1
  4. package/dist/commonjs/providers/replicate.js +12 -0
  5. package/dist/commonjs/snippets/getInferenceSnippets.d.ts.map +1 -1
  6. package/dist/commonjs/snippets/getInferenceSnippets.js +2 -0
  7. package/dist/commonjs/snippets/templates.exported.js +5 -5
  8. package/dist/commonjs/tasks/nlp/chatCompletionStream.d.ts.map +1 -1
  9. package/dist/commonjs/tasks/nlp/chatCompletionStream.js +10 -2
  10. package/dist/esm/package.d.ts +1 -1
  11. package/dist/esm/package.js +1 -1
  12. package/dist/esm/providers/replicate.d.ts.map +1 -1
  13. package/dist/esm/providers/replicate.js +12 -0
  14. package/dist/esm/snippets/getInferenceSnippets.d.ts.map +1 -1
  15. package/dist/esm/snippets/getInferenceSnippets.js +2 -0
  16. package/dist/esm/snippets/templates.exported.js +5 -5
  17. package/dist/esm/tasks/nlp/chatCompletionStream.d.ts.map +1 -1
  18. package/dist/esm/tasks/nlp/chatCompletionStream.js +10 -2
  19. package/package.json +3 -3
  20. package/src/package.ts +1 -1
  21. package/src/providers/replicate.ts +14 -0
  22. package/src/snippets/getInferenceSnippets.ts +4 -0
  23. package/src/snippets/templates.exported.ts +5 -5
  24. package/src/tasks/nlp/chatCompletionStream.ts +10 -2
package/dist/commonjs/package.d.ts CHANGED
@@ -1,3 +1,3 @@
1
- export declare const PACKAGE_VERSION = "4.13.0";
1
+ export declare const PACKAGE_VERSION = "4.13.2";
2
2
  export declare const PACKAGE_NAME = "@huggingface/inference";
3
3
  //# sourceMappingURL=package.d.ts.map
package/dist/commonjs/package.js CHANGED
@@ -2,5 +2,5 @@
2
2
  Object.defineProperty(exports, "__esModule", { value: true });
3
3
  exports.PACKAGE_NAME = exports.PACKAGE_VERSION = void 0;
4
4
  // Generated file from package.json. Issues importing JSON directly when publishing on commonjs/ESM - see https://github.com/microsoft/TypeScript/issues/51783
5
- exports.PACKAGE_VERSION = "4.13.0";
5
+ exports.PACKAGE_VERSION = "4.13.2";
6
6
  exports.PACKAGE_NAME = "@huggingface/inference";
package/dist/commonjs/providers/replicate.d.ts.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"replicate.d.ts","sourceRoot":"","sources":["../../../src/providers/replicate.ts"],"names":[],"mappings":"AAkBA,OAAO,KAAK,EAAE,UAAU,EAAE,YAAY,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAEpF,OAAO,EACN,kBAAkB,EAClB,KAAK,oCAAoC,EACzC,KAAK,sBAAsB,EAC3B,KAAK,qBAAqB,EAC1B,KAAK,qBAAqB,EAC1B,MAAM,qBAAqB,CAAC;AAC7B,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,6BAA6B,CAAC;AACpE,OAAO,KAAK,EAAE,8BAA8B,EAAE,MAAM,8CAA8C,CAAC;AACnG,OAAO,KAAK,EAAE,gCAAgC,EAAE,MAAM,oBAAoB,CAAC;AAE3E,MAAM,WAAW,eAAe;IAC/B,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED,uBAAe,aAAc,SAAQ,kBAAkB;gBAC1C,GAAG,CAAC,EAAE,MAAM;IAIxB,SAAS,CAAC,MAAM,EAAE,SAAS,GAAG,MAAM;IAMpC,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAUlD,cAAc,CAAC,MAAM,EAAE,YAAY,EAAE,MAAM,EAAE,OAAO,GAAG,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC;IAQ7E,OAAO,CAAC,MAAM,EAAE,SAAS,GAAG,MAAM;CAO3C;AAED,qBAAa,wBAAyB,SAAQ,aAAc,YAAW,qBAAqB;IAClF,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAerD,WAAW,CACzB,GAAG,EAAE,eAAe,GAAG,IAAI,EAC3B,GAAG,CAAC,EAAE,MAAM,EACZ,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,EAChC,UAAU,CAAC,EAAE,KAAK,GAAG,MAAM,GAAG,MAAM,GAClC,OAAO,CAAC,MAAM,GAAG,IAAI,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAsBnD;AAED,qBAAa,yBAA0B,SAAQ,aAAa;IAClD,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAarD,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CAiBpE;AAED,qBAAa,wBAAyB,SAAQ,aAAc,YAAW,qBAAqB;IAC5E,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CAcpE;AAED,qBAAa,uCACZ,SAAQ,aACR,YAAW,oCAAoC;IAEtC,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAW9D,mBAAmB,CAAC,IAAI,EAAE,8BAA8B,GAAG,OAAO,CAAC,WAAW,CAAC;IAkBtE,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,gCAAgC,CAAC;CAuBhG;AAED,qBAAa,yBAA0B,SAAQ,aAAc,YAAW,sBAAsB;IACpF,cAAc,CAAC,MAAM,EAAE,UAAU,CAAC,gBAAgB,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAehF,mBAAmB,CAAC,IAAI,EAAE,gBAAgB,GAAG,OAAO,CAAC,WAAW,CAAC;IAcxD,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CA0BpE"}
1
+ {"version":3,"file":"replicate.d.ts","sourceRoot":"","sources":["../../../src/providers/replicate.ts"],"names":[],"mappings":"AAkBA,OAAO,KAAK,EAAE,UAAU,EAAE,YAAY,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAEpF,OAAO,EACN,kBAAkB,EAClB,KAAK,oCAAoC,EACzC,KAAK,sBAAsB,EAC3B,KAAK,qBAAqB,EAC1B,KAAK,qBAAqB,EAC1B,MAAM,qBAAqB,CAAC;AAC7B,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,6BAA6B,CAAC;AACpE,OAAO,KAAK,EAAE,8BAA8B,EAAE,MAAM,8CAA8C,CAAC;AACnG,OAAO,KAAK,EAAE,gCAAgC,EAAE,MAAM,oBAAoB,CAAC;AAE3E,MAAM,WAAW,eAAe;IAC/B,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED,uBAAe,aAAc,SAAQ,kBAAkB;gBAC1C,GAAG,CAAC,EAAE,MAAM;IAIxB,SAAS,CAAC,MAAM,EAAE,SAAS,GAAG,MAAM;IAMpC,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAUlD,cAAc,CAAC,MAAM,EAAE,YAAY,EAAE,MAAM,EAAE,OAAO,GAAG,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC;IAQ7E,OAAO,CAAC,MAAM,EAAE,SAAS,GAAG,MAAM;CAO3C;AAED,qBAAa,wBAAyB,SAAQ,aAAc,YAAW,qBAAqB;IAClF,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAerD,WAAW,CACzB,GAAG,EAAE,eAAe,GAAG,IAAI,EAC3B,GAAG,CAAC,EAAE,MAAM,EACZ,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,EAChC,UAAU,CAAC,EAAE,KAAK,GAAG,MAAM,GAAG,MAAM,GAClC,OAAO,CAAC,MAAM,GAAG,IAAI,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAoCnD;AAED,qBAAa,yBAA0B,SAAQ,aAAa;IAClD,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAarD,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CAiBpE;AAED,qBAAa,wBAAyB,SAAQ,aAAc,YAAW,qBAAqB;IAC5E,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CAcpE;AAED,qBAAa,uCACZ,SAAQ,aACR,YAAW,oCAAoC;IAEtC,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAW9D,mBAAmB,CAAC,IAAI,EAAE,8BAA8B,GAAG,OAAO,CAAC,WAAW,CAAC;IAkBtE,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,gCAAgC,CAAC;CAuBhG;AAED,qBAAa,yBAA0B,SAAQ,aAAc,YAAW,sBAAsB;IACpF,cAAc,CAAC,MAAM,EAAE,UAAU,CAAC,gBAAgB,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAehF,mBAAmB,CAAC,IAAI,EAAE,gBAAgB,GAAG,OAAO,CAAC,WAAW,CAAC;IAcxD,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CA0BpE"}
