@langfuse/client 4.0.0 → 4.1.0-alpha.0

package/dist/index.cjs

@@ -33,16 +33,18 @@ __export(index_exports, {
   ChatMessageType: () => ChatMessageType,
   ChatPromptClient: () => ChatPromptClient,
   DatasetManager: () => DatasetManager,
+  ExperimentManager: () => ExperimentManager,
   LangfuseClient: () => LangfuseClient,
   MediaManager: () => MediaManager,
   PromptManager: () => PromptManager,
   ScoreManager: () => ScoreManager,
-  TextPromptClient: () => TextPromptClient
+  TextPromptClient: () => TextPromptClient,
+  autoevalsToLangfuseEvaluator: () => autoevalsToLangfuseEvaluator
 });
 module.exports = __toCommonJS(index_exports);
 
 // src/LangfuseClient.ts
-var import_core5 = require("@langfuse/core");
+var import_core6 = require("@langfuse/core");
 
 // src/dataset/index.ts
 var DatasetManager = class {
@@ -53,44 +55,91 @@ var DatasetManager = class {
    * @internal
    */
   constructor(params) {
-    this.apiClient = params.apiClient;
+    this.langfuseClient = params.langfuseClient;
   }
   /**
-   * Retrieves a dataset by name along with all its items.
+   * Retrieves a dataset by name with all its items and experiment functionality.
    *
-   * This method automatically handles pagination to fetch all dataset items
-   * and enhances each item with a `link` function for easy experiment tracking.
+   * This method fetches a dataset and all its associated items, with support
+   * for automatic pagination to handle large datasets efficiently. The returned
+   * dataset object includes enhanced functionality for linking items to traces
+   * and running experiments directly on the dataset.
    *
    * @param name - The name of the dataset to retrieve
-   * @param options - Optional configuration for fetching
+   * @param options - Optional configuration for data fetching
    * @param options.fetchItemsPageSize - Number of items to fetch per page (default: 50)
+   * @returns Promise resolving to enhanced dataset with items, linking, and experiment capabilities
    *
-   * @returns Promise that resolves to the dataset with enhanced items
+   * @example Basic dataset retrieval
+   * ```typescript
+   * const dataset = await langfuse.dataset.get("my-evaluation-dataset");
+   * console.log(`Dataset ${dataset.name} has ${dataset.items.length} items`);
    *
-   * @example
+   * // Access dataset properties
+   * console.log(dataset.description);
+   * console.log(dataset.metadata);
+   * ```
+   *
+   * @example Working with dataset items
    * ```typescript
-   * const dataset = await langfuse.dataset.get("my-dataset");
+   * const dataset = await langfuse.dataset.get("qa-dataset");
    *
    * for (const item of dataset.items) {
-   *   // Use the item data for your experiment
-   *   const result = await processItem(item.input);
-   *
-   *   // Link the result to the dataset item
-   *   await item.link(
-   *     { otelSpan: currentSpan },
-   *     "experiment-run-1",
-   *     { description: "Testing new model" }
-   *   );
+   *   console.log("Question:", item.input);
+   *   console.log("Expected Answer:", item.expectedOutput);
+   *
+   *   // Each item has a link function for connecting to traces
+   *   // await item.link(span, "experiment-name");
    * }
    * ```
+   *
+   * @example Running experiments on datasets
+   * ```typescript
+   * const dataset = await langfuse.dataset.get("benchmark-dataset");
+   *
+   * const result = await dataset.runExperiment({
+   *   name: "GPT-4 Benchmark",
+   *   description: "Evaluating GPT-4 on our benchmark tasks",
+   *   task: async ({ input }) => {
+   *     const response = await openai.chat.completions.create({
+   *       model: "gpt-4",
+   *       messages: [{ role: "user", content: input }]
+   *     });
+   *     return response.choices[0].message.content;
+   *   },
+   *   evaluators: [
+   *     async ({ output, expectedOutput }) => ({
+   *       name: "exact_match",
+   *       value: output === expectedOutput ? 1 : 0
+   *     })
+   *   ]
+   * });
+   *
+   * console.log(await result.prettyPrint());
+   * ```
+   *
+   * @example Handling large datasets
+   * ```typescript
+   * // For very large datasets, use smaller page sizes
+   * const largeDataset = await langfuse.dataset.get(
+   *   "large-dataset",
+   *   { fetchItemsPageSize: 100 }
+   * );
+   * ```
+   *
+   * @throws {Error} If the dataset does not exist or cannot be accessed
