langwatch 0.33.2 → 0.34.0

package/dist/index.d.mts

@@ -1,7 +1,7 @@
 import { L as Logger, C as ConsoleLogger, N as NoOpLogger } from './index-D7rKIGrO.mjs';
-export { F as FilterableBatchSpanProcessor, L as LangWatchExporter, S as SpanProcessingExcludeRule, g as getLangWatchLogger, d as getLangWatchTracer } from './implementation-Oz1fNv8p.mjs';
+export { F as FilterableBatchSpanProcessor, L as LangWatchExporter, S as SpanProcessingExcludeRule, g as getLangWatchLogger, d as getLangWatchTracer } from './implementation-BBG46scL.mjs';
 import { z } from 'zod';
-import { p as paths, P as PromptResponse, g as CreatePromptBody, U as UpdatePromptBody, T as TagDefinition, h as CreatedTag, o as operations, i as PromptData, F as FetchPolicy, j as Prompt, a as LangWatchSpan, k as components } from './types-BNbQoBnh.mjs';
+import { p as paths, P as PromptResponse, i as CreatePromptBody, U as UpdatePromptBody, T as TagDefinition, j as CreatedTag, o as operations, k as PromptData, F as FetchPolicy, l as Prompt, a as LangWatchSpan, m as components } from './types-fzTtg5sg.mjs';
 import openApiCreateClient from 'openapi-fetch';
 export { l as attributes } from './types-BvCQH8F1.mjs';
 import '@opentelemetry/sdk-trace-base';
@@ -1268,6 +1268,87 @@ type RunExperimentOptions = {
      */
     onProgress?: (progress: number, total: number) => void;
 };
+/**
+ * Inputs that override a platform experiment or workflow run on the server.
+ *
+ * `data` and `datasetId` are mutually exclusive (the backend rejects passing
+ * both). `parameters` overrides target parameters (e.g. prompt variables) and
+ * `rowIndices` restricts execution to a subset of the configured dataset.
+ */
+type RunWithResultsOptions = {
+    /**
+     * Inline rows to evaluate. Mutually exclusive with `datasetId`.
+     */
+    data?: Array<Record<string, unknown>>;
+    /**
+     * Id of a saved dataset to evaluate. Mutually exclusive with `data`.
+     */
+    datasetId?: string;
+    /**
+     * Target parameter overrides (e.g. prompt variables, model settings).
+     */
+    parameters?: Record<string, string | number | boolean>;
+    /**
+     * Subset of dataset row indices to run.
+     */
+    rowIndices?: number[];
+} & RunExperimentOptions;
+/**
+ * A single per-row result, mirroring one row of the python SDK's results
+ * DataFrame (`_build_df_from_platform`).
+ */
+type ExperimentRowResult = {
+    /**
+     * Index of the dataset row this result belongs to.
+     */
+    index: number;
+    /**
+     * Target id, present only for multi-target runs.
+     */
+    target?: string;
+    /**
+     * Flattened dataset entry fields for this row.
+     */
+    input: Record<string, unknown>;
+    /**
+     * Output produced by the target (`predicted.output`).
+     */
+    output: unknown;
+    /**
+     * Trace id of the target execution for this row.
+     */
+    traceId: string;
+    /**
+     * Cost of the target execution, when measured.
+     */
+    cost?: number;
+    /**
+     * Duration of the target execution in milliseconds, when measured.
+     */
+    duration?: number;
+    /**
+     * Error message for this row, when the target execution failed.
+     */
+    error?: string;
+    /**
+     * Evaluation results for this row keyed by evaluator name.
+     */
+    evaluations: Record<string, {
+        score?: number;
+        passed?: boolean;
+    }>;
+};
+/**
+ * Result of a platform experiment or workflow run, including per-row results
+ * and a link to the run in LangWatch.
+ */
+type ExperimentRunWithResults = {
+    runId: string;
+    runUrl: string;
+    status: string;
+    summary: ExperimentRunSummary;
+    rows: ExperimentRowResult[];
+};
 /**
  * Final result of a platform experiment run
  */
@@ -1310,6 +1391,7 @@ type ExperimentsFacadeConfig = {
  */
 declare class ExperimentsFacade {
     private readonly config;
+    private readonly apiService;
     constructor(config: ExperimentsFacadeConfig);
     /**
      * Initialize a new experiment session (SDK-defined)
@@ -1351,6 +1433,31 @@ declare class ExperimentsFacade {
      * ```
      */
     run(slug: string, options?: RunExperimentOptions): Promise<ExperimentRunResult>;
+    /**
+     * Run a platform experiment and return per-row structured results.
+     *
+     * Starts the run through the unified evaluations-v3 backend (optionally
+     * overriding the configured inputs via `data` / `datasetId` / `parameters` /
+     * `rowIndices`), polls to completion, fetches the per-row results, and maps
+     * them to the same row structure as the python SDK's results DataFrame.
+     *
+     * @param slug - The slug of the experiment (found in the experiment URL)
+     * @param options - Optional inputs and polling configuration
+     * @returns The run id, results URL, status, summary, and per-row results
+     *
+     * @example
+     * ```typescript
+     * const langwatch = new LangWatch();
+     * const { rows, runUrl } = await langwatch.experiments.runWithResults(
+     *   "my-experiment-slug",
+     *   { data: [{ question: "What is 2 + 2?" }] },
+     * );
+     * for (const row of rows) {
+     *   console.log(row.output, row.evaluations);
+     * }
+     * ```
+     */
+    runWithResults(slug: string, options?: RunWithResultsOptions): Promise<ExperimentRunWithResults>;
     /**
      * Run an experiment and wait for completion using polling
      */
@@ -1368,10 +1475,6 @@ declare class ExperimentsFacade {
      */
     private buildResult;
     private sleep;
-    /**
-     * Replace the domain of a URL with a new base URL, preserving the path
-     */
-    private replaceUrlDomain;
 }
 