package/dist/commonjs/providers/replicate.js CHANGED
@@ -74,6 +74,18 @@ class ReplicateTextToImageTask extends ReplicateTask {
74
74
  async getResponse(res, url, headers, outputType) {
75
75
  void url;
76
76
  void headers;
77
+ // Handle string output
78
+ if (typeof res === "object" && "output" in res && typeof res.output === "string" && (0, isUrl_js_1.isUrl)(res.output)) {
79
+ if (outputType === "json") {
80
+ return { ...res };
81
+ }
82
+ if (outputType === "url") {
83
+ return res.output;
84
+ }
85
+ const urlResponse = await fetch(res.output);
86
+ return await urlResponse.blob();
87
+ }
88
+ // Handle array output
77
89
  if (typeof res === "object" &&
78
90
  "output" in res &&
79
91
  Array.isArray(res.output) &&
package/dist/commonjs/snippets/getInferenceSnippets.d.ts.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"getInferenceSnippets.d.ts","sourceRoot":"","sources":["../../../src/snippets/getInferenceSnippets.ts"],"names":[],"mappings":"AACA,OAAO,EACN,KAAK,gBAAgB,EAErB,KAAK,gBAAgB,EAGrB,MAAM,oBAAoB,CAAC;AAK5B,OAAO,KAAK,EAAE,6BAA6B,EAAE,yBAAyB,EAA8B,MAAM,aAAa,CAAC;AAKxH,MAAM,MAAM,uBAAuB,GAAG;IACrC,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,MAAM,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACjC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAmY5B,wBAAgB,oBAAoB,CACnC,KAAK,EAAE,gBAAgB,EACvB,QAAQ,EAAE,yBAAyB,EACnC,wBAAwB,CAAC,EAAE,6BAA6B,EACxD,IAAI,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAC5B,gBAAgB,EAAE,CAIpB"}
1
+ {"version":3,"file":"getInferenceSnippets.d.ts","sourceRoot":"","sources":["../../../src/snippets/getInferenceSnippets.ts"],"names":[],"mappings":"AACA,OAAO,EACN,KAAK,gBAAgB,EAErB,KAAK,gBAAgB,EAGrB,MAAM,oBAAoB,CAAC;AAK5B,OAAO,KAAK,EAAE,6BAA6B,EAAE,yBAAyB,EAA8B,MAAM,aAAa,CAAC;AAKxH,MAAM,MAAM,uBAAuB,GAAG;IACrC,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,MAAM,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACjC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAuY5B,wBAAgB,oBAAoB,CACnC,KAAK,EAAE,gBAAgB,EACvB,QAAQ,EAAE,yBAAyB,EACnC,wBAAwB,CAAC,EAAE,6BAA6B,EACxD,IAAI,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAC5B,gBAAgB,EAAE,CAIpB"}
package/dist/commonjs/snippets/getInferenceSnippets.js CHANGED
@@ -189,6 +189,8 @@ const snippetGenerator = (templateName, inputPreparationFn) => {
189
189
  : providerModelId ?? model.id,
190
190
  billTo: opts?.billTo,
191
191
  endpointUrl: opts?.endpointUrl,
192
+ task,
193
+ directRequest: !!opts?.directRequest,
192
194
  };
193
195
  /// Iterate over clients => check if a snippet exists => generate
194
196
  const clients = provider === "auto" && task !== "conversational" ? CLIENTS_NON_CONVERSATIONAL_AUTO_POLICY : CLIENTS;
package/dist/commonjs/snippets/templates.exported.js CHANGED
@@ -20,8 +20,8 @@ exports.templates = {
20
20
  "basic": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst output = await client.{{ methodName }}({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tmodel: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
21
21
  "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{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
22
22
  "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{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
23
- "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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);",
24
- "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = client.chatCompletionStream({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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}",
23
+ "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n{% if directRequest %}\n provider: \"{{ provider }}\",\n model: \"{{ model.id }}\",\n{% else %}\n model: \"{{ providerModelId }}\",\n{% endif %}\n{{ inputs.asTsString }}\n}{% if billTo %}, {\n billTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(chatCompletion.choices[0].message);",
24
+ "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = client.chatCompletionStream({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n model: \"{{ providerModelId }}\",\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}",
25
25
  "imageToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync(\"{{inputs.asObj.inputs}}\");\n\nconst image = await client.imageToImage({\n{% if endpointUrl %}\n\tendpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tprovider: \"{{provider}}\",\n\tmodel: \"{{model.id}}\",\n\tinputs: data,\n\tparameters: { prompt: \"{{inputs.asObj.parameters.prompt}}\", },\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n/// Use the generated image (it's a Blob)\n// For example, you can save it to a file or display it in an image element\n",
26
26