+   * @see {@link FetchedDataset} for the complete return type specification
+   * @see {@link RunExperimentOnDataset} for experiment execution details
+   * @public
+   * @since 4.0.0
    */
   async get(name, options) {
     var _a;
-    const dataset = await this.apiClient.datasets.get(name);
+    const dataset = await this.langfuseClient.api.datasets.get(name);
     const items = [];
     let page = 1;
     while (true) {
-      const itemsResponse = await this.apiClient.datasetItems.list({
+      const itemsResponse = await this.langfuseClient.api.datasetItems.list({
         datasetName: name,
         limit: (_a = options == null ? void 0 : options.fetchItemsPageSize) != null ? _a : 50,
         page
@@ -101,12 +150,20 @@ var DatasetManager = class {
       }
       page++;
     }
+    const itemsWithLinkMethod = items.map((item) => ({
+      ...item,
+      link: this.createDatasetItemLinkFunction(item)
+    }));
+    const runExperiment = (params) => {
+      return this.langfuseClient.experiment.run({
+        data: items,
+        ...params
+      });
+    };
     const returnDataset = {
       ...dataset,
-      items: items.map((item) => ({
-        ...item,
-        link: this.createDatasetItemLinkFunction(item)
-      }))
+      items: itemsWithLinkMethod,
+      runExperiment
     };
     return returnDataset;
   }
@@ -119,7 +176,7 @@ var DatasetManager = class {
    */
   createDatasetItemLinkFunction(item) {
     const linkFunction = async (obj, runName, runArgs) => {
-      return await this.apiClient.datasetRunItems.create({
+      return await this.langfuseClient.api.datasetRunItems.create({
         runName,
         datasetItemId: item.id,
         traceId: obj.otelSpan.spanContext().traceId,
@@ -131,8 +188,498 @@ var DatasetManager = class {
   }
 };
 
-// src/media/index.ts
+// src/experiment/ExperimentManager.ts
 var import_core = require("@langfuse/core");
+var import_tracing = require("@langfuse/tracing");
+var import_api = require("@opentelemetry/api");
+var ExperimentManager = class {
+  /**
+   * Creates a new ExperimentManager instance.
+   *
+   * @param params - Configuration object
+   * @param params.langfuseClient - The Langfuse client instance for API communication
+   * @internal
+   */
+  constructor(params) {
+    this.langfuseClient = params.langfuseClient;
+  }
+  /**
+   * Gets the global logger instance for experiment-related logging.
+   *
+   * @returns The global logger instance
+   * @internal
+   */
+  get logger() {
+    return (0, import_core.getGlobalLogger)();
+  }
+  /**
+   * Executes an experiment by running a task on each data item and evaluating the results.
+   *
+   * This method orchestrates the complete experiment lifecycle:
+   * 1. Executes the task function on each data item with proper tracing
+   * 2. Runs item-level evaluators on each task output
+   * 3. Executes run-level evaluators on the complete result set
+   * 4. Links results to dataset runs (for Langfuse datasets)
+   * 5. Stores all scores and traces in Langfuse
+   *
+   * @param config - The experiment configuration
+   * @param config.name - Human-readable name for the experiment
+   * @param config.description - Optional description of the experiment's purpose
+   * @param config.metadata - Optional metadata to attach to the experiment run
+   * @param config.data - Array of data items to process (ExperimentItem[] or DatasetItem[])
+   * @param config.task - Function that processes each data item and returns output
+   * @param config.evaluators - Optional array of functions to evaluate each item's output
+   * @param config.runEvaluators - Optional array of functions to evaluate the entire run
+   * @param config.maxConcurrency - Maximum number of concurrent task executions (default: Infinity)
+   *
+   * @returns Promise that resolves to experiment results including:
+   *   - itemResults: Results for each processed data item
+   *   - runEvaluations: Results from run-level evaluators
+   *   - datasetRunId: ID of the dataset run (if using Langfuse datasets)
+   *   - prettyPrint: Function to format and display results
+   *
+   * @throws {Error} When task execution fails and cannot be handled gracefully
+   * @throws {Error} When required evaluators fail critically
+   *
+   * @example Simple experiment
+   * ```typescript
+   * const result = await langfuse.experiment.run({
