openai 4.96.2 → 4.98.0

package/src/resources/fine-tuning/methods.ts

@@ -0,0 +1,152 @@
+// File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+import { APIResource } from '../../resource';
+import * as GraderModelsAPI from '../graders/grader-models';
+
+export class Methods extends APIResource {}
+
+/**
+ * The hyperparameters used for the DPO fine-tuning job.
+ */
+export interface DpoHyperparameters {
+  /**
+   * Number of examples in each batch. A larger batch size means that model
+   * parameters are updated less frequently, but with lower variance.
+   */
+  batch_size?: 'auto' | number;
+
+  /**
+   * The beta value for the DPO method. A higher beta value will increase the weight
+   * of the penalty between the policy and reference model.
+   */
+  beta?: 'auto' | number;
+
+  /**
+   * Scaling factor for the learning rate. A smaller learning rate may be useful to
+   * avoid overfitting.
+   */
+  learning_rate_multiplier?: 'auto' | number;
+
+  /**
+   * The number of epochs to train the model for. An epoch refers to one full cycle
+   * through the training dataset.
+   */
+  n_epochs?: 'auto' | number;
+}
+
+/**
+ * Configuration for the DPO fine-tuning method.
+ */
+export interface DpoMethod {
+  /**
+   * The hyperparameters used for the DPO fine-tuning job.
+   */
+  hyperparameters?: DpoHyperparameters;
+}
+
+/**
+ * The hyperparameters used for the reinforcement fine-tuning job.
+ */
+export interface ReinforcementHyperparameters {
+  /**
+   * Number of examples in each batch. A larger batch size means that model
+   * parameters are updated less frequently, but with lower variance.
+   */
+  batch_size?: 'auto' | number;
+
+  /**
+   * Multiplier on amount of compute used for exploring search space during training.
+   */
+  compute_multiplier?: 'auto' | number;
+
+  /**
+   * The number of training steps between evaluation runs.
+   */
+  eval_interval?: 'auto' | number;
+
+  /**
+   * Number of evaluation samples to generate per training step.
+   */
+  eval_samples?: 'auto' | number;
+
+  /**
+   * Scaling factor for the learning rate. A smaller learning rate may be useful to
+   * avoid overfitting.
+   */
+  learning_rate_multiplier?: 'auto' | number;
+
+  /**
+   * The number of epochs to train the model for. An epoch refers to one full cycle
+   * through the training dataset.
+   */
+  n_epochs?: 'auto' | number;
+
+  /**
+   * Level of reasoning effort.
+   */
+  reasoning_effort?: 'default' | 'low' | 'medium' | 'high';
+}
+
+/**
+ * Configuration for the reinforcement fine-tuning method.
+ */
+export interface ReinforcementMethod {
+  /**
+   * The grader used for the fine-tuning job.
+   */
+  grader:
+    | GraderModelsAPI.StringCheckGrader
+    | GraderModelsAPI.TextSimilarityGrader
+    | GraderModelsAPI.PythonGrader
+    | GraderModelsAPI.ScoreModelGrader
+    | GraderModelsAPI.MultiGrader;
+
+  /**
+   * The hyperparameters used for the reinforcement fine-tuning job.
+   */
+  hyperparameters?: ReinforcementHyperparameters;
+}
+
+/**
+ * The hyperparameters used for the fine-tuning job.
+ */
+export interface SupervisedHyperparameters {
+  /**
+   * Number of examples in each batch. A larger batch size means that model
+   * parameters are updated less frequently, but with lower variance.
+   */
+  batch_size?: 'auto' | number;
+
+  /**
+   * Scaling factor for the learning rate. A smaller learning rate may be useful to
+   * avoid overfitting.
+   */
+  learning_rate_multiplier?: 'auto' | number;
+
+  /**
+   * The number of epochs to train the model for. An epoch refers to one full cycle
+   * through the training dataset.
+   */
+  n_epochs?: 'auto' | number;
+}
+
+/**
+ * Configuration for the supervised fine-tuning method.
+ */
+export interface SupervisedMethod {
+  /**
+   * The hyperparameters used for the fine-tuning job.
+   */
+  hyperparameters?: SupervisedHyperparameters;
+}
+
+export declare namespace Methods {
+  export {
+    type DpoHyperparameters as DpoHyperparameters,
+    type DpoMethod as DpoMethod,
+    type ReinforcementHyperparameters as ReinforcementHyperparameters,
+    type ReinforcementMethod as ReinforcementMethod,
+    type SupervisedHyperparameters as SupervisedHyperparameters,
+    type SupervisedMethod as SupervisedMethod,
+  };
+}

