notdiamond 2.0.0-rc2 → 2.0.0-rc20

package/src/resources/prompt-adaptation.ts

@@ -3,7 +3,6 @@
 import { APIResource } from '../core/resource';
 import * as PromptAdaptationAPI from './prompt-adaptation';
 import { APIPromise } from '../core/api-promise';
-import { buildHeaders } from '../internal/headers';
 import { RequestOptions } from '../internal/request-options';
 import { path } from '../internal/utils/path';
 
@@ -12,14 +11,14 @@ export class PromptAdaptation extends APIResource {
    * Adapt your prompt from one LLM to work optimally across different target LLMs.
    *
    * This endpoint automatically optimizes your prompt (system prompt + user message
-   * template) to achieve better performance when switching between different
-   * language models. Each model has unique characteristics, and what works well for
-   * GPT-4 might not work as well for Claude or Gemini.
+   * template) to improve accuracy on your use case across various models. Each model
+   * has unique characteristics, and what works well for GPT-5 might not work as well
+   * for Claude or Gemini.
    *
    * **How Prompt Adaptation Works:**
    *
-   * 1. You provide your current prompt optimized for an origin model
-   * 2. You specify target models you want to adapt to
+   * 1. You provide your current prompt and optionally your current origin model
+   * 2. You specify the target models you want to adapt your prompt to
    * 3. You provide evaluation examples (golden records) with expected answers
    * 4. The system runs optimization to find the best prompt for each target model
    * 5. You receive adapted prompts that perform well on your target models
@@ -27,14 +26,19 @@ export class PromptAdaptation extends APIResource {
    * **Evaluation Metrics:** Choose either a standard metric or provide custom
    * evaluation:
    *
-   * - **Standard metrics**: LLMaaJ:SQL, LLMaaJ:Sem_Sim_1/3/10 (semantic similarity),
-   *   JSON_Match
+   * - **Standard metrics**: LLMaaJ:Sem_Sim_1 (semantic similarity), JSON_Match
    * - **Custom evaluation**: Provide evaluation_config with your own LLM judge,
    *   prompt, and cutoff
    *
    * **Dataset Requirements:**
    *
-   * - Minimum 5 examples in train_goldens (more examples = better adaptation)
+   * - Minimum 25 examples in train_goldens (more examples = better adaptation)
+   * - **Prototype mode**: Set `prototype_mode: true` to use as few as 3 examples for
+   *   prototyping
+   *   - Recommended when you don't have enough data yet to build a proof-of-concept
+   *   - Note: Performance may be degraded compared to standard mode (25+ examples)
+   *   - Trade-off: Faster iteration with less data vs. potentially less
+   *     generalizability
    * - Each example must have fields matching your template placeholders
    * - Supervised evaluation requires 'answer' field in each golden record
    * - Unsupervised evaluation can work without answers
@@ -45,21 +49,6 @@ export class PromptAdaptation extends APIResource {
    * - Time depends on: number of target models, dataset size, model availability
    * - Use the returned adaptation_run_id to check status and retrieve results
    *
-   * **Subscription Tiers:**
-   *
-   * - Free: 1 target model
-   * - Starter: 3 target models
-   * - Startup: 5 target models
-   * - Enterprise: 10 target models
-   *
-   * **Best Practices:**
-   *
-   * 1. Use diverse, representative examples from your production workload
-   * 2. Include 10-20 examples for best results (5 minimum)
-   * 3. Ensure consistent evaluation across all examples
-   * 4. Test both train_goldens and test_goldens split for validation
-   * 5. Use the same model versions you'll use in production
-   *
    * **Example Workflow:**
    *
    * ```
@@ -80,7 +69,7 @@ export class PromptAdaptation extends APIResource {
    *   system_prompt: 'You are a helpful assistant that answers questions accurately.',
    *   target_models: [
    *     { provider: 'anthropic', model: 'claude-sonnet-4-5-20250929' },
-   *     { provider: 'google', model: 'gemini-1.5-pro' },
+   *     { provider: 'google', model: 'gemini-2.5-flash' },
    *   ],
    *   template: 'Question: {question}\nAnswer:',
    *   evaluation_metric: 'LLMaaJ:Sem_Sim_3',
@@ -155,15 +144,6 @@ export class PromptAdaptation extends APIResource {
    * 3. Apply the optimized prompts when calling the respective target models
    * 4. Compare pre/post optimization scores to see improvement
    *
-   * **Evaluation Scores:**
-   *
-   * - Scores range from 0-10 (higher is better)
-   * - Compare origin_model score with target_models pre_optimization_score for
-   *   baseline
-   * - Compare pre_optimization_score with post_optimization_score to see improvement
-   *   from adaptation
-   * - Typical improvements range from 5-30% on evaluation metrics
-   *
    * **Status Handling:**
    *
    * - If adaptation is still processing, target model results will have
@@ -187,61 +167,17 @@ export class PromptAdaptation extends APIResource {
    *
    * @example
    * ```ts
-   * const adaptationRunResults =
+   * const response =
    *   await client.promptAdaptation.getAdaptResults(
    *     'adaptation_run_id',
    *   );
    * ```
    */
-  getAdaptResults(adaptationRunID: string, options?: RequestOptions): APIPromise<AdaptationRunResults> {
-    return this._client.get(path`/v2/prompt/adaptResults/${adaptationRunID}`, options);
-  }
-
-  /**
-   * Get Adapt Run Results
-   *
-   * @example
-   * ```ts
-   * const adaptationRunResults =
-   *   await client.promptAdaptation.getAdaptRunResults(
-   *     'adaptation_run_id',
-   *     { user_id: 'user_id', 'x-token': 'x-token' },
-   *   );
-   * ```
-   */
-  getAdaptRunResults(
+  getAdaptResults(
     adaptationRunID: string,
-    params: PromptAdaptationGetAdaptRunResultsParams,
-    options?: RequestOptions,
-  ): APIPromise<AdaptationRunResults> {
-    const { user_id, 'x-token': xToken } = params;
-    return this._client.get(path`/v2/prompt/frontendAdaptRunResults/${user_id}/${adaptationRunID}`, {
-      ...options,
-      headers: buildHeaders([{ 'x-token': xToken }, options?.headers]),
-    });
-  }
-
-  /**
-   * Get Adapt Runs
-   *
-   * @example
-   * ```ts
-   * const adaptationRunResults =
-   *   await client.promptAdaptation.getAdaptRuns('user_id', {
-   *     'x-token': 'x-token',
-   *   });
-   * ```
-   */
-  getAdaptRuns(
-    userID: string,
-    params: PromptAdaptationGetAdaptRunsParams,
     options?: RequestOptions,
-  ): APIPromise<PromptAdaptationGetAdaptRunsResponse> {
-    const { 'x-token': xToken } = params;
-    return this._client.get(path`/v2/prompt/frontendAdaptRuns/${userID}`, {
-      ...options,
-      headers: buildHeaders([{ 'x-token': xToken }, options?.headers]),
-    });
+  ): APIPromise<PromptAdaptationGetAdaptResultsResponse> {
+    return this._client.get(path`/v2/prompt/adaptResults/${adaptationRunID}`, options);
   }
 
