feat(api): Tool calling features

Add parallel tool calling option to chat completions
Allow 'required' as a function call option

.stats.yml

@@ -1,2 +1,2 @@
 configured_endpoints: 7
-openapi_spec_url: https://storage.googleapis.com/stainless-sdk-openapi-specs/groqcloud%2Fgroqcloud-0f93f9ac6d4ad16dacaddd7608e104374c83f3bd9d0b9ed4c0273eb27ed998b7.yml
+openapi_spec_url: https://storage.googleapis.com/stainless-sdk-openapi-specs/groqcloud%2Fgroqcloud-a4c1214ecaa24ad37fbb3c12b8392787ebe0fd51c5e7e08bdf25d7608dc7900b.yml

README.md

@@ -1,6 +1,6 @@
 # Groq Node API Library
 
-[![NPM version](https://img.shields.io/npm/v/groq-sdk.svg)](https://npmjs.org/package/groq-sdk)
+[![NPM version](https://img.shields.io/npm/v/groq-sdk.svg)](https://npmjs.org/package/groq-sdk) ![npm bundle size](https://img.shields.io/bundlephobia/minzip/groq-sdk)
 
 This library provides convenient access to the Groq REST API from server-side TypeScript or JavaScript.
 
@@ -22,12 +22,14 @@ The full API of this library can be found in [api.md](api.md).
 ```js
 import Groq from 'groq-sdk';
 
-const groq = new Groq();
+const groq = new Groq({
+  apiKey: process.env['GROQ_API_KEY'], // This is the default and can be omitted
+});
 
 async function main() {
   const chatCompletion = await groq.chat.completions.create({
     messages: [{ role: 'user', content: 'Explain the importance of low latency LLMs' }],
-    model: 'mixtral-8x7b-32768',
+    model: 'llama3-8b-8192',
   });
 
   console.log(chatCompletion.choices[0].message.content);
@@ -44,15 +46,17 @@ This library includes TypeScript definitions for all request params and response
 ```ts
 import Groq from 'groq-sdk';
 
-const groq = new Groq();
+const groq = new Groq({
+  apiKey: process.env['GROQ_API_KEY'], // This is the default and can be omitted
+});
 
 async function main() {
   const params: Groq.Chat.CompletionCreateParams = {
     messages: [
       { role: 'system', content: 'You are a helpful assistant.' },
       { role: 'user', content: 'Explain the importance of low latency LLMs' },
     ],
-    model: 'mixtral-8x7b-32768',
+    model: 'llama3-8b-8192',
   };
   const chatCompletion: Groq.Chat.ChatCompletion = await groq.chat.completions.create(params);
 }
@@ -77,7 +81,7 @@ async function main() {
         { role: 'system', content: 'You are a helpful assistant.' },
         { role: 'user', content: 'Explain the importance of low latency LLMs' },
       ],
-      model: 'mixtral-8x7b-32768',
+      model: 'llama3-8b-8192',
     })
     .catch(async (err) => {
       if (err instanceof Groq.APIError) {
@@ -122,7 +126,7 @@ const groq = new Groq({
 });
 
 // Or, configure per-request:
-await groq.chat.completions.create({ messages: [{ role: 'system', content: 'You are a helpful assistant.' }, { role: 'user', content: 'Explain the importance of low latency LLMs' }], model: 'mixtral-8x7b-32768' }, {
+await groq.chat.completions.create({ messages: [{ role: 'system', content: 'You are a helpful assistant.' }, { role: 'user', content: 'Explain the importance of low latency LLMs' }], model: 'llama3-8b-8192' }, {
   maxRetries: 5,
 });
 ```
@@ -139,7 +143,7 @@ const groq = new Groq({
 });
 
 // Override per-request:
-await groq.chat.completions.create({ messages: [{ role: 'system', content: 'You are a helpful assistant.' }, { role: 'user', content: 'Explain the importance of low latency LLMs' }], model: 'mixtral-8x7b-32768' }, {
+await groq.chat.completions.create({ messages: [{ role: 'system', content: 'You are a helpful assistant.' }, { role: 'user', content: 'Explain the importance of low latency LLMs' }], model: 'llama3-8b-8192' }, {
   timeout: 5 * 1000,
 });
 ```
@@ -166,7 +170,7 @@ const response = await groq.chat.completions
       { role: 'system', content: 'You are a helpful assistant.' },
       { role: 'user', content: 'Explain the importance of low latency LLMs' },
     ],
-    model: 'mixtral-8x7b-32768',
+    model: 'llama3-8b-8192',
   })
   .asResponse();
 console.log(response.headers.get('X-My-Header'));
@@ -178,7 +182,7 @@ const { data: chatCompletion, response: raw } = await groq.chat.completions
       { role: 'system', content: 'You are a helpful assistant.' },
       { role: 'user', content: 'Explain the importance of low latency LLMs' },
     ],
-    model: 'mixtral-8x7b-32768',
+    model: 'llama3-8b-8192',
   })
   .withResponse();
 console.log(raw.headers.get('X-My-Header'));
@@ -292,7 +296,7 @@ await groq.chat.completions.create(
       { role: 'system', content: 'You are a helpful assistant.' },
       { role: 'user', content: 'Explain the importance of low latency LLMs' },
     ],
-    model: 'mixtral-8x7b-32768',
+    model: 'llama3-8b-8192',
   },
   {
     httpAgent: new http.Agent({ keepAlive: false }),