  "imageToVideo": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync(\"{{inputs.asObj.inputs}}\");\n\nconst video = await client.imageToVideo({\n{% if endpointUrl %}\n\tendpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tprovider: \"{{provider}}\",\n\tmodel: \"{{model.id}}\",\n\tinputs: data,\n\tparameters: { prompt: \"{{inputs.asObj.parameters.prompt}}\", },\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\n/// Use the generated video (it's a Blob)\n// For example, you can save it to a file or display it in a video element\n",
27
27
  "textToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst image = await client.textToImage({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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)",
@@ -43,12 +43,12 @@ exports.templates = {
43
43
  "basic": "result = client.{{ methodName }}(\n {{ inputs.asObj.inputs }},\n model=\"{{ model.id }}\",\n)",
44
44
  "basicAudio": "output = client.{{ methodName }}({{ inputs.asObj.inputs }}, model=\"{{ model.id }}\")",
45
45
  "basicImage": "output = client.{{ methodName }}({{ inputs.asObj.inputs }}, model=\"{{ model.id }}\")",
46
- "conversational": "completion = client.chat.completions.create(\n model=\"{{ model.id }}\",\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
47
- "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=\"\") ",
46
+ "conversational": "completion = client.chat.completions.create(\n{% if directRequest %}\n model=\"{{ model.id }}\",\n{% else %}\n model=\"{{ providerModelId }}\",\n{% endif %}\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
47
+ "conversationalStream": "stream = 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=\"\") ",
48
48
  "documentQuestionAnswering": "output = client.document_question_answering(\n \"{{ inputs.asObj.image }}\",\n question=\"{{ inputs.asObj.question }}\",\n model=\"{{ model.id }}\",\n) ",
49
49
  "imageToImage": "with open(\"{{ inputs.asObj.inputs }}\", \"rb\") as image_file:\n input_image = image_file.read()\n\n# output is a PIL.Image object\nimage = client.image_to_image(\n input_image,\n prompt=\"{{ inputs.asObj.parameters.prompt }}\",\n model=\"{{ model.id }}\",\n)\n",
50
50
  "imageToVideo": "with open(\"{{ inputs.asObj.inputs }}\", \"rb\") as image_file:\n input_image = image_file.read()\n\nvideo = client.image_to_video(\n input_image,\n prompt=\"{{ inputs.asObj.parameters.prompt }}\",\n model=\"{{ model.id }}\",\n) \n",
51
- "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n{% if endpointUrl %}\n base_url=\"{{ baseUrl }}\",\n{% endif %}\n provider=\"{{ provider }}\",\n api_key=\"{{ accessToken }}\",\n{% if billTo %}\n bill_to=\"{{ billTo }}\",\n{% endif %}\n)",
51
+ "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n{% if endpointUrl %}\n base_url=\"{{ baseUrl }}\",\n{% endif %}\n{% if task != \"conversational\" or directRequest %}\n provider=\"{{ provider }}\",\n{% endif %}\n api_key=\"{{ accessToken }}\",\n{% if billTo %}\n bill_to=\"{{ billTo }}\",\n{% endif %}\n)",
52
52
  "questionAnswering": "answer = client.question_answering(\n question=\"{{ inputs.asObj.question }}\",\n context=\"{{ inputs.asObj.context }}\",\n model=\"{{ model.id }}\",\n) ",
53
53
  "tableQuestionAnswering": "answer = client.table_question_answering(\n query=\"{{ inputs.asObj.query }}\",\n table={{ inputs.asObj.table }},\n model=\"{{ model.id }}\",\n) ",
54
54
  "textToImage": "# output is a PIL.Image object\nimage = client.text_to_image(\n {{ inputs.asObj.inputs }},\n model=\"{{ model.id }}\",\n) ",
package/dist/commonjs/tasks/nlp/chatCompletionStream.d.ts.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"chatCompletionStream.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/chatCompletionStream.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,mBAAmB,EAAE,0BAA0B,EAAE,MAAM,oBAAoB,CAAC;AAG1F,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAGxD;;GAEG;AACH,wBAAuB,oBAAoB,CAC1C,IAAI,EAAE,QAAQ,GAAG,mBAAmB,EACpC,OAAO,CAAC,EAAE,OAAO,GACf,cAAc,CAAC,0BAA0B,CAAC,CAO5C"}
1
+ {"version":3,"file":"chatCompletionStream.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/chatCompletionStream.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,mBAAmB,EAAE,0BAA0B,EAAE,MAAM,oBAAoB,CAAC;AAG1F,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAKxD;;GAEG;AACH,wBAAuB,oBAAoB,CAC1C,IAAI,EAAE,QAAQ,GAAG,mBAAmB,EACpC,OAAO,CAAC,EAAE,OAAO,GACf,cAAc,CAAC,0BAA0B,CAAC,CAa5C"}
package/dist/commonjs/tasks/nlp/chatCompletionStream.js CHANGED
@@ -4,12 +4,20 @@ exports.chatCompletionStream = chatCompletionStream;
4
4
  const getInferenceProviderMapping_js_1 = require("../../lib/getInferenceProviderMapping.js");
5
5
  const getProviderHelper_js_1 = require("../../lib/getProviderHelper.js");
6
6
  const request_js_1 = require("../../utils/request.js");
7
+ const providerHelper_js_1 = require("../../providers/providerHelper.js");
7
8
  /**
8
9
  * Use to continue text from a prompt. Same as `textGeneration` but returns generator that can be read one token at a time
9
10
  */
10
11
  async function* chatCompletionStream(args, options) {
11
- const provider = await (0, getInferenceProviderMapping_js_1.resolveProvider)(args.provider, args.model, args.endpointUrl);
12
- const providerHelper = (0, getProviderHelper_js_1.getProviderHelper)(provider, "conversational");
12
+ let providerHelper;
13
+ if (!args.provider || args.provider === "auto") {
14
+ // Special case: we have a dedicated auto-router for conversational models. No need to fetch provider mapping.