 /**
@@ -1660,11 +1763,41 @@ type WorkflowResponse = NonNullable<paths["/api/workflows"]["get"]["responses"][
 type WorkflowDeleteResponse = paths["/api/workflows/{id}"]["delete"]["responses"]["200"]["content"]["application/json"];
 declare class WorkflowsApiService {
     private readonly apiClient;
-    constructor(config?: Pick<InternalConfig, "langwatchApiClient">);
+    private readonly experimentsApiService;
+    private readonly endpoint;
+    constructor(config?: Pick<InternalConfig, "langwatchApiClient"> & {
+        endpoint?: string;
+    });
     private handleApiError;
+    private postUndeclaredEndpoint;
     getAll(): Promise<WorkflowResponse[]>;
     get(id: string): Promise<WorkflowResponse>;
     delete(id: string): Promise<WorkflowDeleteResponse>;
+    /**
+     * Run a studio workflow evaluation and return per-row structured results.
+     *
+     * Starts the evaluation through the unified evaluations-v3 backend
+     * (`POST /api/workflows/{workflowId}/evaluate`), polls to completion, fetches
+     * the per-row results, and maps them to the same row structure as the python
+     * SDK's results DataFrame.
+     *
+     * @param workflowId - The studio workflow id
+     * @param options - Optional inputs, committed version override, and polling
+     *                   configuration. `data` and `datasetId` are mutually
+     *                   exclusive.
+     * @returns The run id, results URL, status, summary, and per-row results
+     *
+     * @example
+     * ```typescript
+     * const langwatch = new LangWatch();
+     * const { rows, runUrl } = await langwatch.workflows.run("workflow_123", {
+     *   data: [{ question: "What is 2 + 2?" }],
+     * });
+     * ```
+     */
+    run(workflowId: string, options?: RunWithResultsOptions & {
+        versionId?: string;
+    }): Promise<ExperimentRunWithResults>;
 }
 
 interface AgentResponse {
@@ -1995,4 +2128,4 @@ declare const logger: {
     NoOpLogger: typeof NoOpLogger;
 };
 