package.json

@@ -16,7 +16,6 @@
   "scripts": {
     "test": "./scripts/test",
     "build": "./scripts/build",
-    "prepack": "echo 'to pack, run yarn build && (cd dist; yarn pack)' && exit 1",
     "prepublishOnly": "echo 'to publish, run yarn build && (cd dist; yarn publish)' && exit 1",
     "format": "prettier --write --cache --cache-strategy metadata . !dist",
     "prepare": "if ./scripts/utils/check-is-in-git-install.sh; then ./scripts/build; fi",

src/core.ts

@@ -19,7 +19,7 @@ import {
   type HeadersInit,
 } from './_shims/index';
 export { type Response };
-import { isMultipartBody } from './uploads';
+import { BlobLike, isBlobLike, isMultipartBody } from './uploads';
 export {
   maybeMultipartFormRequestOptions,
   multipartFormRequestOptions,
@@ -249,7 +249,17 @@ export abstract class APIClient {
     path: string,
     opts?: PromiseOrValue<RequestOptions<Req>>,
   ): APIPromise<Rsp> {
-    return this.request(Promise.resolve(opts).then((opts) => ({ method, path, ...opts })));
+    return this.request(
+      Promise.resolve(opts).then(async (opts) => {
+        const body =
+          opts && isBlobLike(opts?.body) ? new DataView(await opts.body.arrayBuffer())
+          : opts?.body instanceof DataView ? opts.body
+          : opts?.body instanceof ArrayBuffer ? new DataView(opts.body)
+          : opts && ArrayBuffer.isView(opts?.body) ? new DataView(opts.body.buffer)
+          : opts?.body;
+        return { method, path, ...opts, body };
+      }),
+    );
   }
 
   getAPIList<Item, PageClass extends AbstractPage<Item> = AbstractPage<Item>>(
@@ -271,6 +281,8 @@ export abstract class APIClient {
         const encoded = encoder.encode(body);
         return encoded.length.toString();
       }
+    } else if (ArrayBuffer.isView(body)) {
+      return body.byteLength.toString();
     }
 
     return null;
@@ -280,7 +292,9 @@ export abstract class APIClient {
     const { method, path, query, headers: headers = {} } = options;
 
     const body =
-      isMultipartBody(options.body) ? options.body.body
+      ArrayBuffer.isView(options.body) || (options.__binaryRequest && typeof options.body === 'string') ?
+        options.body
+      : isMultipartBody(options.body) ? options.body.body
       : options.body ? JSON.stringify(options.body, null, 2)
       : null;
     const contentLength = this.calculateContentLength(body);
@@ -735,7 +749,9 @@ export type Headers = Record<string, string | null | undefined>;
 export type DefaultQuery = Record<string, string | undefined>;
 export type KeysEnum<T> = { [P in keyof Required<T>]: true };
 
-export type RequestOptions<Req = unknown | Record<string, unknown> | Readable> = {
+export type RequestOptions<
+  Req = unknown | Record<string, unknown> | Readable | BlobLike | ArrayBufferView | ArrayBuffer,
+> = {
   method?: HTTPMethod;
   path?: string;
   query?: Req | undefined;
@@ -749,6 +765,7 @@ export type RequestOptions<Req = unknown | Record<string, unknown> | Readable> =
   signal?: AbortSignal | undefined | null;
   idempotencyKey?: string;
 
+  __binaryRequest?: boolean | undefined;
   __binaryResponse?: boolean | undefined;
   __streamClass?: typeof Stream;
 };
@@ -770,6 +787,7 @@ const requestOptionsKeys: KeysEnum<RequestOptions> = {
   signal: true,
   idempotencyKey: true,
 
+  __binaryRequest: true,
   __binaryResponse: true,
   __streamClass: true,
 };
@@ -783,10 +801,11 @@ export const isRequestOptions = (obj: unknown): obj is RequestOptions => {
   );
 };
 
-export type FinalRequestOptions<Req = unknown | Record<string, unknown> | Readable> = RequestOptions<Req> & {
-  method: HTTPMethod;
-  path: string;
-};
+export type FinalRequestOptions<Req = unknown | Record<string, unknown> | Readable | DataView> =
+  RequestOptions<Req> & {
+    method: HTTPMethod;
+    path: string;
+  };
 
 declare const Deno: any;
 declare const EdgeRuntime: any;

src/resources/chat/completions.ts

@@ -623,7 +623,7 @@ export interface ChatCompletionTool {
  * `none` is the default when no tools are present. `auto` is the default if tools
  * are present.
  */
-export type ChatCompletionToolChoiceOption = 'none' | 'auto' | ChatCompletionNamedToolChoice;
+export type ChatCompletionToolChoiceOption = 'none' | 'auto' | 'required' | ChatCompletionNamedToolChoice;
 
 export interface ChatCompletionToolMessageParam {
   /**
@@ -695,7 +695,7 @@ export interface ChatCompletionCreateParamsBase {
    * `none` is the default when no functions are present. `auto` is the default if
    * functions are present.
    */
-  function_call?: 'none' | 'auto' | ChatCompletionFunctionCallOption | null;
+  function_call?: 'none' | 'auto' | 'required' | ChatCompletionFunctionCallOption | null;
 