+   *   name: "Translation Quality Test",
+   *   data: [
+   *     { input: "Hello world", expectedOutput: "Hola mundo" },
+   *     { input: "Good morning", expectedOutput: "Buenos días" }
+   *   ],
+   *   task: async ({ input }) => translateText(input, 'es'),
+   *   evaluators: [
+   *     async ({ output, expectedOutput }) => ({
+   *       name: "bleu_score",
+   *       value: calculateBleuScore(output, expectedOutput)
+   *     })
+   *   ]
+   * });
+   * ```
+   *
+   * @example Experiment with concurrency control
+   * ```typescript
+   * const result = await langfuse.experiment.run({
+   *   name: "Large Scale Evaluation",
+   *   data: largeBatchOfItems,
+   *   task: expensiveModelCall,
+   *   maxConcurrency: 5, // Process max 5 items simultaneously
+   *   evaluators: [myEvaluator],
+   *   runEvaluators: [
+   *     async ({ itemResults }) => ({
+   *       name: "average_score",
+   *       value: itemResults.reduce((acc, r) => acc + r.evaluations[0].value, 0) / itemResults.length
+   *     })
+   *   ]
+   * });
+   * ```
+   *
+   * @see {@link ExperimentParams} for detailed parameter documentation
+   * @see {@link ExperimentResult} for detailed return value documentation
+   * @see {@link Evaluator} for evaluator function specifications
+   * @see {@link RunEvaluator} for run evaluator function specifications
+   *
+   * @public
+   */
+  async run(config) {
+    const {
+      data,
+      evaluators,
+      task,
+      name,
+      description,
+      metadata,
+      maxConcurrency: batchSize = Infinity,
+      runEvaluators
+    } = config;
+    if (!this.isOtelRegistered()) {
+      this.logger.warn(
+        "OpenTelemetry has not been set up. Traces will not be sent to Langfuse.See our docs on how to set up OpenTelemetry: https://langfuse.com/docs/observability/sdk/typescript/setup#tracing-setup"
+      );
+    }
+    const itemResults = [];
+    for (let i = 0; i < data.length; i += batchSize) {
+      const batch = data.slice(i, i + batchSize);
+      const promises = batch.map(
+        async (item) => {
+          return this.runItem({
+            item,
+            evaluators,
+            task,
+            experimentName: name,
+            experimentDescription: description,
+            experimentMetadata: metadata
+          });
+        }
+      );
+      const results = await Promise.all(promises);
+      itemResults.push(...results);
+    }
+    const datasetRunId = itemResults.length > 0 ? itemResults[0].datasetRunId : void 0;
+    let datasetRunUrl = void 0;
+    if (datasetRunId && data.length > 0 && "datasetId" in data[0]) {
+      const datasetId = data[0].datasetId;
+      const projectUrl = (await this.langfuseClient.getTraceUrl("mock")).split(
+        "/traces"
+      )[0];
+      datasetRunUrl = `${projectUrl}/datasets/${datasetId}/runs/${datasetRunId}`;
+    }
+    let runEvaluations = [];
+    if (runEvaluators && (runEvaluators == null ? void 0 : runEvaluators.length) > 0) {
+      const promises = runEvaluators.map(async (runEvaluator) => {
+        return runEvaluator({ itemResults }).then((result) => {
+          return Array.isArray(result) ? result : [result];
+        }).catch((err) => {
+          this.logger.error("Run evaluator failed with error ", err);
+          throw err;
+        });
+      });
+      runEvaluations = (await Promise.allSettled(promises)).reduce(
+        (acc, settledPromise) => {
+          if (settledPromise.status === "fulfilled") {
+            acc.push(...settledPromise.value);
+          }
+          return acc;
+        },
+        []
+      );
+      if (datasetRunId) {
+        runEvaluations.forEach(
+          (runEval) => this.langfuseClient.score.create({ datasetRunId, ...runEval })
+        );
+      }
+    }
+    await this.langfuseClient.score.flush();
+    return {
+      itemResults,
+      datasetRunId,
+      runEvaluations,
+      prettyPrint: async (options) => {
+        var _a;
+        return await this.prettyPrintResults({
+          datasetRunUrl,
+          itemResults,
+          originalData: data,
+          runEvaluations,
+          name: config.name,
+          description: config.description,
+          includeItemResults: (_a = options == null ? void 0 : options.includeItemResults) != null ? _a : true
+        });
+      }
+    };
+  }
+  /**
+   * Executes the task and evaluators for a single data item.