-export { type BatchCreateRecordsResponse, type CreateDatasetOptions, type CreateEvaluatorBody, type CreateFromUploadResponse, type Dataset, DatasetApiError, type DatasetColumnType, type DatasetEntry, DatasetError, type DatasetListItem, type DatasetMetadata, DatasetNotFoundError, DatasetPlanLimitError, type DatasetRecordResponse, DatasetValidationError, DatasetsFacade, type DeleteRecordsResponse, type EvaluateOptions, type EvaluationCost, EvaluationError, type EvaluationResult, type EvaluationStatus, EvaluationsApiError, EvaluationsFacade, EvaluatorCallError, EvaluatorError, type EvaluatorField, EvaluatorNotFoundError, type EvaluatorResponse, EvaluatorsApiError, EvaluatorsApiService, Experiment, ExperimentApiError, ExperimentError, type EvaluateOptions$1 as ExperimentEvaluateOptions, type EvaluationResult$1 as ExperimentEvaluationResult, type EvaluationStatus$1 as ExperimentEvaluationStatus, ExperimentInitError, type ExperimentInitOptions, ExperimentsFacade, FetchPolicy, type GetDatasetOptions, type GetPromptOptions, LangWatch, type ListDatasetsApiResponse, type ListDatasetsOptions, type ListRecordsApiResponse, type ListRecordsOptions, type LogOptions, type PaginatedResponse, type Pagination, type RunCallback, type RunContext, type RunOptions, type TargetInfo, type TargetMetadata, TargetMetadataConflictError, type TargetType, type UpdateDatasetOptions, type UploadResponse, logger };
+export { type BatchCreateRecordsResponse, type CreateDatasetOptions, type CreateEvaluatorBody, type CreateFromUploadResponse, type Dataset, DatasetApiError, type DatasetColumnType, type DatasetEntry, DatasetError, type DatasetListItem, type DatasetMetadata, DatasetNotFoundError, DatasetPlanLimitError, type DatasetRecordResponse, DatasetValidationError, DatasetsFacade, type DeleteRecordsResponse, type EvaluateOptions, type EvaluationCost, EvaluationError, type EvaluationResult, type EvaluationStatus, EvaluationsApiError, EvaluationsFacade, EvaluatorCallError, EvaluatorError, type EvaluatorField, EvaluatorNotFoundError, type EvaluatorResponse, EvaluatorsApiError, EvaluatorsApiService, Experiment, ExperimentApiError, ExperimentError, type EvaluateOptions$1 as ExperimentEvaluateOptions, type EvaluationResult$1 as ExperimentEvaluationResult, type EvaluationStatus$1 as ExperimentEvaluationStatus, ExperimentInitError, type ExperimentInitOptions, type ExperimentRowResult, type ExperimentRunWithResults, ExperimentsFacade, FetchPolicy, type GetDatasetOptions, type GetPromptOptions, LangWatch, type ListDatasetsApiResponse, type ListDatasetsOptions, type ListRecordsApiResponse, type ListRecordsOptions, type LogOptions, type PaginatedResponse, type Pagination, type RunCallback, type RunContext, type RunOptions, type RunWithResultsOptions, type TargetInfo, type TargetMetadata, TargetMetadataConflictError, type TargetType, type UpdateDatasetOptions, type UploadResponse, logger };