package/src/resources/graders/grader-models.ts

@@ -0,0 +1,296 @@
+// File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+import { APIResource } from '../../resource';
+import * as ResponsesAPI from '../responses/responses';
+
+export class GraderModels extends APIResource {}
+
+/**
+ * A LabelModelGrader object which uses a model to assign labels to each item in
+ * the evaluation.
+ */
+export interface LabelModelGrader {
+  input: Array<LabelModelGrader.Input>;
+
+  /**
+   * The labels to assign to each item in the evaluation.
+   */
+  labels: Array<string>;
+
+  /**
+   * The model to use for the evaluation. Must support structured outputs.
+   */
+  model: string;
+
+  /**
+   * The name of the grader.
+   */
+  name: string;
+
+  /**
+   * The labels that indicate a passing result. Must be a subset of labels.
+   */
+  passing_labels: Array<string>;
+
+  /**
+   * The object type, which is always `label_model`.
+   */
+  type: 'label_model';
+}
+
+export namespace LabelModelGrader {
+  /**
+   * A message input to the model with a role indicating instruction following
+   * hierarchy. Instructions given with the `developer` or `system` role take
+   * precedence over instructions given with the `user` role. Messages with the
+   * `assistant` role are presumed to have been generated by the model in previous
+   * interactions.
+   */
+  export interface Input {
+    /**
+     * Text inputs to the model - can contain template strings.
+     */
+    content: string | ResponsesAPI.ResponseInputText | Input.OutputText;
+
+    /**
+     * The role of the message input. One of `user`, `assistant`, `system`, or
+     * `developer`.
+     */
+    role: 'user' | 'assistant' | 'system' | 'developer';
+
+    /**
+     * The type of the message input. Always `message`.
+     */
+    type?: 'message';
+  }
+
+  export namespace Input {
+    /**
+     * A text output from the model.
+     */
+    export interface OutputText {
+      /**
+       * The text output from the model.
+       */
+      text: string;
+
+      /**
+       * The type of the output text. Always `output_text`.
+       */
+      type: 'output_text';
+    }
+  }
+}
+
+/**
+ * A MultiGrader object combines the output of multiple graders to produce a single
+ * score.
+ */
+export interface MultiGrader {
+  /**
+   * A formula to calculate the output based on grader results.
+   */
+  calculate_output: string;
+
+  graders: Record<
+    string,
+    StringCheckGrader | TextSimilarityGrader | PythonGrader | ScoreModelGrader | LabelModelGrader
+  >;
+
+  /**
+   * The name of the grader.
+   */
+  name: string;
+
+  /**
+   * The type of grader.
+   */
+  type: 'multi';
+}
+
+/**
+ * A PythonGrader object that runs a python script on the input.
+ */
+export interface PythonGrader {
+  /**
+   * The name of the grader.
+   */
+  name: string;
+
+  /**
+   * The source code of the python script.
+   */
+  source: string;
+
+  /**
+   * The object type, which is always `python`.
+   */
+  type: 'python';
+
+  /**
+   * The image tag to use for the python script.
+   */
+  image_tag?: string;
+}
+
+/**
+ * A ScoreModelGrader object that uses a model to assign a score to the input.
+ */
+export interface ScoreModelGrader {
+  /**
+   * The input text. This may include template strings.
+   */
+  input: Array<ScoreModelGrader.Input>;
+
+  /**
+   * The model to use for the evaluation.
+   */
+  model: string;
+
+  /**
+   * The name of the grader.
+   */
+  name: string;
+
+  /**
+   * The object type, which is always `score_model`.
+   */
+  type: 'score_model';
+
+  /**
+   * The range of the score. Defaults to `[0, 1]`.
+   */
+  range?: Array<number>;
+
+  /**
+   * The sampling parameters for the model.
+   */
+  sampling_params?: unknown;
+}
+
+export namespace ScoreModelGrader {
+  /**
+   * A message input to the model with a role indicating instruction following
+   * hierarchy. Instructions given with the `developer` or `system` role take
+   * precedence over instructions given with the `user` role. Messages with the
+   * `assistant` role are presumed to have been generated by the model in previous
+   * interactions.
+   */
+  export interface Input {
+    /**
+     * Text inputs to the model - can contain template strings.
+     */
+    content: string | ResponsesAPI.ResponseInputText | Input.OutputText;
+
+    /**
+     * The role of the message input. One of `user`, `assistant`, `system`, or
+     * `developer`.
+     */
+    role: 'user' | 'assistant' | 'system' | 'developer';
+
+    /**
+     * The type of the message input. Always `message`.
+     */
+    type?: 'message';
+  }
+
+  export namespace Input {
+    /**
+     * A text output from the model.
+     */
+    export interface OutputText {
+      /**
+       * The text output from the model.
+       */
+      text: string;
+
+      /**
+       * The type of the output text. Always `output_text`.
+       */
+      type: 'output_text';
+    }
+  }
+}
+
+/**
+ * A StringCheckGrader object that performs a string comparison between input and
+ * reference using a specified operation.
+ */
+export interface StringCheckGrader {
+  /**
+   * The input text. This may include template strings.
+   */
+  input: string;
+
+  /**
+   * The name of the grader.
+   */
+  name: string;
+
+  /**
+   * The string check operation to perform. One of `eq`, `ne`, `like`, or `ilike`.
+   */
+  operation: 'eq' | 'ne' | 'like' | 'ilike';
+
+  /**
+   * The reference text. This may include template strings.
+   */
+  reference: string;
+
+  /**
+   * The object type, which is always `string_check`.
+   */
+  type: 'string_check';
+}
+
+/**
+ * A TextSimilarityGrader object which grades text based on similarity metrics.
+ */
+export interface TextSimilarityGrader {
+  /**
+   * The evaluation metric to use. One of `fuzzy_match`, `bleu`, `gleu`, `meteor`,
+   * `rouge_1`, `rouge_2`, `rouge_3`, `rouge_4`, `rouge_5`, or `rouge_l`.
+   */
+  evaluation_metric:
+    | 'fuzzy_match'
+    | 'bleu'
+    | 'gleu'
+    | 'meteor'
+    | 'rouge_1'
+    | 'rouge_2'
+    | 'rouge_3'
+    | 'rouge_4'
+    | 'rouge_5'
+    | 'rouge_l';
+
+  /**
+   * The text being graded.
+   */
+  input: string;
+
+  /**
+   * The name of the grader.
+   */
+  name: string;
+
+  /**
+   * The text being graded against.
+   */
+  reference: string;
+
+  /**
+   * The type of grader.
+   */
+  type: 'text_similarity';
+}
+
+export declare namespace GraderModels {
+  export {
+    type LabelModelGrader as LabelModelGrader,
+    type MultiGrader as MultiGrader,
+    type PythonGrader as PythonGrader,
+    type ScoreModelGrader as ScoreModelGrader,
+    type StringCheckGrader as StringCheckGrader,
+    type TextSimilarityGrader as TextSimilarityGrader,
+  };
+}