+   *
+   * This method handles the complete processing pipeline for one data item:
+   * 1. Executes the task within a traced observation span
+   * 2. Links the result to a dataset run (if applicable)
+   * 3. Runs all item-level evaluators on the output
+   * 4. Stores evaluation scores in Langfuse
+   * 5. Handles errors gracefully by continuing with remaining evaluators
+   *
+   * @param params - Parameters for item execution
+   * @param params.experimentName - Name of the parent experiment
+   * @param params.experimentDescription - Description of the parent experiment
+   * @param params.experimentMetadata - Metadata for the parent experiment
+   * @param params.item - The data item to process
+   * @param params.task - The task function to execute
+   * @param params.evaluators - Optional evaluators to run on the output
+   *
+   * @returns Promise resolving to the item result with output, evaluations, and trace info
+   *
+   * @throws {Error} When task execution fails (propagated from task function)
+   *
+   * @internal
+   */
+  async runItem(params) {
+    const { item, evaluators = [], task } = params;
+    const { output, traceId } = await (0, import_tracing.startActiveObservation)(
+      "experiment-item-run",
+      async (span) => {
+        const output2 = await task(item);
+        span.update({
+          input: item.input,
+          output: output2
+        });
+        return { output: output2, traceId: span.traceId };
+      }
+    );
+    let datasetRunId = void 0;
+    if ("id" in item) {
+      await this.langfuseClient.api.datasetRunItems.create({
+        runName: params.experimentName,
+        runDescription: params.experimentDescription,
+        metadata: params.experimentMetadata,
+        datasetItemId: item.id,
+        traceId
+      }).then((result) => {
+        datasetRunId = result.datasetRunId;
+      }).catch(
+        (err) => this.logger.error("Linking dataset run item failed", err)
+      );
+    }
+    const evalPromises = evaluators.map(
+      async (evaluator) => {
+        const params2 = {
+          input: item.input,
+          expectedOutput: item.expectedOutput,
+          output
+        };
+        return evaluator(params2).then((result) => {
+          return Array.isArray(result) ? result : [result];
+        }).catch((err) => {
+          this.logger.error(
+            `Evaluator '${evaluator.name}' failed for params 
+
+${JSON.stringify(params2)}
+
+ with error: ${err}`
+          );
+          throw err;
+        });
+      }
+    );
+    const evals = (await Promise.allSettled(evalPromises)).reduce(
+      (acc, promiseResult) => {
+        if (promiseResult.status === "fulfilled") {
+          acc.push(...promiseResult.value.flat());
+        }
+        return acc;
+      },
+      []
+    );
+    for (const ev of evals) {
+      this.langfuseClient.score.create({
+        traceId,
+        name: ev.name,
+        comment: ev.comment,
+        value: ev.value,
+        metadata: ev.metadata,
+        dataType: ev.dataType
+      });
+    }
+    return {
+      output,
+      evaluations: evals,
+      traceId,
+      datasetRunId
+    };
+  }
+  /**
+   * Formats experiment results into a human-readable string representation.
+   *
+   * Creates a comprehensive, nicely formatted summary of the experiment including:
+   * - Individual item results with inputs, outputs, expected values, and scores
+   * - Dataset item and trace links (when available)
+   * - Experiment overview with aggregate statistics
+   * - Average scores across all evaluations
+   * - Run-level evaluation results
+   * - Links to dataset runs in the Langfuse UI
+   *
+   * @param params - Formatting parameters
+   * @param params.datasetRunUrl - Optional URL to the dataset run in Langfuse UI
+   * @param params.itemResults - Results from processing each data item
+   * @param params.originalData - The original input data items
+   * @param params.runEvaluations - Results from run-level evaluators
+   * @param params.name - Name of the experiment
+   * @param params.description - Optional description of the experiment
+   * @param params.includeItemResults - Whether to include individual item details (default: true)
+   *
+   * @returns Promise resolving to formatted string representation
+   *
+   * @example Output format
+   * ```
+   * 1. Item 1:
+   *    Input:    What is the capital of France?