package/dist/index.d.ts

@@ -1,7 +1,7 @@
 import { L as Logger, C as ConsoleLogger, N as NoOpLogger } from './index-D7rKIGrO.js';
-export { F as FilterableBatchSpanProcessor, L as LangWatchExporter, S as SpanProcessingExcludeRule, g as getLangWatchLogger, d as getLangWatchTracer } from './implementation-Cbelb5TK.js';
+export { F as FilterableBatchSpanProcessor, L as LangWatchExporter, S as SpanProcessingExcludeRule, g as getLangWatchLogger, d as getLangWatchTracer } from './implementation-BIfcyP32.js';
 import { z } from 'zod';
-import { p as paths, P as PromptResponse, g as CreatePromptBody, U as UpdatePromptBody, T as TagDefinition, h as CreatedTag, o as operations, i as PromptData, F as FetchPolicy, j as Prompt, a as LangWatchSpan, k as components } from './types-CmMOt86P.js';
+import { p as paths, P as PromptResponse, i as CreatePromptBody, U as UpdatePromptBody, T as TagDefinition, j as CreatedTag, o as operations, k as PromptData, F as FetchPolicy, l as Prompt, a as LangWatchSpan, m as components } from './types-CYfKDbKj.js';
 import openApiCreateClient from 'openapi-fetch';
 export { l as attributes } from './types-BvCQH8F1.js';
 import '@opentelemetry/sdk-trace-base';
@@ -1268,6 +1268,87 @@ type RunExperimentOptions = {
      */
     onProgress?: (progress: number, total: number) => void;
 };
+/**
+ * Inputs that override a platform experiment or workflow run on the server.
+ *
+ * `data` and `datasetId` are mutually exclusive (the backend rejects passing
+ * both). `parameters` overrides target parameters (e.g. prompt variables) and
+ * `rowIndices` restricts execution to a subset of the configured dataset.
+ */
+type RunWithResultsOptions = {
+    /**
+     * Inline rows to evaluate. Mutually exclusive with `datasetId`.
+     */
+    data?: Array<Record<string, unknown>>;
+    /**
+     * Id of a saved dataset to evaluate. Mutually exclusive with `data`.
+     */
+    datasetId?: string;
+    /**
+     * Target parameter overrides (e.g. prompt variables, model settings).
+     */
+    parameters?: Record<string, string | number | boolean>;
+    /**
+     * Subset of dataset row indices to run.
+     */
+    rowIndices?: number[];
+} & RunExperimentOptions;
+/**
+ * A single per-row result, mirroring one row of the python SDK's results
+ * DataFrame (`_build_df_from_platform`).
+ */
+type ExperimentRowResult = {
+    /**
+     * Index of the dataset row this result belongs to.
+     */
+    index: number;
+    /**
+     * Target id, present only for multi-target runs.
+     */
+    target?: string;
+    /**
+     * Flattened dataset entry fields for this row.
+     */
+    input: Record<string, unknown>;
+    /**
+     * Output produced by the target (`predicted.output`).
+     */
+    output: unknown;
+    /**
+     * Trace id of the target execution for this row.
+     */
+    traceId: string;
+    /**
+     * Cost of the target execution, when measured.
+     */
+    cost?: number;
+    /**
+     * Duration of the target execution in milliseconds, when measured.
+     */
+    duration?: number;
+    /**
+     * Error message for this row, when the target execution failed.
+     */
+    error?: string;
+    /**
+     * Evaluation results for this row keyed by evaluator name.
+     */
+    evaluations: Record<string, {
+        score?: number;
+        passed?: boolean;
+    }>;
+};
+/**
+ * Result of a platform experiment or workflow run, including per-row results
+ * and a link to the run in LangWatch.
+ */
+type ExperimentRunWithResults = {
+    runId: string;
+    runUrl: string;
+    status: string;
+    summary: ExperimentRunSummary;
+    rows: ExperimentRowResult[];
+};
 /**
  * Final result of a platform experiment run
  */
@@ -1310,6 +1391,7 @@ type ExperimentsFacadeConfig = {
  */
 declare class ExperimentsFacade {
     private readonly config;
+    private readonly apiService;
     constructor(config: ExperimentsFacadeConfig);
     /**
      * Initialize a new experiment session (SDK-defined)
@@ -1351,6 +1433,31 @@ declare class ExperimentsFacade {
      * ```
      */
     run(slug: string, options?: RunExperimentOptions): Promise<ExperimentRunResult>;
+    /**
+     * Run a platform experiment and return per-row structured results.
+     *
+     * Starts the run through the unified evaluations-v3 backend (optionally
+     * overriding the configured inputs via `data` / `datasetId` / `parameters` /
+     * `rowIndices`), polls to completion, fetches the per-row results, and maps
+     * them to the same row structure as the python SDK's results DataFrame.
+     *
+     * @param slug - The slug of the experiment (found in the experiment URL)
+     * @param options - Optional inputs and polling configuration
+     * @returns The run id, results URL, status, summary, and per-row results
+     *
+     * @example
+     * ```typescript
+     * const langwatch = new LangWatch();
+     * const { rows, runUrl } = await langwatch.experiments.runWithResults(
+     *   "my-experiment-slug",
+     *   { data: [{ question: "What is 2 + 2?" }] },
+     * );
+     * for (const row of rows) {
+     *   console.log(row.output, row.evaluations);
+     * }
+     * ```
+     */
+    runWithResults(slug: string, options?: RunWithResultsOptions): Promise<ExperimentRunWithResults>;
     /**
      * Run an experiment and wait for completion using polling
      */
@@ -1368,10 +1475,6 @@ declare class ExperimentsFacade {
      */
     private buildResult;
     private sleep;
-    /**
-     * Replace the domain of a URL with a new base URL, preserving the path
-     */
-    private replaceUrlDomain;
 }
 