package/src/resources/graders/graders.ts

@@ -0,0 +1,31 @@
+// File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+import { APIResource } from '../../resource';
+import * as GraderModelsAPI from './grader-models';
+import {
+  GraderModels,
+  LabelModelGrader,
+  MultiGrader,
+  PythonGrader,
+  ScoreModelGrader,
+  StringCheckGrader,
+  TextSimilarityGrader,
+} from './grader-models';
+
+export class Graders extends APIResource {
+  graderModels: GraderModelsAPI.GraderModels = new GraderModelsAPI.GraderModels(this._client);
+}
+
+Graders.GraderModels = GraderModels;
+
+export declare namespace Graders {
+  export {
+    GraderModels as GraderModels,
+    type LabelModelGrader as LabelModelGrader,
+    type MultiGrader as MultiGrader,
+    type PythonGrader as PythonGrader,
+    type ScoreModelGrader as ScoreModelGrader,
+    type StringCheckGrader as StringCheckGrader,
+    type TextSimilarityGrader as TextSimilarityGrader,
+  };
+}

package/src/resources/graders/index.ts

@@ -0,0 +1,12 @@
+// File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+export {
+  GraderModels,
+  type LabelModelGrader,
+  type MultiGrader,
+  type PythonGrader,
+  type ScoreModelGrader,
+  type StringCheckGrader,
+  type TextSimilarityGrader,
+} from './grader-models';
+export { Graders } from './graders';

package/src/resources/graders.ts

@@ -0,0 +1,3 @@
+// File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+export * from './graders/index';