   /**
    * Deprecated in favor of `tools`.
@@ -726,11 +726,16 @@ export interface ChatCompletionCreateParamsBase {
 
   /**
    * How many chat completion choices to generate for each input message. Note that
-   * you will be charged based on the number of generated tokens across all of the
-   * choices. Keep `n` as `1` to minimize costs.
+   * the current moment, only n=1 is supported. Other values will result in a 400
+   * response.
    */
   n?: number | null;
 
+  /**
+   * Whether to enable parallel function calling during tool use.
+   */
+  parallel_tool_calls?: boolean | null;
+
   /**
    * Number between -2.0 and 2.0. Positive values penalize new tokens based on
    * whether they appear in the text so far, increasing the model's likelihood to
@@ -841,7 +846,7 @@ export namespace CompletionCreateParams {
 
     /**
      * The parameters the functions accepts, described as a JSON Schema object. See the
-     * [guide](/docs/guides/text-generation/function-calling) for examples, and the
+     * docs on [tool use](/docs/tool-use) for examples, and the
      * [JSON Schema reference](https://json-schema.org/understanding-json-schema/) for
      * documentation about the format.
      *

src/resources/embeddings.ts

@@ -88,7 +88,7 @@ export interface EmbeddingCreateParams {
   /**
    * ID of the model to use.
    */
-  model: string;
+  model: (string & {}) | 'nomic-embed-text-v1_5';
 
   /**
    * The format to return the embeddings in. Can only be `float` or `base64`.

src/resources/shared.ts

@@ -25,7 +25,7 @@ export interface FunctionDefinition {
 
   /**
    * The parameters the functions accepts, described as a JSON Schema object. See the
-   * [guide](/docs/guides/text-generation/function-calling) for examples, and the
+   * docs on [tool use](/docs/tool-use) for examples, and the
    * [JSON Schema reference](https://json-schema.org/understanding-json-schema/) for
    * documentation about the format.
    *
@@ -36,7 +36,7 @@ export interface FunctionDefinition {
 
 /**
  * The parameters the functions accepts, described as a JSON Schema object. See the
- * [guide](/docs/guides/text-generation/function-calling) for examples, and the
+ * docs on [tool use](/docs/tool-use) for examples, and the
  * [JSON Schema reference](https://json-schema.org/understanding-json-schema/) for
  * documentation about the format.
  *

tests/api-resources/chat/completions.test.ts

@@ -11,11 +11,7 @@ const groq = new Groq({
 describe('resource completions', () => {
   test('create: only required params', async () => {
     const responsePromise = groq.chat.completions.create({
-      messages: [
-        { content: 'string', role: 'system' },
-        { content: 'string', role: 'system' },
-        { content: 'string', role: 'system' },
-      ],
+      messages: [{ content: 'string', role: 'system' }],
       model: 'string',
     });
     const rawResponse = await responsePromise.asResponse();
@@ -29,11 +25,7 @@ describe('resource completions', () => {
 
   test('create: required and optional params', async () => {
     const response = await groq.chat.completions.create({
-      messages: [
-        { content: 'string', role: 'system', name: 'string', tool_call_id: 'string' },
-        { content: 'string', role: 'system', name: 'string', tool_call_id: 'string' },
-        { content: 'string', role: 'system', name: 'string', tool_call_id: 'string' },
-      ],
+      messages: [{ content: 'string', role: 'system', name: 'string' }],
       model: 'string',
       frequency_penalty: -2,
       function_call: 'none',
@@ -46,20 +38,21 @@ describe('resource completions', () => {
       logprobs: true,
       max_tokens: 0,
       n: 1,
+      parallel_tool_calls: true,
       presence_penalty: -2,
-      response_format: { type: 'string' },
+      response_format: { type: 'json_object' },
       seed: 0,
       stop: '\n',
       stream: true,
-      temperature: 0,
+      temperature: 1,
       tool_choice: 'none',
       tools: [
-        { function: { description: 'string', name: 'string', parameters: { foo: 'bar' } }, type: 'function' },
-        { function: { description: 'string', name: 'string', parameters: { foo: 'bar' } }, type: 'function' },
-        { function: { description: 'string', name: 'string', parameters: { foo: 'bar' } }, type: 'function' },
+        { type: 'function', function: { description: 'string', name: 'string', parameters: { foo: 'bar' } } },
+        { type: 'function', function: { description: 'string', name: 'string', parameters: { foo: 'bar' } } },
+        { type: 'function', function: { description: 'string', name: 'string', parameters: { foo: 'bar' } } },
       ],
       top_logprobs: 0,
-      top_p: 0,
+      top_p: 1,
       user: 'string',
     });
   });