feat(api): api update

Author: stainless-app[bot]

.stats.yml

@@ -1,4 +1,4 @@
 configured_endpoints: 35
-openapi_spec_url: https://storage.googleapis.com/stainless-sdk-openapi-specs/togetherai%2Ftogetherai-7d7d90647aab5086171c692e905111458f67092e1a424eb5fb13024579758c42.yml
-openapi_spec_hash: d9020f812eee9eb0d050c760bf97510b
+openapi_spec_url: https://storage.googleapis.com/stainless-sdk-openapi-specs/togetherai%2Ftogetherai-98498fdf25ae4e0659b9e7251f44484cbdd1c994164a1a9baf18a16ef5e0267a.yml
+openapi_spec_hash: ddd9362aae34ac124dca824c679a7e43
 config_hash: 2887483246d24a361689cbd1258db7cf

src/resources/chat/completions.ts

@@ -380,6 +380,13 @@ export interface CompletionCreateParamsBase {
    */
   presence_penalty?: number;
 
+  /**
+   * Controls the level of reasoning effort the model should apply when generating
+   * responses. Higher values may result in more thoughtful and detailed responses
+   * but may take longer to generate.
+   */
+  reasoning_effort?: 'low' | 'medium' | 'high';
+
   /**
    * A number that controls the diversity of generated text by reducing the
    * likelihood of repeated sequences. Higher values decrease repetition.

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

@@ -36,6 +36,7 @@ describe('resource completions', () => {
       min_p: 0,
       n: 1,
       presence_penalty: 0,
+      reasoning_effort: 'medium',
       repetition_penalty: 0,
       response_format: { schema: { foo: 'bar' }, type: 'json' },
       safety_model: 'safety_model_name',