feat(api): support storing chat completions, enabling evals and model…

… distillation in the dashboard (#1112)

Learn more at http://openai.com/devday2024

src/resources/chat/chat.ts

@@ -16,6 +16,7 @@ export type ChatModel =
   | 'gpt-4o'
   | 'gpt-4o-2024-08-06'
   | 'gpt-4o-2024-05-13'
+  | 'gpt-4o-realtime-preview-2024-10-01'
   | 'chatgpt-4o-latest'
   | 'gpt-4o-mini'
   | 'gpt-4o-mini-2024-07-18'

src/resources/chat/completions.ts

@@ -727,8 +727,12 @@ export type ChatCompletionCreateParams =
 
 export interface ChatCompletionCreateParamsBase {
   /**
-   * A list of messages comprising the conversation so far.
-   * [Example Python code](https://cookbook.openai.com/examples/how_to_format_inputs_to_chatgpt_models).
+   * A list of messages comprising the conversation so far. Depending on the
+   * [model](https://platform.openai.com/docs/models) you use, different message
+   * types (modalities) are supported, like
+   * [text](https://platform.openai.com/docs/guides/text-generation),
+   * [images](https://platform.openai.com/docs/guides/vision), and
+   * [audio](https://platform.openai.com/docs/guides/audio).
    */
   messages: Array<ChatCompletionMessageParam>;
 
@@ -806,6 +810,12 @@ export interface ChatCompletionCreateParamsBase {
    */
   max_tokens?: number | null;
 
+  /**
+   * Developer-defined tags and values used for filtering completions in the
+   * [dashboard](https://platform.openai.com/completions).
+   */
+  metadata?: Record<string, string> | null;
+
   /**
    * 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
@@ -889,6 +899,12 @@ export interface ChatCompletionCreateParamsBase {
    */
   stop?: string | null | Array<string>;
 
+  /**
+   * Whether or not to store the output of this completion request for traffic
+   * logging in the [dashboard](https://platform.openai.com/completions).
+   */
+  store?: boolean | null;
+
   /**
    * If set, partial message deltas will be sent, like in ChatGPT. Tokens will be
    * sent as data-only

src/resources/completions.ts

@@ -125,18 +125,43 @@ export interface CompletionUsage {
    * Breakdown of tokens used in a completion.
    */
   completion_tokens_details?: CompletionUsage.CompletionTokensDetails;
+
+  /**
+   * Breakdown of tokens used in the prompt.
+   */
+  prompt_tokens_details?: CompletionUsage.PromptTokensDetails;
 }
 
 export namespace CompletionUsage {
   /**
    * Breakdown of tokens used in a completion.
    */
   export interface CompletionTokensDetails {
+    /**
+     * Audio input tokens generated by the model.
+     */
+    audio_tokens?: number;
+
     /**
      * Tokens generated by the model for reasoning.
      */
     reasoning_tokens?: number;
   }
+
+  /**
+   * Breakdown of tokens used in the prompt.
+   */
+  export interface PromptTokensDetails {
+    /**
+     * Audio input tokens present in the prompt.
+     */
+    audio_tokens?: number;
+
+    /**
+     * Cached tokens present in the prompt.
+     */
+    cached_tokens?: number;
+  }
 }
 
 export type CompletionCreateParams = CompletionCreateParamsNonStreaming | CompletionCreateParamsStreaming;

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

@@ -34,13 +34,15 @@ describe('resource completions', () => {
       logprobs: true,
       max_completion_tokens: 0,
       max_tokens: 0,
+      metadata: { foo: 'string' },
       n: 1,
       parallel_tool_calls: true,
       presence_penalty: -2,
       response_format: { type: 'text' },
       seed: -9007199254740991,
       service_tier: 'auto',
       stop: 'string',
+      store: true,
       stream: false,
       stream_options: { include_usage: true },
       temperature: 1,