15
+ providerHelper = new providerHelper_js_1.AutoRouterConversationalTask();
16
+ }
17
+ else {
18
+ const provider = await (0, getInferenceProviderMapping_js_1.resolveProvider)(args.provider, args.model, args.endpointUrl);
19
+ providerHelper = (0, getProviderHelper_js_1.getProviderHelper)(provider, "conversational");
20
+ }
13
21
  yield* (0, request_js_1.innerStreamingRequest)(args, providerHelper, {
14
22
  ...options,
15
23
  task: "conversational",
package/dist/esm/package.d.ts CHANGED
@@ -1,3 +1,3 @@
1
- export declare const PACKAGE_VERSION = "4.13.0";
1
+ export declare const PACKAGE_VERSION = "4.13.2";
2
2
  export declare const PACKAGE_NAME = "@huggingface/inference";
3
3
  //# sourceMappingURL=package.d.ts.map
package/dist/esm/package.js CHANGED
@@ -1,3 +1,3 @@
1
1
  // Generated file from package.json. Issues importing JSON directly when publishing on commonjs/ESM - see https://github.com/microsoft/TypeScript/issues/51783
2
- export const PACKAGE_VERSION = "4.13.0";
2
+ export const PACKAGE_VERSION = "4.13.2";
3
3
  export const PACKAGE_NAME = "@huggingface/inference";
package/dist/esm/providers/replicate.d.ts.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"replicate.d.ts","sourceRoot":"","sources":["../../../src/providers/replicate.ts"],"names":[],"mappings":"AAkBA,OAAO,KAAK,EAAE,UAAU,EAAE,YAAY,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAEpF,OAAO,EACN,kBAAkB,EAClB,KAAK,oCAAoC,EACzC,KAAK,sBAAsB,EAC3B,KAAK,qBAAqB,EAC1B,KAAK,qBAAqB,EAC1B,MAAM,qBAAqB,CAAC;AAC7B,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,6BAA6B,CAAC;AACpE,OAAO,KAAK,EAAE,8BAA8B,EAAE,MAAM,8CAA8C,CAAC;AACnG,OAAO,KAAK,EAAE,gCAAgC,EAAE,MAAM,oBAAoB,CAAC;AAE3E,MAAM,WAAW,eAAe;IAC/B,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED,uBAAe,aAAc,SAAQ,kBAAkB;gBAC1C,GAAG,CAAC,EAAE,MAAM;IAIxB,SAAS,CAAC,MAAM,EAAE,SAAS,GAAG,MAAM;IAMpC,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAUlD,cAAc,CAAC,MAAM,EAAE,YAAY,EAAE,MAAM,EAAE,OAAO,GAAG,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC;IAQ7E,OAAO,CAAC,MAAM,EAAE,SAAS,GAAG,MAAM;CAO3C;AAED,qBAAa,wBAAyB,SAAQ,aAAc,YAAW,qBAAqB;IAClF,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAerD,WAAW,CACzB,GAAG,EAAE,eAAe,GAAG,IAAI,EAC3B,GAAG,CAAC,EAAE,MAAM,EACZ,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,EAChC,UAAU,CAAC,EAAE,KAAK,GAAG,MAAM,GAAG,MAAM,GAClC,OAAO,CAAC,MAAM,GAAG,IAAI,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAsBnD;AAED,qBAAa,yBAA0B,SAAQ,aAAa;IAClD,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAarD,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CAiBpE;AAED,qBAAa,wBAAyB,SAAQ,aAAc,YAAW,qBAAqB;IAC5E,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CAcpE;AAED,qBAAa,uCACZ,SAAQ,aACR,YAAW,oCAAoC;IAEtC,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAW9D,mBAAmB,CAAC,IAAI,EAAE,8BAA8B,GAAG,OAAO,CAAC,WAAW,CAAC;IAkBtE,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,gCAAgC,CAAC;CAuBhG;AAED,qBAAa,yBAA0B,SAAQ,aAAc,YAAW,sBAAsB;IACpF,cAAc,CAAC,MAAM,EAAE,UAAU,CAAC,gBAAgB,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAehF,mBAAmB,CAAC,IAAI,EAAE,gBAAgB,GAAG,OAAO,CAAC,WAAW,CAAC;IAcxD,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CA0BpE"}
1
+ {"version":3,"file":"replicate.d.ts","sourceRoot":"","sources":["../../../src/providers/replicate.ts"],"names":[],"mappings":"AAkBA,OAAO,KAAK,EAAE,UAAU,EAAE,YAAY,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAEpF,OAAO,EACN,kBAAkB,EAClB,KAAK,oCAAoC,EACzC,KAAK,sBAAsB,EAC3B,KAAK,qBAAqB,EAC1B,KAAK,qBAAqB,EAC1B,MAAM,qBAAqB,CAAC;AAC7B,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,6BAA6B,CAAC;AACpE,OAAO,KAAK,EAAE,8BAA8B,EAAE,MAAM,8CAA8C,CAAC;AACnG,OAAO,KAAK,EAAE,gCAAgC,EAAE,MAAM,oBAAoB,CAAC;AAE3E,MAAM,WAAW,eAAe;IAC/B,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED,uBAAe,aAAc,SAAQ,kBAAkB;gBAC1C,GAAG,CAAC,EAAE,MAAM;IAIxB,SAAS,CAAC,MAAM,EAAE,SAAS,GAAG,MAAM;IAMpC,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAUlD,cAAc,CAAC,MAAM,EAAE,YAAY,EAAE,MAAM,EAAE,OAAO,GAAG,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC;IAQ7E,OAAO,CAAC,MAAM,EAAE,SAAS,GAAG,MAAM;CAO3C;AAED,qBAAa,wBAAyB,SAAQ,aAAc,YAAW,qBAAqB;IAClF,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAerD,WAAW,CACzB,GAAG,EAAE,eAAe,GAAG,IAAI,EAC3B,GAAG,CAAC,EAAE,MAAM,EACZ,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,EAChC,UAAU,CAAC,EAAE,KAAK,GAAG,MAAM,GAAG,MAAM,GAClC,OAAO,CAAC,MAAM,GAAG,IAAI,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAoCnD;AAED,qBAAa,yBAA0B,SAAQ,aAAa;IAClD,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAarD,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CAiBpE;AAED,qBAAa,wBAAyB,SAAQ,aAAc,YAAW,qBAAqB;IAC5E,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CAcpE;AAED,qBAAa,uCACZ,SAAQ,aACR,YAAW,oCAAoC;IAEtC,cAAc,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAW9D,mBAAmB,CAAC,IAAI,EAAE,8BAA8B,GAAG,OAAO,CAAC,WAAW,CAAC;IAkBtE,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,gCAAgC,CAAC;CAuBhG;AAED,qBAAa,yBAA0B,SAAQ,aAAc,YAAW,sBAAsB;IACpF,cAAc,CAAC,MAAM,EAAE,UAAU,CAAC,gBAAgB,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC;IAehF,mBAAmB,CAAC,IAAI,EAAE,gBAAgB,GAAG,OAAO,CAAC,WAAW,CAAC;IAcxD,WAAW,CAAC,QAAQ,EAAE,eAAe,GAAG,OAAO,CAAC,IAAI,CAAC;CA0BpE"}