   /**
@@ -292,28 +228,143 @@ export class PromptAdaptation extends APIResource {
   }
 
   /**
-   * Get LLM costs for a specific adaptation run
+   * Get LLM usage costs for a specific prompt adaptation run.
+   *
+   * This endpoint returns the total cost and detailed usage records for all LLM
+   * requests made during a prompt adaptation run. Use this to track costs associated
+   * with optimizing prompts for different target models.
+   *
+   * **Cost Breakdown:**
+   *
+   * - Total cost across all models used in the adaptation
+   * - Individual usage records with provider, model, tokens, and costs
+   * - Timestamps for each LLM request
+   *
+   * **Access Control:**
+   *
+   * - Only accessible by the user who created the adaptation run
+   * - Requires prompt adaptation access
    *
    * @example
    * ```ts
-   * const response =
-   *   await client.promptAdaptation.retrieveCosts(
-   *     'adaptation_run_id',
-   *   );
+   * const response = await client.promptAdaptation.getCost(
+   *   'adaptation_run_id',
+   * );
    * ```
    */
-  retrieveCosts(
-    adaptationRunID: string,
-    options?: RequestOptions,
-  ): APIPromise<PromptAdaptationRetrieveCostsResponse> {
-    return this._client.get(path`/v1/adaptation-runs/${adaptationRunID}/costs`, options);
+  getCost(adaptationRunID: string, options?: RequestOptions): APIPromise<PromptAdaptationGetCostResponse> {
+    return this._client.get(path`/v2/prompt/adapt/${adaptationRunID}/costs`, options);
   }
 }
 