 /**
@@ -1660,11 +1763,41 @@ type WorkflowResponse = NonNullable<paths["/api/workflows"]["get"]["responses"][
 type WorkflowDeleteResponse = paths["/api/workflows/{id}"]["delete"]["responses"]["200"]["content"]["application/json"];
 declare class WorkflowsApiService {
     private readonly apiClient;
-    constructor(config?: Pick<InternalConfig, "langwatchApiClient">);
+    private readonly experimentsApiService;
+    private readonly endpoint;
+    constructor(config?: Pick<InternalConfig, "langwatchApiClient"> & {
+        endpoint?: string;
+    });
     private handleApiError;
+    private postUndeclaredEndpoint;
     getAll(): Promise<WorkflowResponse[]>;
     get(id: string): Promise<WorkflowResponse>;
     delete(id: string): Promise<WorkflowDeleteResponse>;
+    /**
+     * Run a studio workflow evaluation and return per-row structured results.
+     *
+     * Starts the evaluation through the unified evaluations-v3 backend
+     * (`POST /api/workflows/{workflowId}/evaluate`), polls to completion, fetches
+     * the per-row results, and maps them to the same row structure as the python
+     * SDK's results DataFrame.
+     *
+     * @param workflowId - The studio workflow id
+     * @param options - Optional inputs, committed version override, and polling
+     *                   configuration. `data` and `datasetId` are mutually
+     *                   exclusive.
+     * @returns The run id, results URL, status, summary, and per-row results
+     *
+     * @example
+     * ```typescript
+     * const langwatch = new LangWatch();
+     * const { rows, runUrl } = await langwatch.workflows.run("workflow_123", {
+     *   data: [{ question: "What is 2 + 2?" }],
+     * });
+     * ```
+     */
+    run(workflowId: string, options?: RunWithResultsOptions & {
+        versionId?: string;
+    }): Promise<ExperimentRunWithResults>;
 }
 
 interface AgentResponse {
@@ -1995,4 +2128,4 @@ declare const logger: {
     NoOpLogger: typeof NoOpLogger;
 };
 
-export { type BatchCreateRecordsResponse, type CreateDatasetOptions, type CreateEvaluatorBody, type CreateFromUploadResponse, type Dataset, DatasetApiError, type DatasetColumnType, type DatasetEntry, DatasetError, type DatasetListItem, type DatasetMetadata, DatasetNotFoundError, DatasetPlanLimitError, type DatasetRecordResponse, DatasetValidationError, DatasetsFacade, type DeleteRecordsResponse, type EvaluateOptions, type EvaluationCost, EvaluationError, type EvaluationResult, type EvaluationStatus, EvaluationsApiError, EvaluationsFacade, EvaluatorCallError, EvaluatorError, type EvaluatorField, EvaluatorNotFoundError, type EvaluatorResponse, EvaluatorsApiError, EvaluatorsApiService, Experiment, ExperimentApiError, ExperimentError, type EvaluateOptions$1 as ExperimentEvaluateOptions, type EvaluationResult$1 as ExperimentEvaluationResult, type EvaluationStatus$1 as ExperimentEvaluationStatus, ExperimentInitError, type ExperimentInitOptions, ExperimentsFacade, FetchPolicy, type GetDatasetOptions, type GetPromptOptions, LangWatch, type ListDatasetsApiResponse, type ListDatasetsOptions, type ListRecordsApiResponse, type ListRecordsOptions, type LogOptions, type PaginatedResponse, type Pagination, type RunCallback, type RunContext, type RunOptions, type TargetInfo, type TargetMetadata, TargetMetadataConflictError, type TargetType, type UpdateDatasetOptions, type UploadResponse, logger };
+export { type BatchCreateRecordsResponse, type CreateDatasetOptions, type CreateEvaluatorBody, type CreateFromUploadResponse, type Dataset, DatasetApiError, type DatasetColumnType, type DatasetEntry, DatasetError, type DatasetListItem, type DatasetMetadata, DatasetNotFoundError, DatasetPlanLimitError, type DatasetRecordResponse, DatasetValidationError, DatasetsFacade, type DeleteRecordsResponse, type EvaluateOptions, type EvaluationCost, EvaluationError, type EvaluationResult, type EvaluationStatus, EvaluationsApiError, EvaluationsFacade, EvaluatorCallError, EvaluatorError, type EvaluatorField, EvaluatorNotFoundError, type EvaluatorResponse, EvaluatorsApiError, EvaluatorsApiService, Experiment, ExperimentApiError, ExperimentError, type EvaluateOptions$1 as ExperimentEvaluateOptions, type EvaluationResult$1 as ExperimentEvaluationResult, type EvaluationStatus$1 as ExperimentEvaluationStatus, ExperimentInitError, type ExperimentInitOptions, type ExperimentRowResult, type ExperimentRunWithResults, ExperimentsFacade, FetchPolicy, type GetDatasetOptions, type GetPromptOptions, LangWatch, type ListDatasetsApiResponse, type ListDatasetsOptions, type ListRecordsApiResponse, type ListRecordsOptions, type LogOptions, type PaginatedResponse, type Pagination, type RunCallback, type RunContext, type RunOptions, type RunWithResultsOptions, type TargetInfo, type TargetMetadata, TargetMetadataConflictError, type TargetType, type UpdateDatasetOptions, type UploadResponse, logger };