package/dist/esm/providers/replicate.js CHANGED
@@ -71,6 +71,18 @@ export class ReplicateTextToImageTask extends ReplicateTask {
71
71
  async getResponse(res, url, headers, outputType) {
72
72
  void url;
73
73
  void headers;
74
+ // Handle string output
75
+ if (typeof res === "object" && "output" in res && typeof res.output === "string" && isUrl(res.output)) {
76
+ if (outputType === "json") {
77
+ return { ...res };
78
+ }
79
+ if (outputType === "url") {
80
+ return res.output;
81
+ }
82
+ const urlResponse = await fetch(res.output);
83
+ return await urlResponse.blob();
84
+ }
85
+ // Handle array output
74
86
  if (typeof res === "object" &&
75
87
  "output" in res &&
76
88
  Array.isArray(res.output) &&
package/dist/esm/snippets/getInferenceSnippets.d.ts.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"getInferenceSnippets.d.ts","sourceRoot":"","sources":["../../../src/snippets/getInferenceSnippets.ts"],"names":[],"mappings":"AACA,OAAO,EACN,KAAK,gBAAgB,EAErB,KAAK,gBAAgB,EAGrB,MAAM,oBAAoB,CAAC;AAK5B,OAAO,KAAK,EAAE,6BAA6B,EAAE,yBAAyB,EAA8B,MAAM,aAAa,CAAC;AAKxH,MAAM,MAAM,uBAAuB,GAAG;IACrC,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,MAAM,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACjC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAmY5B,wBAAgB,oBAAoB,CACnC,KAAK,EAAE,gBAAgB,EACvB,QAAQ,EAAE,yBAAyB,EACnC,wBAAwB,CAAC,EAAE,6BAA6B,EACxD,IAAI,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAC5B,gBAAgB,EAAE,CAIpB"}
1
+ {"version":3,"file":"getInferenceSnippets.d.ts","sourceRoot":"","sources":["../../../src/snippets/getInferenceSnippets.ts"],"names":[],"mappings":"AACA,OAAO,EACN,KAAK,gBAAgB,EAErB,KAAK,gBAAgB,EAGrB,MAAM,oBAAoB,CAAC;AAK5B,OAAO,KAAK,EAAE,6BAA6B,EAAE,yBAAyB,EAA8B,MAAM,aAAa,CAAC;AAKxH,MAAM,MAAM,uBAAuB,GAAG;IACrC,SAAS,CAAC,EAAE,OAAO,CAAC;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,aAAa,CAAC,EAAE,OAAO,CAAC;IACxB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,MAAM,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACjC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAuY5B,wBAAgB,oBAAoB,CACnC,KAAK,EAAE,gBAAgB,EACvB,QAAQ,EAAE,yBAAyB,EACnC,wBAAwB,CAAC,EAAE,6BAA6B,EACxD,IAAI,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAC5B,gBAAgB,EAAE,CAIpB"}
package/dist/esm/snippets/getInferenceSnippets.js CHANGED
@@ -186,6 +186,8 @@ const snippetGenerator = (templateName, inputPreparationFn) => {
186
186
  : providerModelId ?? model.id,
187
187
  billTo: opts?.billTo,
188
188
  endpointUrl: opts?.endpointUrl,
189
+ task,
190
+ directRequest: !!opts?.directRequest,
189
191
  };
190
192
  /// Iterate over clients => check if a snippet exists => generate
191
193
  const clients = provider === "auto" && task !== "conversational" ? CLIENTS_NON_CONVERSATIONAL_AUTO_POLICY : CLIENTS;
package/dist/esm/snippets/templates.exported.js CHANGED
@@ -17,8 +17,8 @@ export const templates = {
17
17
  "basic": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst output = await client.{{ methodName }}({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tmodel: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
18
18
  "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{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
19
19
  "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{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
20
- "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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);",
21
- "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = client.chatCompletionStream({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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}",
20
+ "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n{% if directRequest %}\n provider: \"{{ provider }}\",\n model: \"{{ model.id }}\",\n{% else %}\n model: \"{{ providerModelId }}\",\n{% endif %}\n{{ inputs.asTsString }}\n}{% if billTo %}, {\n billTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(chatCompletion.choices[0].message);",
21
+ "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = client.chatCompletionStream({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n model: \"{{ providerModelId }}\",\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}",
22
22
  "imageToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync(\"{{inputs.asObj.inputs}}\");\n\nconst image = await client.imageToImage({\n{% if endpointUrl %}\n\tendpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tprovider: \"{{provider}}\",\n\tmodel: \"{{model.id}}\",\n\tinputs: data,\n\tparameters: { prompt: \"{{inputs.asObj.parameters.prompt}}\", },\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n/// Use the generated image (it's a Blob)\n// For example, you can save it to a file or display it in an image element\n",
23
23