 /**
- * Complete results for a prompt adaptation run including all target models.
+ * A training or test example for prompt adaptation.
  */
-export interface AdaptationRunResults {
+export interface GoldenRecord {
+  /**
+   * Dictionary mapping field names to their values. Keys must match the fields
+   * specified in the template
+   */
+  fields: { [key: string]: string };
+
+  /**
+   * Expected answer for supervised evaluation. Required for supervised metrics,
+   * optional for unsupervised
+   */
+  answer?: string | null;
+}
+
+/**
+ * Status enum for asynchronous jobs (prompt adaptation, custom router training,
+ * etc.).
+ *
+ * Represents the current state of a long-running operation:
+ *
+ * - **created**: Job has been initialized but not yet queued
+ * - **queued**: Job is waiting in the queue to be processed
+ * - **processing**: Job is currently being executed
+ * - **completed**: Job finished successfully and results are available
+ * - **failed**: Job encountered an error and did not complete
+ */
+export type JobStatus = 'created' | 'queued' | 'processing' | 'completed' | 'failed';
+
+/**
+ * Model for specifying an LLM provider in API requests.
+ */
+export interface RequestProvider {
+  /**
+   * Model name (e.g., 'gpt-4o', 'claude-sonnet-4-5-20250929')
+   */
+  model: string;
+
+  /**
+   * Provider name (e.g., 'openai', 'anthropic', 'google')
+   */
+  provider: string;
+
+  /**
+   * Maximum context length for the model (required for custom models)
+   */
+  context_length?: number | null;
+
+  /**
+   * Input token price per million tokens in USD (required for custom models)
+   */
+  input_price?: number | null;
+
+  /**
+   * Whether this is a custom model not in Not Diamond's supported model list
+   */
+  is_custom?: boolean;
+
+  /**
+   * Average latency in seconds (required for custom models)
+   */
+  latency?: number | null;
+
+  /**
+   * Output token price per million tokens in USD (required for custom models)
+   */
+  output_price?: number | null;
+}
+
+/**
+ * Response model for POST /v2/prompt/adapt endpoint.
+ *
+ * Returned immediately after submitting a prompt adaptation request. The
+ * adaptation process runs asynchronously, so use the returned adaptation_run_id to
+ * track progress and retrieve results when complete.
+ *
+ * **Next steps:**
+ *
+ * 1. Store the adaptation_run_id
+ * 2. Poll GET /v2/prompt/adaptStatus/{adaptation_run_id} to check progress
+ * 3. When status is 'completed', retrieve optimized prompts from GET
+ *    /v2/prompt/adaptResults/{adaptation_run_id}
+ * 4. Use the optimized prompts with your target models
+ */
+export interface PromptAdaptationAdaptResponse {
+  /**
+   * Unique identifier for this adaptation run. Use this to poll status and retrieve
+   * optimized prompts when complete
+   */
+  adaptation_run_id: string;
+}
+
+/**
+ * Response model for GET /v2/prompt/adaptResults/{adaptation_run_id} endpoint.
+ *
+ * Contains the complete results of a prompt adaptation run, including optimized
+ * prompts and evaluation metrics for all target models. Use this to retrieve your
+ * adapted prompts after the adaptation status is 'completed'.
+ *
+ * The response includes:
+ *
+ * - Baseline performance of your original prompt on the origin model
+ * - Optimized prompts for each target model with pre/post optimization scores
+ * - Evaluation metrics and cost information for each model
+ */
+export interface PromptAdaptationGetAdaptResultsResponse {
   /**
    * Unique ID for this adaptation run
    */
@@ -325,14 +376,14 @@ export interface AdaptationRunResults {
   created_at: string;
 
   /**
-   * Overall status of the adaptation run
+   * Overall status of the adaptation run (queued, running, completed, failed)
    */
   job_status: JobStatus;
 
   /**
-   * Results for each target model with optimized prompts
+   * Results for each target model with optimized prompts and improvement scores
    */
-  target_models: Array<AdaptationRunResults.TargetModel>;
+  target_models: Array<PromptAdaptationGetAdaptResultsResponse.TargetModel>;
 
