@arizeai/phoenix-evals 0.7.0 → 0.9.0

package/dist/src/llm/createToolResponseHandlingEvaluator.d.ts

@@ -0,0 +1,78 @@
+import { CreateClassificationEvaluatorArgs } from "../types/evals";
+import { ClassificationEvaluator } from "./ClassificationEvaluator";
+export interface ToolResponseHandlingEvaluatorArgs<RecordType extends Record<string, unknown> = ToolResponseHandlingEvaluationRecord> extends Omit<CreateClassificationEvaluatorArgs<RecordType>, "promptTemplate" | "choices" | "optimizationDirection" | "name"> {
+    optimizationDirection?: CreateClassificationEvaluatorArgs<RecordType>["optimizationDirection"];
+    name?: CreateClassificationEvaluatorArgs<RecordType>["name"];
+    choices?: CreateClassificationEvaluatorArgs<RecordType>["choices"];
+    promptTemplate?: CreateClassificationEvaluatorArgs<RecordType>["promptTemplate"];
+}
+/**
+ * A record to be evaluated by the tool response handling evaluator.
+ */
+export type ToolResponseHandlingEvaluationRecord = {
+    /**
+     * The user query or conversation context.
+     */
+    input: string;
+    /**
+     * The tool invocation(s) made by the agent, including arguments.
+     */
+    toolCall: string;
+    /**
+     * The tool's response (data, errors, or partial results).
+     */
+    toolResult: string;
+    /**
+     * The agent's handling after receiving the tool result
+     * (may include retries, follow-ups, or final response).
+     */
+    output: string;
+};
+/**
+ * Creates a tool response handling evaluator function.
+ *
+ * This function returns an evaluator that determines whether an AI agent properly
+ * handled a tool's response, including error handling, data extraction,
+ * transformation, and safe information disclosure.
+ *
+ * @param args - The arguments for creating the tool response handling evaluator.
+ * @param args.model - The model to use for classification.
+ * @param args.choices - The possible classification choices (defaults to correct/incorrect).
+ * @param args.promptTemplate - The prompt template to use.
+ * @param args.telemetry - The telemetry to use for the evaluator.
+ *
+ * @returns An evaluator function that takes a {@link ToolResponseHandlingEvaluationRecord}
+ * and returns a classification result indicating whether the tool response handling
+ * is correct or incorrect.
+ *
+ * @example
+ * ```ts
+ * const evaluator = createToolResponseHandlingEvaluator({ model: openai("gpt-4o-mini") });
+ *
+ * // Example: Correct extraction from tool result
+ * const result = await evaluator.evaluate({
+ *   input: "What's the weather in Seattle?",
+ *   toolCall: 'get_weather(location="Seattle")',
+ *   toolResult: JSON.stringify({
+ *     temperature: 58,
+ *     unit: "fahrenheit",
+ *     conditions: "partly cloudy"
+ *   }),
+ *   output: "The weather in Seattle is 58°F and partly cloudy."
+ * });
+ * console.log(result.label); // "correct"
+ *
+ * // Example: Hallucinated data (incorrect)
+ * const resultHallucinated = await evaluator.evaluate({
+ *   input: "What restaurants are nearby?",
+ *   toolCall: 'search_restaurants(location="downtown")',
+ *   toolResult: JSON.stringify({
+ *     results: [{ name: "Cafe Luna", rating: 4.2 }]
+ *   }),
+ *   output: "I found Cafe Luna (4.2 stars) and Mario's Italian (4.8 stars) nearby."
+ * });
+ * console.log(resultHallucinated.label); // "incorrect" - Mario's was hallucinated
+ * ```
+ */
+export declare function createToolResponseHandlingEvaluator<RecordType extends Record<string, unknown> = ToolResponseHandlingEvaluationRecord>(args: ToolResponseHandlingEvaluatorArgs<RecordType>): ClassificationEvaluator<RecordType>;
+//# sourceMappingURL=createToolResponseHandlingEvaluator.d.ts.map

package/dist/src/llm/createToolResponseHandlingEvaluator.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"createToolResponseHandlingEvaluator.d.ts","sourceRoot":"","sources":["../../../src/llm/createToolResponseHandlingEvaluator.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,iCAAiC,EAAE,MAAM,gBAAgB,CAAC;AAEnE,OAAO,EAAE,uBAAuB,EAAE,MAAM,2BAA2B,CAAC;AAGpE,MAAM,WAAW,iCAAiC,CAChD,UAAU,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GACxC,oCAAoC,CACtC,SAAQ,IAAI,CACZ,iCAAiC,CAAC,UAAU,CAAC,EAC7C,gBAAgB,GAAG,SAAS,GAAG,uBAAuB,GAAG,MAAM,CAChE;IACC,qBAAqB,CAAC,EAAE,iCAAiC,CAAC,UAAU,CAAC,CAAC,uBAAuB,CAAC,CAAC;IAC/F,IAAI,CAAC,EAAE,iCAAiC,CAAC,UAAU,CAAC,CAAC,MAAM,CAAC,CAAC;IAC7D,OAAO,CAAC,EAAE,iCAAiC,CAAC,UAAU,CAAC,CAAC,SAAS,CAAC,CAAC;IACnE,cAAc,CAAC,EAAE,iCAAiC,CAAC,UAAU,CAAC,CAAC,gBAAgB,CAAC,CAAC;CAClF;AAED;;GAEG;AACH,MAAM,MAAM,oCAAoC,GAAG;IACjD;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,QAAQ,EAAE,MAAM,CAAC;IACjB;;OAEG;IACH,UAAU,EAAE,MAAM,CAAC;IACnB;;;OAGG;IACH,MAAM,EAAE,MAAM,CAAC;CAChB,CAAC;AAEF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA6CG;AACH,wBAAgB,mCAAmC,CACjD,UAAU,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GACxC,oCAAoC,EAEtC,IAAI,EAAE,iCAAiC,CAAC,UAAU,CAAC,GAClD,uBAAuB,CAAC,UAAU,CAAC,CAerC"}