  "imageToVideo": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync(\"{{inputs.asObj.inputs}}\");\n\nconst video = await client.imageToVideo({\n{% if endpointUrl %}\n\tendpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tprovider: \"{{provider}}\",\n\tmodel: \"{{model.id}}\",\n\tinputs: data,\n\tparameters: { prompt: \"{{inputs.asObj.parameters.prompt}}\", },\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\n/// Use the generated video (it's a Blob)\n// For example, you can save it to a file or display it in a video element\n",
24
24
  "textToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst image = await client.textToImage({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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)",
@@ -40,12 +40,12 @@ export const templates = {
40
40
  "basic": "result = client.{{ methodName }}(\n {{ inputs.asObj.inputs }},\n model=\"{{ model.id }}\",\n)",
41
41
  "basicAudio": "output = client.{{ methodName }}({{ inputs.asObj.inputs }}, model=\"{{ model.id }}\")",
42
42
  "basicImage": "output = client.{{ methodName }}({{ inputs.asObj.inputs }}, model=\"{{ model.id }}\")",
43
- "conversational": "completion = client.chat.completions.create(\n model=\"{{ model.id }}\",\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
44
- "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=\"\") ",
43
+ "conversational": "completion = client.chat.completions.create(\n{% if directRequest %}\n model=\"{{ model.id }}\",\n{% else %}\n model=\"{{ providerModelId }}\",\n{% endif %}\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
44
+ "conversationalStream": "stream = 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=\"\") ",
45
45
  "documentQuestionAnswering": "output = client.document_question_answering(\n \"{{ inputs.asObj.image }}\",\n question=\"{{ inputs.asObj.question }}\",\n model=\"{{ model.id }}\",\n) ",
46
46
  "imageToImage": "with open(\"{{ inputs.asObj.inputs }}\", \"rb\") as image_file:\n input_image = image_file.read()\n\n# output is a PIL.Image object\nimage = client.image_to_image(\n input_image,\n prompt=\"{{ inputs.asObj.parameters.prompt }}\",\n model=\"{{ model.id }}\",\n)\n",
47
47
  "imageToVideo": "with open(\"{{ inputs.asObj.inputs }}\", \"rb\") as image_file:\n input_image = image_file.read()\n\nvideo = client.image_to_video(\n input_image,\n prompt=\"{{ inputs.asObj.parameters.prompt }}\",\n model=\"{{ model.id }}\",\n) \n",
48
- "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n{% if endpointUrl %}\n base_url=\"{{ baseUrl }}\",\n{% endif %}\n provider=\"{{ provider }}\",\n api_key=\"{{ accessToken }}\",\n{% if billTo %}\n bill_to=\"{{ billTo }}\",\n{% endif %}\n)",
48
+ "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n{% if endpointUrl %}\n base_url=\"{{ baseUrl }}\",\n{% endif %}\n{% if task != \"conversational\" or directRequest %}\n provider=\"{{ provider }}\",\n{% endif %}\n api_key=\"{{ accessToken }}\",\n{% if billTo %}\n bill_to=\"{{ billTo }}\",\n{% endif %}\n)",
49
49
  "questionAnswering": "answer = client.question_answering(\n question=\"{{ inputs.asObj.question }}\",\n context=\"{{ inputs.asObj.context }}\",\n model=\"{{ model.id }}\",\n) ",
50
50
  "tableQuestionAnswering": "answer = client.table_question_answering(\n query=\"{{ inputs.asObj.query }}\",\n table={{ inputs.asObj.table }},\n model=\"{{ model.id }}\",\n) ",
51
51
  "textToImage": "# output is a PIL.Image object\nimage = client.text_to_image(\n {{ inputs.asObj.inputs }},\n model=\"{{ model.id }}\",\n) ",
package/dist/esm/tasks/nlp/chatCompletionStream.d.ts.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"file":"chatCompletionStream.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/chatCompletionStream.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,mBAAmB,EAAE,0BAA0B,EAAE,MAAM,oBAAoB,CAAC;AAG1F,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAGxD;;GAEG;AACH,wBAAuB,oBAAoB,CAC1C,IAAI,EAAE,QAAQ,GAAG,mBAAmB,EACpC,OAAO,CAAC,EAAE,OAAO,GACf,cAAc,CAAC,0BAA0B,CAAC,CAO5C"}
1
+ {"version":3,"file":"chatCompletionStream.d.ts","sourceRoot":"","sources":["../../../../src/tasks/nlp/chatCompletionStream.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,mBAAmB,EAAE,0BAA0B,EAAE,MAAM,oBAAoB,CAAC;AAG1F,OAAO,KAAK,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAKxD;;GAEG;AACH,wBAAuB,oBAAoB,CAC1C,IAAI,EAAE,QAAQ,GAAG,mBAAmB,EACpC,OAAO,CAAC,EAAE,OAAO,GACf,cAAc,CAAC,0BAA0B,CAAC,CAa5C"}
package/dist/esm/tasks/nlp/chatCompletionStream.js CHANGED
@@ -1,12 +1,20 @@
1
1
  import { resolveProvider } from "../../lib/getInferenceProviderMapping.js";
2
2
  import { getProviderHelper } from "../../lib/getProviderHelper.js";
3
3
  import { innerStreamingRequest } from "../../utils/request.js";
4
+ import { AutoRouterConversationalTask } from "../../providers/providerHelper.js";
4
5
  /**
5
6
  * Use to continue text from a prompt. Same as `textGeneration` but returns generator that can be read one token at a time
6
7
  */
7
8
  export async function* chatCompletionStream(args, options) {
8
- const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
9
- const providerHelper = getProviderHelper(provider, "conversational");
9
+ let providerHelper;
10
+ if (!args.provider || args.provider === "auto") {
11
+ // Special case: we have a dedicated auto-router for conversational models. No need to fetch provider mapping.