   /**
    * Timestamp of last update to this adaptation run
@@ -344,19 +395,58 @@ export interface AdaptationRunResults {
   evaluation_metric?: string | null;
 
   /**
-   * Metrics for the LLM requests made during the adaptation run
+   * Metrics for the LLM requests made during the adaptation run (e.g.,
+   * total_requests, avg_latency)
    */
   llm_request_metrics?: { [key: string]: number };
 
   /**
-   * Results for the origin model (baseline performance)
+   * Baseline results for the origin model in prompt adaptation.
+   *
+   * Part of AdaptationRunResultsResponse. Contains the performance metrics and
+   * prompt configuration for your original prompt on the origin model. This serves
+   * as the baseline to compare against optimized prompts for target models.
+   *
+   * **Fields include:**
+   *
+   * - Original system prompt and user message template
+   * - Baseline performance score and evaluation metrics
+   * - Cost of running the baseline evaluation
+   * - Job status for the origin model evaluation
    */
-  origin_model?: AdaptationRunResults.OriginModel | null;
+  origin_model?: PromptAdaptationGetAdaptResultsResponse.OriginModel | null;
+
+  /**
+   * Whether this adaptation run was created with prototype mode (3-24 training
+   * examples allowed). Prototype mode may have degraded performance compared to
+   * standard mode (25+ examples)
+   */
+  prototype_mode?: boolean;
 }
 
-export namespace AdaptationRunResults {
+export namespace PromptAdaptationGetAdaptResultsResponse {
   /**
-   * Results for a single target model adaptation.
+   * Optimized prompt results for a single target model in prompt adaptation.
+   *
+   * Part of AdaptationRunResultsResponse. Contains the optimized system prompt and
+   * user message template for a specific target model, along with performance scores
+   * before and after optimization. Use these optimized prompts with the target model
+   * to achieve better performance than the original prompt.
+   *
+   * **Key metrics:**
+   *
+   * - **pre_optimization_score**: Performance with original prompt on this target
+   *   model
+   * - **post_optimization_score**: Performance with optimized prompt on this target
+   *   model
+   * - **Score improvement**: post - pre shows how much optimization helped
+   *
+   * **Usage:**
+   *
+   * 1. Extract the optimized system_prompt and user_message_template
+   * 2. Replace placeholders in user_message_template using fields from your data
+   * 3. Use these prompts when calling this target model
+   * 4. Compare pre/post scores to see improvement gained
    */
   export interface TargetModel {
     cost: number | null;
@@ -371,31 +461,54 @@ export namespace AdaptationRunResults {
 
     pre_optimization_score: number | null;
 
+    task_type: string | null;
+
     /**
-     * Status of this specific target model adaptation
+     * Status enum for asynchronous jobs (prompt adaptation, custom router training,
+     * etc.).
+     *
+     * Represents the current state of a long-running operation:
+     *
+     * - **created**: Job has been initialized but not yet queued
+     * - **queued**: Job is waiting in the queue to be processed
+     * - **processing**: Job is currently being executed
+     * - **completed**: Job finished successfully and results are available
+     * - **failed**: Job encountered an error and did not complete
      */
-    result_status: PromptAdaptationAPI.JobStatus | null;
+    result_status?: PromptAdaptationAPI.JobStatus | null;
 
     /**
-     * Optimized system prompt for this target model
+     * Optimized system prompt for this target model. Use this as the system message in
+     * your LLM calls
      */
-    system_prompt: string | null;
-
-    task_type: string | null;
+    system_prompt?: string | null;
 
     /**
-     * Optimized user message template for this target model
+     * Optimized user message template with placeholders. Substitute fields using your
+     * data before calling the LLM
      */
-    user_message_template: string | null;
+    user_message_template?: string | null;
 
     /**
-     * Field names used in the optimized template
+     * List of field names to substitute in the template (e.g., ['question',
+     * 'context']). These match the curly-brace placeholders in user_message_template
      */
-    user_message_template_fields: Array<string> | null;
+    user_message_template_fields?: Array<string> | null;
   }
 
   /**
-   * Results for the origin model (baseline performance)
+   * Baseline results for the origin model in prompt adaptation.
+   *
+   * Part of AdaptationRunResultsResponse. Contains the performance metrics and
+   * prompt configuration for your original prompt on the origin model. This serves
+   * as the baseline to compare against optimized prompts for target models.
+   *
+   * **Fields include:**
+   *
+   * - Original system prompt and user message template
+   * - Baseline performance score and evaluation metrics
+   * - Cost of running the baseline evaluation
+   * - Job status for the origin model evaluation
    */
   export interface OriginModel {
     cost: number | null;
@@ -404,85 +517,170 @@ export namespace AdaptationRunResults {
 
     model_name: string | null;
 