+   *    Expected: Paris
+   *    Actual:   Paris
+   *    Scores:
+   *      • exact_match: 1.000
+   *      • similarity: 0.95
+   *        💭 Very close match with expected output
+   *
+   *    Dataset Item:
+   *    https://cloud.langfuse.com/project/123/datasets/456/items/789
+   *
+   *    Trace:
+   *    https://cloud.langfuse.com/project/123/traces/abc123
+   *
+   * ──────────────────────────────────────────────────
+   * 📊 Translation Quality Test - Testing model accuracy
+   * 2 items
+   * Evaluations:
+   *   • exact_match
+   *   • similarity
+   *
+   * Average Scores:
+   *   • exact_match: 0.850
+   *   • similarity: 0.923
+   *
+   * Run Evaluations:
+   *   • overall_quality: 0.887
+   *     💭 Good performance with room for improvement
+   *
+   * 🔗 Dataset Run:
+   *    https://cloud.langfuse.com/project/123/datasets/456/runs/def456
+   * ```
+   *
+   * @internal
+   */
+  async prettyPrintResults(params) {
+    var _a, _b;
+    const {
+      itemResults,
+      originalData,
+      runEvaluations,
+      name,
+      description,
+      includeItemResults = true
+    } = params;
+    if (itemResults.length === 0) {
+      return "No experiment results to display.";
+    }
+    let output = "";
+    if (includeItemResults) {
+      for (let index = 0; index < itemResults.length; index++) {
+        const result = itemResults[index];
+        const originalItem = originalData[index];
+        output += `
+${index + 1}. Item ${index + 1}:
+`;
+        if ((originalItem == null ? void 0 : originalItem.input) !== void 0) {
+          output += `   Input:    ${this.formatValue(originalItem.input)}
+`;
+        }
+        const expectedOutput = (_b = (_a = originalItem == null ? void 0 : originalItem.expectedOutput) != null ? _a : result.expectedOutput) != null ? _b : null;
+        output += `   Expected: ${expectedOutput !== null ? this.formatValue(expectedOutput) : "null"}
+`;
+        output += `   Actual:   ${this.formatValue(result.output)}
+`;
+        if (result.evaluations.length > 0) {
+          output += `   Scores:
+`;
+          result.evaluations.forEach((evaluation) => {
+            const score = typeof evaluation.value === "number" ? evaluation.value.toFixed(3) : evaluation.value;
+            output += `     \u2022 ${evaluation.name}: ${score}`;
+            if (evaluation.comment) {
+              output += `
+       \u{1F4AD} ${evaluation.comment}`;
+            }
+            output += "\n";
+          });
+        }
+        if (originalItem && "id" in originalItem && "datasetId" in originalItem) {
+          const projectUrl = (await this.langfuseClient.getTraceUrl("mock")).split("/traces")[0];
+          const datasetItemUrl = `${projectUrl}/datasets/${originalItem.datasetId}/items/${originalItem.id}`;
+          output += `
+   Dataset Item:
+   ${datasetItemUrl}
+`;
+        }
+        if (result.traceId) {
+          const traceUrl = await this.langfuseClient.getTraceUrl(
+            result.traceId
+          );
+          output += `
+   Trace:
+   ${traceUrl}
+`;
+        }
+      }
+    } else {
+      output += `Individual Results: Hidden (${itemResults.length} items)
+`;
+      output += "\u{1F4A1} Call prettyPrint({ includeItemResults: true }) to view them\n";
+    }
+    const totalItems = itemResults.length;
+    const evaluationNames = new Set(
+      itemResults.flatMap((r) => r.evaluations.map((e) => e.name))
+    );
+    output += `
+${"\u2500".repeat(50)}
+`;
+    output += `\u{1F4CA} ${name}`;
+    if (description) {
+      output += ` - ${description}`;
+    }
+    output += `
+${totalItems} items`;
+    if (evaluationNames.size > 0) {
+      output += `
+Evaluations:`;
+      Array.from(evaluationNames).forEach((evalName) => {
+        output += `
+  \u2022 ${evalName}`;
+      });
+      output += "\n";
+    }
+    if (evaluationNames.size > 0) {
+      output += `
+Average Scores:`;
+      for (const evalName of evaluationNames) {
+        const scores = itemResults.flatMap((r) => r.evaluations).filter((e) => e.name === evalName && typeof e.value === "number").map((e) => e.value);
+        if (scores.length > 0) {
+          const avg = scores.reduce((a, b) => a + b, 0) / scores.length;
+          output += `
+  \u2022 ${evalName}: ${avg.toFixed(3)}`;
+        }
+      }
+      output += "\n";
+    }
+    if (runEvaluations.length > 0) {
+      output += `
+Run Evaluations:`;
+      runEvaluations.forEach((runEval) => {
+        const score = typeof runEval.value === "number" ? runEval.value.toFixed(3) : runEval.value;
+        output += `
+  \u2022 ${runEval.name}: ${score}`;
+        if (runEval.comment) {
+          output += `
+    \u{1F4AD} ${runEval.comment}`;
+        }
+      });
+      output += "\n";
+    }
+    if (params.datasetRunUrl) {
+      output += `
+\u{1F517} Dataset Run:
+   ${params.datasetRunUrl}`;
+    }
+    return output;
+  }
+  /**
+   * Formats a value for display in pretty-printed output.