12
+ providerHelper = new AutoRouterConversationalTask();
13
+ }
14
+ else {
15
+ const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
16
+ providerHelper = getProviderHelper(provider, "conversational");
17
+ }
10
18
  yield* innerStreamingRequest(args, providerHelper, {
11
19
  ...options,
12
20
  task: "conversational",
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@huggingface/inference",
3
- "version": "4.13.0",
3
+ "version": "4.13.2",
4
4
  "license": "MIT",
5
5
  "author": "Hugging Face and Tim Mikeladze <tim.mikeladze@gmail.com>",
6
6
  "description": "Typescript client for the Hugging Face Inference Providers and Inference Endpoints",
@@ -40,8 +40,8 @@
40
40
  },
41
41
  "type": "module",
42
42
  "dependencies": {
43
- "@huggingface/tasks": "^0.19.58",
44
- "@huggingface/jinja": "^0.5.1"
43
+ "@huggingface/jinja": "^0.5.1",
44
+ "@huggingface/tasks": "^0.19.63"
45
45
  },
46
46
  "devDependencies": {
47
47
  "@types/node": "18.13.0"
package/src/package.ts CHANGED
@@ -1,3 +1,3 @@
1
1
  // Generated file from package.json. Issues importing JSON directly when publishing on commonjs/ESM - see https://github.com/microsoft/TypeScript/issues/51783
2
- export const PACKAGE_VERSION = "4.13.0";
2
+ export const PACKAGE_VERSION = "4.13.2";
3
3
  export const PACKAGE_NAME = "@huggingface/inference";
package/src/providers/replicate.ts CHANGED
@@ -95,6 +95,20 @@ export class ReplicateTextToImageTask extends ReplicateTask implements TextToIma
95
95
  ): Promise<string | Blob | Record<string, unknown>> {
96
96
  void url;
97
97
  void headers;
98
+
99
+ // Handle string output
100
+ if (typeof res === "object" && "output" in res && typeof res.output === "string" && isUrl(res.output)) {
101
+ if (outputType === "json") {
102
+ return { ...res };
103
+ }
104
+ if (outputType === "url") {
105
+ return res.output;
106
+ }
107
+ const urlResponse = await fetch(res.output);
108
+ return await urlResponse.blob();
109
+ }
110
+
111
+ // Handle array output
98
112
  if (
99
113
  typeof res === "object" &&
100
114
  "output" in res &&
package/src/snippets/getInferenceSnippets.ts CHANGED
@@ -60,6 +60,8 @@ interface TemplateParams {
60
60
  importBase64?: boolean; // specific to snippetImportRequests
61
61
  importJson?: boolean; // specific to snippetImportRequests
62
62
  endpointUrl?: string;
63
+ task?: InferenceTask;
64
+ directRequest?: boolean;
63
65
  }
64
66
 
65
67
  // Helpers to find + load templates
@@ -263,6 +265,8 @@ const snippetGenerator = (templateName: string, inputPreparationFn?: InputPrepar
263
265
  : providerModelId ?? model.id,
264
266
  billTo: opts?.billTo,
265
267
  endpointUrl: opts?.endpointUrl,
268
+ task,
269
+ directRequest: !!opts?.directRequest,
266
270
  };
267
271
 
268
272
  /// Iterate over clients => check if a snippet exists => generate
package/src/snippets/templates.exported.ts CHANGED
@@ -17,8 +17,8 @@ export const templates: Record<string, Record<string, Record<string, string>>> =
17
17
  "basic": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst output = await client.{{ methodName }}({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tmodel: \"{{ model.id }}\",\n\tinputs: {{ inputs.asObj.inputs }},\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
18
18
  "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{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
19
19
  "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{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tdata,\n\tmodel: \"{{ model.id }}\",\n\tprovider: \"{{ provider }}\",\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(output);",
20
- "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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);",
21
- "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = client.chatCompletionStream({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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}",
20
+ "conversational": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst chatCompletion = await client.chatCompletion({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n{% if directRequest %}\n provider: \"{{ provider }}\",\n model: \"{{ model.id }}\",\n{% else %}\n model: \"{{ providerModelId }}\",\n{% endif %}\n{{ inputs.asTsString }}\n}{% if billTo %}, {\n billTo: \"{{ billTo }}\",\n}{% endif %});\n\nconsole.log(chatCompletion.choices[0].message);",
21
+ "conversationalStream": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nlet out = \"\";\n\nconst stream = client.chatCompletionStream({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n model: \"{{ providerModelId }}\",\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}",
22
22
  "imageToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync(\"{{inputs.asObj.inputs}}\");\n\nconst image = await client.imageToImage({\n{% if endpointUrl %}\n\tendpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tprovider: \"{{provider}}\",\n\tmodel: \"{{model.id}}\",\n\tinputs: data,\n\tparameters: { prompt: \"{{inputs.asObj.parameters.prompt}}\", },\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n/// Use the generated image (it's a Blob)\n// For example, you can save it to a file or display it in an image element\n",
23
23
