Fix: patch streaming

src/core.ts

@@ -1,4 +1,5 @@
 import { VERSION } from './version';
+import { Stream } from './lib/streaming';
 import {
   GroqError,
   APIError,
@@ -38,6 +39,19 @@ type APIResponseProps = {
 
 async function defaultParseResponse<T>(props: APIResponseProps): Promise<T> {
   const { response } = props;
+  if (props.options.stream) {
+    debug('response', response.status, response.url, response.headers, response.body);
+
+    // Note: there is an invariant here that isn't represented in the type system
+    // that if you set `stream: true` the response type must also be `Stream<T>`
+
+    if (props.options.__streamClass) {
+      return props.options.__streamClass.fromSSEResponse(response, props.controller) as any;
+    }
+
+    return Stream.fromSSEResponse(response, props.controller) as any;
+  }
+
   // fetch refuses to read the body when the status code is 204.
   if (response.status === 204) {
     return null as T;
@@ -736,6 +750,7 @@ export type RequestOptions<Req = unknown | Record<string, unknown> | Readable> =
   idempotencyKey?: string;
 
   __binaryResponse?: boolean | undefined;
+  __streamClass?: typeof Stream;
 };
 
 // This is required so that we can determine if a given object matches the RequestOptions
@@ -756,6 +771,7 @@ const requestOptionsKeys: KeysEnum<RequestOptions> = {
   idempotencyKey: true,
 
   __binaryResponse: true,
+  __streamClass: true,
 };
 
 export const isRequestOptions = (obj: unknown): obj is RequestOptions => {

src/lib/streaming.ts

@@ -1,7 +1,7 @@
-import { ReadableStream, type Response } from 'groq-sdk/_shims/index';
-import { GroqError } from 'groq-sdk/error';
+import { ReadableStream, type Response } from '../_shims/index';
+import { GroqError } from '../error';
 
-import { APIError } from 'groq-sdk/error';
+import { APIError } from '../error';
 
 type Bytes = string | ArrayBuffer | Uint8Array | Buffer | null | undefined;
 

src/resources/chat/completions.ts

@@ -5,13 +5,33 @@ import { APIResource } from '../../resource';
 import * as ChatCompletionsAPI from './completions';
 import * as CompletionsAPI from '../completions';
 import * as Shared from '../shared';
+import { Stream } from '../../lib/streaming';
 
 export class Completions extends APIResource {
   /**
    * Creates a model response for the given chat conversation.
    */
-  create(body: CompletionCreateParams, options?: Core.RequestOptions): Core.APIPromise<ChatCompletion> {
-    return this._client.post('/openai/v1/chat/completions', { body, ...options });
+  create(
+    body: ChatCompletionCreateParamsNonStreaming,
+    options?: Core.RequestOptions,
+  ): Core.APIPromise<ChatCompletion>;
+  create(
+    body: ChatCompletionCreateParamsStreaming,
+    options?: Core.RequestOptions,
+  ): Core.APIPromise<Stream<ChatCompletionChunk>>;
+  create(
+    body: ChatCompletionCreateParamsBase,
+    options?: Core.RequestOptions,
+  ): Core.APIPromise<Stream<ChatCompletionChunk> | ChatCompletion>;
+  create(
+    body: ChatCompletionCreateParams,
+    options?: Core.RequestOptions,
+  ): Core.APIPromise<ChatCompletion> | Core.APIPromise<Stream<ChatCompletionChunk>> {
+    return this._client.post('/openai/v1/chat/completions', {
+      body,
+      ...options,
+      stream: body.stream ?? false,
+    }) as Core.APIPromise<ChatCompletion> | Core.APIPromise<Stream<ChatCompletionChunk>>;
   }
 }
 
@@ -205,7 +225,7 @@ export namespace ChatCompletionChunk {
      * number of tokens specified in the request was reached, `tool_calls` if the model
      * called a tool, or `function_call` (deprecated) if the model called a function.
      */
-    finish_reason: 'stop' | 'length' | 'tool_calls' | 'content_filter' | 'function_call' | null;
+    finish_reason: 'stop' | 'length' | 'tool_calls' | 'function_call' | null;
 
     /**
      * The index of the choice in the list of choices.
@@ -640,7 +660,11 @@ export interface ChatCompletionUserMessageParam {
   name?: string;
 }
 
-export interface CompletionCreateParams {
+export type ChatCompletionCreateParams =
+  | ChatCompletionCreateParamsNonStreaming
+  | ChatCompletionCreateParamsStreaming;
+
+export interface ChatCompletionCreateParamsBase {
   /**
    * A list of messages comprising the conversation so far.
    */
@@ -843,6 +867,26 @@ export namespace CompletionCreateParams {
   }
 }
 
+export interface ChatCompletionCreateParamsNonStreaming extends ChatCompletionCreateParamsBase {
+  /**
+   * If set, partial message deltas will be sent. Tokens will be sent as data-only
+   * [server-sent events](https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events#Event_stream_format)
+   * as they become available, with the stream terminated by a `data: [DONE]`
+   * message. [Example code](/docs/text-chat#streaming-a-chat-completion).
+   */
+  stream?: false | null;
+}
+
+export interface ChatCompletionCreateParamsStreaming extends ChatCompletionCreateParamsBase {
+  /**
+   * If set, partial message deltas will be sent. Tokens will be sent as data-only
+   * [server-sent events](https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events#Event_stream_format)
+   * as they become available, with the stream terminated by a `data: [DONE]`
+   * message. [Example code](/docs/text-chat#streaming-a-chat-completion).
+   */
+  stream: true;
+}
+
 export namespace Completions {
   export import ChatCompletion = ChatCompletionsAPI.ChatCompletion;
   export import ChatCompletionAssistantMessageParam = ChatCompletionsAPI.ChatCompletionAssistantMessageParam;