+   *
+   * Handles different value types appropriately:
+   * - Strings: Truncates long strings to 50 characters with "..."
+   * - Objects/Arrays: Converts to JSON string representation
+   * - Primitives: Uses toString() representation
+   *
+   * @param value - The value to format
+   * @returns Formatted string representation suitable for display
+   *
+   * @internal
+   */
+  formatValue(value) {
+    if (typeof value === "string") {
+      return value.length > 50 ? `${value.substring(0, 47)}...` : value;
+    }
+    return JSON.stringify(value);
+  }
+  isOtelRegistered() {
+    let tracerProvider = import_api.trace.getTracerProvider();
+    if (tracerProvider instanceof import_api.ProxyTracerProvider) {
+      tracerProvider = tracerProvider.getDelegate();
+    }
+    return tracerProvider.constructor.name !== "NoopTracerProvider";
+  }
+};
+
+// src/media/index.ts
+var import_core2 = require("@langfuse/core");
 var MediaManager = class _MediaManager {
   /**
    * Creates a new MediaManager instance.
@@ -210,14 +757,14 @@ var MediaManager = class _MediaManager {
               const uint8Content = new Uint8Array(
                 await mediaContent.arrayBuffer()
               );
-              const base64MediaContent = (0, import_core.bytesToBase64)(uint8Content);
+              const base64MediaContent = (0, import_core2.bytesToBase64)(uint8Content);
               const base64DataUri = `data:${mediaData.contentType};base64,${base64MediaContent}`;
               referenceStringToMediaContentMap.set(
                 referenceString,
                 base64DataUri
               );
             } catch (error) {
-              (0, import_core.getGlobalLogger)().warn(
+              (0, import_core2.getGlobalLogger)().warn(
                 "Error fetching media content for reference string",
                 referenceString,
                 error
@@ -293,10 +840,10 @@ var MediaManager = class _MediaManager {
 };
 
 // src/prompt/promptManager.ts
-var import_core3 = require("@langfuse/core");
+var import_core4 = require("@langfuse/core");
 
 // src/prompt/promptCache.ts
-var import_core2 = require("@langfuse/core");
+var import_core3 = require("@langfuse/core");
 var DEFAULT_PROMPT_CACHE_TTL_SECONDS = 60;
 var LangfusePromptCacheItem = class {
   constructor(value, ttlSeconds) {
@@ -348,7 +895,7 @@ var LangfusePromptCache = class {
     return this._refreshingKeys.has(key);
   }
   invalidate(promptName) {
-    (0, import_core2.getGlobalLogger)().debug(
+    (0, import_core3.getGlobalLogger)().debug(
       "Invalidating cache keys for",
       promptName,
       this._cache.keys()
@@ -692,7 +1239,7 @@ var PromptManager = class {
     this.cache = new LangfusePromptCache();
   }
   get logger() {
-    return (0, import_core3.getGlobalLogger)();
+    return (0, import_core4.getGlobalLogger)();
   }
   /**
    * Creates a new prompt in Langfuse.