package/src/resources/images.ts

@@ -6,6 +6,13 @@ import * as Core from '../core';
 export class Images extends APIResource {
   /**
    * Creates a variation of a given image. This endpoint only supports `dall-e-2`.
+   *
+   * @example
+   * ```ts
+   * const imagesResponse = await client.images.createVariation({
+   *   image: fs.createReadStream('otter.png'),
+   * });
+   * ```
    */
   createVariation(
     body: ImageCreateVariationParams,
@@ -17,6 +24,14 @@ export class Images extends APIResource {
   /**
    * Creates an edited or extended image given one or more source images and a
    * prompt. This endpoint only supports `gpt-image-1` and `dall-e-2`.
+   *
+   * @example
+   * ```ts
+   * const imagesResponse = await client.images.edit({
+   *   image: fs.createReadStream('path/to/file'),
+   *   prompt: 'A cute baby sea otter wearing a beret',
+   * });
+   * ```
    */
   edit(body: ImageEditParams, options?: Core.RequestOptions): Core.APIPromise<ImagesResponse> {
     return this._client.post('/images/edits', Core.multipartFormRequestOptions({ body, ...options }));
@@ -25,6 +40,13 @@ export class Images extends APIResource {
   /**
    * Creates an image given a prompt.
    * [Learn more](https://platform.openai.com/docs/guides/images).
+   *
+   * @example
+   * ```ts
+   * const imagesResponse = await client.images.generate({
+   *   prompt: 'A cute baby sea otter',
+   * });
+   * ```
    */
   generate(body: ImageGenerateParams, options?: Core.RequestOptions): Core.APIPromise<ImagesResponse> {
     return this._client.post('/images/generations', { body, ...options });
@@ -162,10 +184,13 @@ export interface ImageCreateVariationParams {
 
 export interface ImageEditParams {
   /**
-   * The image(s) to edit. Must be a supported image file or an array of images. For
-   * `gpt-image-1`, each image should be a `png`, `webp`, or `jpg` file less than
-   * 25MB. For `dall-e-2`, you can only provide one image, and it should be a square
-   * `png` file less than 4MB.
+   * The image(s) to edit. Must be a supported image file or an array of images.
+   *
+   * For `gpt-image-1`, each image should be a `png`, `webp`, or `jpg` file less than
+   * 25MB. You can provide up to 16 images.
+   *
+   * For `dall-e-2`, you can only provide one image, and it should be a square `png`
+   * file less than 4MB.
    */
   image: Core.Uploadable | Array<Core.Uploadable>;
 
@@ -175,6 +200,17 @@ export interface ImageEditParams {
    */
   prompt: string;
 
+  /**
+   * Allows to set transparency for the background of the generated image(s). This
+   * parameter is only supported for `gpt-image-1`. Must be one of `transparent`,
+   * `opaque` or `auto` (default value). When `auto` is used, the model will
+   * automatically determine the best background for the image.
+   *
+   * If `transparent`, the output format needs to support transparency, so it should
+   * be set to either `png` (default value) or `webp`.
+   */
+  background?: 'transparent' | 'opaque' | 'auto' | null;
+
   /**
    * An additional image whose fully transparent areas (e.g. where alpha is zero)
    * indicate where `image` should be edited. If there are multiple images provided,
@@ -215,7 +251,7 @@ export interface ImageEditParams {
    * (landscape), `1024x1536` (portrait), or `auto` (default value) for
    * `gpt-image-1`, and one of `256x256`, `512x512`, or `1024x1024` for `dall-e-2`.
    */
-  size?: '256x256' | '512x512' | '1024x1024' | null;
+  size?: '256x256' | '512x512' | '1024x1024' | '1536x1024' | '1024x1536' | 'auto' | null;
 
   /**
    * A unique identifier representing your end-user, which can help OpenAI to monitor

package/src/resources/index.ts

@@ -33,10 +33,7 @@ export {
   EvalListResponsesPage,
   Evals,
   type EvalCustomDataSourceConfig,
-  type EvalLabelModelGrader,
   type EvalStoredCompletionsDataSourceConfig,
-  type EvalStringCheckGrader,
-  type EvalTextSimilarityGrader,
   type EvalCreateResponse,
   type EvalRetrieveResponse,
   type EvalUpdateResponse,
@@ -57,6 +54,7 @@ export {
   type FileListParams,
 } from './files';
 export { FineTuning } from './fine-tuning/fine-tuning';
+export { Graders } from './graders/graders';
 export {
   Images,
   type Image,

package/src/resources/responses/input-items.ts

@@ -10,6 +10,16 @@ import { type CursorPageParams } from '../../pagination';
 export class InputItems extends APIResource {
   /**
    * Returns a list of input items for a given response.
+   *
+   * @example
+   * ```ts
+   * // Automatically fetches more pages as needed.
+   * for await (const responseItem of client.responses.inputItems.list(
+   *   'response_id',
+   * )) {
+   *   // ...
+   * }
+   * ```
    */
   list(
     responseId: string,