-    result_status: PromptAdaptationAPI.JobStatus | null;
-
     score: number | null;
 
-    system_prompt: string | null;
-
-    user_message_template: string | null;
-  }
-}
+    /**
+     * Status enum for asynchronous jobs (prompt adaptation, custom router training,
+     * etc.).
+     *
+     * Represents the current state of a long-running operation:
+     *
+     * - **created**: Job has been initialized but not yet queued
+     * - **queued**: Job is waiting in the queue to be processed
+     * - **processing**: Job is currently being executed
+     * - **completed**: Job finished successfully and results are available
+     * - **failed**: Job encountered an error and did not complete
+     */
+    result_status?: PromptAdaptationAPI.JobStatus | null;
 
-export type JobStatus = 'created' | 'queued' | 'processing' | 'completed' | 'failed';
+    /**
+     * Original system prompt used for the origin model
+     */
+    system_prompt?: string | null;
 
-/**
- * Response from prompt adaptation request.
- */
-export interface PromptAdaptationAdaptResponse {
-  /**
-   * Unique ID for this adaptation run. Use this to check status and retrieve results
-   */
-  adaptation_run_id: string;
+    /**
+     * Original user message template used for the origin model
+     */
+    user_message_template?: string | null;
+  }
 }
 
-export type PromptAdaptationGetAdaptRunsResponse = Array<AdaptationRunResults>;
-
 /**
- * Status response for a prompt adaptation run.
+ * Response model for GET /v2/prompt/adaptStatus/{adaptation_run_id} endpoint.
+ *
+ * Returns the current status of an asynchronous prompt adaptation job. Poll this
+ * endpoint periodically to track progress. When status is 'completed', you can
+ * retrieve the optimized prompts using the /adaptResults endpoint.
+ *
+ * **Status values:**
+ *
+ * - **created**: Job has been initialized
+ * - **queued**: Waiting in queue (check queue_position for your place in line)
+ * - **processing**: Currently running optimization
+ * - **completed**: Finished successfully, results available via /adaptResults
+ * - **failed**: Encountered an error during processing
+ *
+ * **Polling recommendations:**
+ *
+ * - Poll every 30-60 seconds while status is incomplete
+ * - Stop polling once status is 'completed' or 'failed'
+ * - Adaptation typically takes 10-30 minutes total
  */
 export interface PromptAdaptationGetAdaptStatusResponse {
   /**
-   * Unique ID for this adaptation run. Use this to check status and retrieve results
+   * Unique identifier for this adaptation run. Use this to poll status and retrieve
+   * optimized prompts when complete
    */
   adaptation_run_id: string;
 
   /**
-   * Current status of the adaptation run (created, queued, processing, completed, or
-   * failed)
+   * Current status of the adaptation run. Poll until this is 'completed' or 'failed'
    */
   status: JobStatus;
 
   /**
-   * Position in queue if status is 'queued'. Lower numbers mean earlier processing
+   * Position in queue when status is 'queued'. Lower numbers process sooner. Null
+   * when not queued
    */
   queue_position?: number | null;
 }
 
-export interface PromptAdaptationRetrieveCostsResponse {
+/**
+ * Response model for GET /v2/prompt/adapt/{adaptation_run_id}/costs endpoint.
+ *
+ * Contains the total LLM costs and detailed usage records for a prompt adaptation
+ * run. Use this to track costs associated with optimizing prompts for different
+ * target models.
+ */
+export interface PromptAdaptationGetCostResponse {
+  /**
+   * Unique identifier for the adaptation run
+   */
   adaptation_run_id: string;
 
+  /**
+   * Total cost in USD across all LLM requests in this adaptation run
+   */
   total_cost: number;
 
-  usage_records: Array<PromptAdaptationRetrieveCostsResponse.UsageRecord>;
+  /**
+   * Detailed usage records for each LLM request made during the adaptation
+   */
+  usage_records: Array<PromptAdaptationGetCostResponse.UsageRecord>;
 }
 