@@ -919,8 +1466,8 @@ var PromptManager = class {
 };
 
 // src/score/index.ts
-var import_core4 = require("@langfuse/core");
-var import_api = require("@opentelemetry/api");
+var import_core5 = require("@langfuse/core");
+var import_api2 = require("@opentelemetry/api");
 var MAX_QUEUE_SIZE = 1e5;
 var MAX_BATCH_SIZE = 100;
 var ScoreManager = class {
@@ -935,13 +1482,13 @@ var ScoreManager = class {
     this.flushPromise = null;
     this.flushTimer = null;
     this.apiClient = params.apiClient;
-    const envFlushAtCount = (0, import_core4.getEnv)("LANGFUSE_FLUSH_AT");
-    const envFlushIntervalSeconds = (0, import_core4.getEnv)("LANGFUSE_FLUSH_INTERVAL");
+    const envFlushAtCount = (0, import_core5.getEnv)("LANGFUSE_FLUSH_AT");
+    const envFlushIntervalSeconds = (0, import_core5.getEnv)("LANGFUSE_FLUSH_INTERVAL");
     this.flushAtCount = envFlushAtCount ? Number(envFlushAtCount) : 10;
     this.flushIntervalSeconds = envFlushIntervalSeconds ? Number(envFlushIntervalSeconds) : 1;
   }
   get logger() {
-    return (0, import_core4.getGlobalLogger)();
+    return (0, import_core5.getGlobalLogger)();
   }
   /**
    * Creates a new score event and adds it to the processing queue.
@@ -966,11 +1513,11 @@ var ScoreManager = class {
     var _a, _b;
     const scoreData = {
       ...data,
-      id: (_a = data.id) != null ? _a : (0, import_core4.generateUUID)(),
-      environment: (_b = data.environment) != null ? _b : (0, import_core4.getEnv)("LANGFUSE_TRACING_ENVIRONMENT")
+      id: (_a = data.id) != null ? _a : (0, import_core5.generateUUID)(),
+      environment: (_b = data.environment) != null ? _b : (0, import_core5.getEnv)("LANGFUSE_TRACING_ENVIRONMENT")
     };
     const scoreIngestionEvent = {
-      id: (0, import_core4.generateUUID)(),
+      id: (0, import_core5.generateUUID)(),
       type: "score-create",
       timestamp: (/* @__PURE__ */ new Date()).toISOString(),
       body: scoreData
@@ -982,10 +1529,14 @@ var ScoreManager = class {
       return;
     }
     this.eventQueue.push(scoreIngestionEvent);
+    this.logger.debug(
+      "Added score event to queue:\n",
+      JSON.stringify(scoreIngestionEvent, null, 2)
+    );
     if (this.eventQueue.length >= this.flushAtCount) {
       this.flushPromise = this.flush();
     } else if (!this.flushTimer) {
-      this.flushTimer = (0, import_core4.safeSetTimeout)(() => {
+      this.flushTimer = (0, import_core5.safeSetTimeout)(() => {
         this.flushPromise = this.flush();
       }, this.flushIntervalSeconds * 1e3);
     }
@@ -1068,7 +1619,7 @@ var ScoreManager = class {
    * ```
    */
   activeObservation(data) {
-    const currentOtelSpan = import_api.trace.getActiveSpan();
+    const currentOtelSpan = import_api2.trace.getActiveSpan();
     if (!currentOtelSpan) {
       this.logger.warn("No active span in context to score.");
       return;
@@ -1104,7 +1655,7 @@ var ScoreManager = class {
    * ```
    */
   activeTrace(data) {
-    const currentOtelSpan = import_api.trace.getActiveSpan();
+    const currentOtelSpan = import_api2.trace.getActiveSpan();
     if (!currentOtelSpan) {
       this.logger.warn("No active span in context to score trace.");
       return;
@@ -1204,10 +1755,10 @@ var LangfuseClient = class {
   constructor(params) {
     this.projectId = null;
     var _a, _b, _c, _d, _e, _f, _g;
-    const logger = (0, import_core5.getGlobalLogger)();
-    const publicKey = (_a = params == null ? void 0 : params.publicKey) != null ? _a : (0, import_core5.getEnv)("LANGFUSE_PUBLIC_KEY");