  "imageToVideo": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst data = fs.readFileSync(\"{{inputs.asObj.inputs}}\");\n\nconst video = await client.imageToVideo({\n{% if endpointUrl %}\n\tendpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\n\tprovider: \"{{provider}}\",\n\tmodel: \"{{model.id}}\",\n\tinputs: data,\n\tparameters: { prompt: \"{{inputs.asObj.parameters.prompt}}\", },\n}{% if billTo %}, {\n\tbillTo: \"{{ billTo }}\",\n}{% endif %});\n\n/// Use the generated video (it's a Blob)\n// For example, you can save it to a file or display it in a video element\n",
24
24
  "textToImage": "import { InferenceClient } from \"@huggingface/inference\";\n\nconst client = new InferenceClient(\"{{ accessToken }}\");\n\nconst image = await client.textToImage({\n{% if endpointUrl %}\n endpointUrl: \"{{ endpointUrl }}\",\n{% endif %}\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)",
@@ -40,12 +40,12 @@ export const templates: Record<string, Record<string, Record<string, string>>> =
40
40
  "basic": "result = client.{{ methodName }}(\n {{ inputs.asObj.inputs }},\n model=\"{{ model.id }}\",\n)",
41
41
  "basicAudio": "output = client.{{ methodName }}({{ inputs.asObj.inputs }}, model=\"{{ model.id }}\")",
42
42
  "basicImage": "output = client.{{ methodName }}({{ inputs.asObj.inputs }}, model=\"{{ model.id }}\")",
43
- "conversational": "completion = client.chat.completions.create(\n model=\"{{ model.id }}\",\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
44
- "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=\"\") ",
43
+ "conversational": "completion = client.chat.completions.create(\n{% if directRequest %}\n model=\"{{ model.id }}\",\n{% else %}\n model=\"{{ providerModelId }}\",\n{% endif %}\n{{ inputs.asPythonString }}\n)\n\nprint(completion.choices[0].message) ",
44
+ "conversationalStream": "stream = 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=\"\") ",
45
45
  "documentQuestionAnswering": "output = client.document_question_answering(\n \"{{ inputs.asObj.image }}\",\n question=\"{{ inputs.asObj.question }}\",\n model=\"{{ model.id }}\",\n) ",
46
46
  "imageToImage": "with open(\"{{ inputs.asObj.inputs }}\", \"rb\") as image_file:\n input_image = image_file.read()\n\n# output is a PIL.Image object\nimage = client.image_to_image(\n input_image,\n prompt=\"{{ inputs.asObj.parameters.prompt }}\",\n model=\"{{ model.id }}\",\n)\n",
47
47
  "imageToVideo": "with open(\"{{ inputs.asObj.inputs }}\", \"rb\") as image_file:\n input_image = image_file.read()\n\nvideo = client.image_to_video(\n input_image,\n prompt=\"{{ inputs.asObj.parameters.prompt }}\",\n model=\"{{ model.id }}\",\n) \n",
48
- "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n{% if endpointUrl %}\n base_url=\"{{ baseUrl }}\",\n{% endif %}\n provider=\"{{ provider }}\",\n api_key=\"{{ accessToken }}\",\n{% if billTo %}\n bill_to=\"{{ billTo }}\",\n{% endif %}\n)",
48
+ "importInferenceClient": "from huggingface_hub import InferenceClient\n\nclient = InferenceClient(\n{% if endpointUrl %}\n base_url=\"{{ baseUrl }}\",\n{% endif %}\n{% if task != \"conversational\" or directRequest %}\n provider=\"{{ provider }}\",\n{% endif %}\n api_key=\"{{ accessToken }}\",\n{% if billTo %}\n bill_to=\"{{ billTo }}\",\n{% endif %}\n)",
49
49
  "questionAnswering": "answer = client.question_answering(\n question=\"{{ inputs.asObj.question }}\",\n context=\"{{ inputs.asObj.context }}\",\n model=\"{{ model.id }}\",\n) ",
50
50
  "tableQuestionAnswering": "answer = client.table_question_answering(\n query=\"{{ inputs.asObj.query }}\",\n table={{ inputs.asObj.table }},\n model=\"{{ model.id }}\",\n) ",
51
51
  "textToImage": "# output is a PIL.Image object\nimage = client.text_to_image(\n {{ inputs.asObj.inputs }},\n model=\"{{ model.id }}\",\n) ",
package/src/tasks/nlp/chatCompletionStream.ts CHANGED
@@ -3,6 +3,8 @@ import { resolveProvider } from "../../lib/getInferenceProviderMapping.js";
3
3
  import { getProviderHelper } from "../../lib/getProviderHelper.js";
4
4
  import type { BaseArgs, Options } from "../../types.js";
5
5
  import { innerStreamingRequest } from "../../utils/request.js";
6
+ import type { ConversationalTaskHelper, TaskProviderHelper } from "../../providers/providerHelper.js";
7
+ import { AutoRouterConversationalTask } from "../../providers/providerHelper.js";
6
8
 
7
9
  /**
8
10
  * Use to continue text from a prompt. Same as `textGeneration` but returns generator that can be read one token at a time
@@ -11,8 +13,14 @@ export async function* chatCompletionStream(
11
13
  args: BaseArgs & ChatCompletionInput,
12
14
  options?: Options
13
15
  ): AsyncGenerator<ChatCompletionStreamOutput> {
14
- const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
15
- const providerHelper = getProviderHelper(provider, "conversational");
16
+ let providerHelper: ConversationalTaskHelper & TaskProviderHelper;
17
+ if (!args.provider || args.provider === "auto") {
18
+ // Special case: we have a dedicated auto-router for conversational models. No need to fetch provider mapping.
19
+ providerHelper = new AutoRouterConversationalTask();
20
+ } else {
21
+ const provider = await resolveProvider(args.provider, args.model, args.endpointUrl);
22
+ providerHelper = getProviderHelper(provider, "conversational");
23
+ }
16
24
  yield* innerStreamingRequest<ChatCompletionStreamOutput>(args, providerHelper, {
17
25
  ...options,
18
26
  task: "conversational",