-export namespace PromptAdaptationRetrieveCostsResponse {
+export namespace PromptAdaptationGetCostResponse {
+  /**
+   * Individual LLM usage record with token counts and cost breakdown.
+   *
+   * Returned by GET /llm-usage endpoint and included in AdaptationRunCostResponse.
+   * Each record represents a single LLM API call with detailed usage metrics.
+   */
   export interface UsageRecord {
+    /**
+     * Unique identifier for this usage record
+     */
     id: string;
 
+    /**
+     * Adaptation run ID this usage is associated with
+     */
     adaptation_run_id: string;
 
+    /**
+     * Cost of input tokens in USD
+     */
     input_cost: number;
 
+    /**
+     * Number of input tokens consumed
+     */
     input_tokens: number;
 
+    /**
+     * Model name (e.g., 'gpt-4', 'claude-3-opus-20240229')
+     */
     model: string;
 
+    /**
+     * Organization ID associated with the request
+     */
     organization_id: string;
 
+    /**
+     * Cost of output tokens in USD
+     */
     output_cost: number;
 
+    /**
+     * Number of output tokens generated
+     */
     output_tokens: number;
 
+    /**
+     * LLM provider (e.g., 'openai', 'anthropic', 'google')
+     */
     provider: string;
 
+    /**
+     * Type of task: 'pre-optimization evaluation', 'optimization', or
+     * 'post-optimization evaluation'
+     */
     task_type: string;
 
+    /**
+     * Unix timestamp when the request was made
+     */
     timestamp: number;
 
+    /**
+     * Total cost (input + output) in USD
+     */
     total_cost: number;
 
+    /**
+     * User ID who made the request
+     */
     user_id: string;
   }
 }
@@ -502,9 +700,9 @@ export interface PromptAdaptationAdaptParams {
 
   /**
    * List of models to adapt the prompt for. Maximum count depends on your
-   * subscription tier
+   * subscription tier (Free: 1, Starter: 3, Startup: 5, Enterprise: 10)
    */
-  target_models: Array<PromptAdaptationAdaptParams.TargetModel>;
+  target_models: Array<RequestProvider>;
 
   /**
    * User message template with placeholders for fields. Use curly braces for field
@@ -518,190 +716,50 @@ export interface PromptAdaptationAdaptParams {
 
   /**
    * Training examples (legacy parameter). Use train_goldens and test_goldens for
-   * better control
+   * better control. Minimum 25 examples (or 3 with prototype_mode=true)
    */
-  goldens?: Array<PromptAdaptationAdaptParams.Golden> | null;
+  goldens?: Array<GoldenRecord> | null;
 
   /**
    * Model for specifying an LLM provider in API requests.
    */
-  origin_model?: PromptAdaptationAdaptParams.OriginModel | null;
+  origin_model?: RequestProvider | null;
 
   /**
-   * Optional baseline score for the origin model
+   * Optional baseline score for the origin model. If provided, can skip origin model
+   * evaluation
    */
   origin_model_evaluation_score?: number | null;
 
   /**
-   * Test examples for evaluation. Required if train_goldens is provided
+   * Enable prototype mode to use as few as 3 training examples (instead of 25).
+   * Note: Performance may be degraded with fewer examples. Recommended for
+   * prototyping AI applications when you don't have enough data yet
    */
-  test_goldens?: Array<PromptAdaptationAdaptParams.TestGolden> | null;
+  prototype_mode?: boolean;
 
   /**
-   * Training examples for prompt optimization. Minimum 5 examples required
+   * Test examples for evaluation. Required if train_goldens is provided. Used to
+   * measure final performance on held-out data
    */
-  train_goldens?: Array<PromptAdaptationAdaptParams.TrainGolden> | null;
-}
+  test_goldens?: Array<GoldenRecord> | null;
 