-    const secretKey = (_b = params == null ? void 0 : params.secretKey) != null ? _b : (0, import_core5.getEnv)("LANGFUSE_SECRET_KEY");
-    this.baseUrl = (_e = (_d = (_c = params == null ? void 0 : params.baseUrl) != null ? _c : (0, import_core5.getEnv)("LANGFUSE_BASE_URL")) != null ? _d : (0, import_core5.getEnv)("LANGFUSE_BASEURL")) != null ? _e : (
+    const logger = (0, import_core6.getGlobalLogger)();
+    const publicKey = (_a = params == null ? void 0 : params.publicKey) != null ? _a : (0, import_core6.getEnv)("LANGFUSE_PUBLIC_KEY");
+    const secretKey = (_b = params == null ? void 0 : params.secretKey) != null ? _b : (0, import_core6.getEnv)("LANGFUSE_SECRET_KEY");
+    this.baseUrl = (_e = (_d = (_c = params == null ? void 0 : params.baseUrl) != null ? _c : (0, import_core6.getEnv)("LANGFUSE_BASE_URL")) != null ? _d : (0, import_core6.getEnv)("LANGFUSE_BASEURL")) != null ? _e : (
       // legacy v2
       "https://cloud.langfuse.com"
     );
@@ -1221,13 +1772,13 @@ var LangfuseClient = class {
         "No secret key provided in constructor or as LANGFUSE_SECRET_KEY env var. Client operations will fail."
       );
     }
-    const timeoutSeconds = (_g = params == null ? void 0 : params.timeout) != null ? _g : Number((_f = (0, import_core5.getEnv)("LANGFUSE_TIMEOUT")) != null ? _f : 5);
-    this.api = new import_core5.LangfuseAPIClient({
+    const timeoutSeconds = (_g = params == null ? void 0 : params.timeout) != null ? _g : Number((_f = (0, import_core6.getEnv)("LANGFUSE_TIMEOUT")) != null ? _f : 5);
+    this.api = new import_core6.LangfuseAPIClient({
       baseUrl: this.baseUrl,
       username: publicKey,
       password: secretKey,
       xLangfusePublicKey: publicKey,
-      xLangfuseSdkVersion: import_core5.LANGFUSE_SDK_VERSION,
+      xLangfuseSdkVersion: import_core6.LANGFUSE_SDK_VERSION,
       xLangfuseSdkName: "javascript",
       environment: "",
       // noop as baseUrl is set
@@ -1239,9 +1790,10 @@ var LangfuseClient = class {
       timeoutSeconds
     });
     this.prompt = new PromptManager({ apiClient: this.api });
-    this.dataset = new DatasetManager({ apiClient: this.api });
+    this.dataset = new DatasetManager({ langfuseClient: this });
     this.score = new ScoreManager({ apiClient: this.api });
     this.media = new MediaManager({ apiClient: this.api });
+    this.experiment = new ExperimentManager({ langfuseClient: this });
     this.getPrompt = this.prompt.get.bind(this.prompt);
     this.createPrompt = this.prompt.create.bind(this.prompt);
     this.updatePrompt = this.prompt.update.bind(this.prompt);
@@ -1316,15 +1868,36 @@ var LangfuseClient = class {
     return traceUrl;
   }
 };
+
+// src/experiment/adapters.ts
+function autoevalsToLangfuseEvaluator(autoevalEvaluator, params) {
+  const langfuseEvaluator = async (langfuseEvaluatorParams) => {
+    var _a;
+    const score = await autoevalEvaluator({
+      ...params != null ? params : {},
+      input: langfuseEvaluatorParams.input,
+      output: langfuseEvaluatorParams.output,
+      expected: langfuseEvaluatorParams.expectedOutput
+    });
+    return {
+      name: score.name,
+      value: (_a = score.score) != null ? _a : 0,
+      metadata: score.metadata
+    };
+  };
+  return langfuseEvaluator;
+}
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
   ChatMessageType,
   ChatPromptClient,
   DatasetManager,
+  ExperimentManager,
   LangfuseClient,
   MediaManager,
   PromptManager,
   ScoreManager,
-  TextPromptClient
+  TextPromptClient,
+  autoevalsToLangfuseEvaluator
 });
 //# sourceMappingURL=index.cjs.map