-export namespace PromptAdaptationAdaptParams {
   /**
-   * Model for specifying an LLM provider in API requests.
+   * Training examples for prompt optimization. Minimum 25 examples required (or 3
+   * with prototype_mode=true). Cannot be used with 'goldens' parameter
    */
-  export interface TargetModel {
-    /**
-     * Model name (e.g., 'gpt-4o', 'claude-sonnet-4-5-20250929')
-     */
-    model: string;
-
-    /**
-     * Provider name (e.g., 'openai', 'anthropic', 'google')
-     */
-    provider: string;
-
-    /**
-     * Maximum context length for the model (required for custom models)
-     */
-    context_length?: number | null;
-
-    /**
-     * Input token price per million tokens in USD (required for custom models)
-     */
-    input_price?: number | null;
-
-    /**
-     * Whether this is a custom model not in Not Diamond's supported model list
-     */
-    is_custom?: boolean;
-
-    /**
-     * Average latency in seconds (required for custom models)
-     */
-    latency?: number | null;
-
-    /**
-     * Output token price per million tokens in USD (required for custom models)
-     */
-    output_price?: number | null;
-  }
-
-  /**
-   * A training or test example for prompt adaptation.
-   */
-  export interface Golden {
-    /**
-     * Dictionary mapping field names to their values. Keys must match the fields
-     * specified in the template
-     */
-    fields: { [key: string]: string };
-
-    /**
-     * Expected answer for supervised evaluation. Required for supervised metrics,
-     * optional for unsupervised
-     */
-    answer?: string | null;
-  }
-
-  /**
-   * Model for specifying an LLM provider in API requests.
-   */
-  export interface OriginModel {
-    /**
-     * Model name (e.g., 'gpt-4o', 'claude-sonnet-4-5-20250929')
-     */
-    model: string;
-
-    /**
-     * Provider name (e.g., 'openai', 'anthropic', 'google')
-     */
-    provider: string;
-
-    /**
-     * Maximum context length for the model (required for custom models)
-     */
-    context_length?: number | null;
-
-    /**
-     * Input token price per million tokens in USD (required for custom models)
-     */
-    input_price?: number | null;
-
-    /**
-     * Whether this is a custom model not in Not Diamond's supported model list
-     */
-    is_custom?: boolean;
-
-    /**
-     * Average latency in seconds (required for custom models)
-     */
-    latency?: number | null;
-
-    /**
-     * Output token price per million tokens in USD (required for custom models)
-     */
-    output_price?: number | null;
-  }
-
-  /**
-   * A training or test example for prompt adaptation.
-   */
-  export interface TestGolden {
-    /**
-     * Dictionary mapping field names to their values. Keys must match the fields
-     * specified in the template
-     */
-    fields: { [key: string]: string };
-
-    /**
-     * Expected answer for supervised evaluation. Required for supervised metrics,
-     * optional for unsupervised
-     */
-    answer?: string | null;
-  }
-
-  /**
-   * A training or test example for prompt adaptation.
-   */
-  export interface TrainGolden {
-    /**
-     * Dictionary mapping field names to their values. Keys must match the fields
-     * specified in the template
-     */
-    fields: { [key: string]: string };
-
-    /**
-     * Expected answer for supervised evaluation. Required for supervised metrics,
-     * optional for unsupervised
-     */
-    answer?: string | null;
-  }
-}
-
-export interface PromptAdaptationGetAdaptRunResultsParams {
-  /**
-   * Path param:
-   */
-  user_id: string;
-
-  /**
-   * Header param:
-   */
-  'x-token': string;
-}
-
-export interface PromptAdaptationGetAdaptRunsParams {
-  'x-token': string;
+  train_goldens?: Array<GoldenRecord> | null;
 }
 
 export declare namespace PromptAdaptation {
   export {
-    type AdaptationRunResults as AdaptationRunResults,
+    type GoldenRecord as GoldenRecord,
     type JobStatus as JobStatus,
+    type RequestProvider as RequestProvider,
     type PromptAdaptationAdaptResponse as PromptAdaptationAdaptResponse,
-    type PromptAdaptationGetAdaptRunsResponse as PromptAdaptationGetAdaptRunsResponse,
+    type PromptAdaptationGetAdaptResultsResponse as PromptAdaptationGetAdaptResultsResponse,
     type PromptAdaptationGetAdaptStatusResponse as PromptAdaptationGetAdaptStatusResponse,
-    type PromptAdaptationRetrieveCostsResponse as PromptAdaptationRetrieveCostsResponse,
+    type PromptAdaptationGetCostResponse as PromptAdaptationGetCostResponse,
     type PromptAdaptationAdaptParams as PromptAdaptationAdaptParams,
-    type PromptAdaptationGetAdaptRunResultsParams as PromptAdaptationGetAdaptRunResultsParams,
-    type PromptAdaptationGetAdaptRunsParams as PromptAdaptationGetAdaptRunsParams,
   };
 }