package/dist/src/llm/createToolResponseHandlingEvaluator.js

@@ -0,0 +1,70 @@
+"use strict";
+var __rest = (this && this.__rest) || function (s, e) {
+    var t = {};
+    for (var p in s) if (Object.prototype.hasOwnProperty.call(s, p) && e.indexOf(p) < 0)
+        t[p] = s[p];
+    if (s != null && typeof Object.getOwnPropertySymbols === "function")
+        for (var i = 0, p = Object.getOwnPropertySymbols(s); i < p.length; i++) {
+            if (e.indexOf(p[i]) < 0 && Object.prototype.propertyIsEnumerable.call(s, p[i]))
+                t[p[i]] = s[p[i]];
+        }
+    return t;
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.createToolResponseHandlingEvaluator = createToolResponseHandlingEvaluator;
+const default_templates_1 = require("../__generated__/default_templates");
+const createClassificationEvaluator_1 = require("./createClassificationEvaluator");
+/**
+ * Creates a tool response handling evaluator function.
+ *
+ * This function returns an evaluator that determines whether an AI agent properly
+ * handled a tool's response, including error handling, data extraction,
+ * transformation, and safe information disclosure.
+ *
+ * @param args - The arguments for creating the tool response handling evaluator.
+ * @param args.model - The model to use for classification.
+ * @param args.choices - The possible classification choices (defaults to correct/incorrect).
+ * @param args.promptTemplate - The prompt template to use.
+ * @param args.telemetry - The telemetry to use for the evaluator.
+ *
+ * @returns An evaluator function that takes a {@link ToolResponseHandlingEvaluationRecord}
+ * and returns a classification result indicating whether the tool response handling
+ * is correct or incorrect.
+ *
+ * @example
+ * ```ts
+ * const evaluator = createToolResponseHandlingEvaluator({ model: openai("gpt-4o-mini") });
+ *
+ * // Example: Correct extraction from tool result
+ * const result = await evaluator.evaluate({
+ *   input: "What's the weather in Seattle?",
+ *   toolCall: 'get_weather(location="Seattle")',
+ *   toolResult: JSON.stringify({
+ *     temperature: 58,
+ *     unit: "fahrenheit",
+ *     conditions: "partly cloudy"
+ *   }),
+ *   output: "The weather in Seattle is 58°F and partly cloudy."
+ * });
+ * console.log(result.label); // "correct"
+ *
+ * // Example: Hallucinated data (incorrect)
+ * const resultHallucinated = await evaluator.evaluate({
+ *   input: "What restaurants are nearby?",
+ *   toolCall: 'search_restaurants(location="downtown")',
+ *   toolResult: JSON.stringify({
+ *     results: [{ name: "Cafe Luna", rating: 4.2 }]
+ *   }),
+ *   output: "I found Cafe Luna (4.2 stars) and Mario's Italian (4.8 stars) nearby."
+ * });
+ * console.log(resultHallucinated.label); // "incorrect" - Mario's was hallucinated
+ * ```
+ */
+function createToolResponseHandlingEvaluator(args) {
+    const { choices = default_templates_1.TOOL_RESPONSE_HANDLING_CLASSIFICATION_EVALUATOR_CONFIG.choices, promptTemplate = default_templates_1.TOOL_RESPONSE_HANDLING_CLASSIFICATION_EVALUATOR_CONFIG.template, optimizationDirection = default_templates_1.TOOL_RESPONSE_HANDLING_CLASSIFICATION_EVALUATOR_CONFIG.optimizationDirection, name = default_templates_1.TOOL_RESPONSE_HANDLING_CLASSIFICATION_EVALUATOR_CONFIG.name } = args, rest = __rest(args, ["choices", "promptTemplate", "optimizationDirection", "name"]);
+    return (0, createClassificationEvaluator_1.createClassificationEvaluator)(Object.assign(Object.assign({}, rest), { promptTemplate,
+        choices,
+        optimizationDirection,
+        name }));
+}
+//# sourceMappingURL=createToolResponseHandlingEvaluator.js.map

package/dist/src/llm/createToolResponseHandlingEvaluator.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"createToolResponseHandlingEvaluator.js","sourceRoot":"","sources":["../../../src/llm/createToolResponseHandlingEvaluator.ts"],"names":[],"mappings":";;;;;;;;;;;;;AAwFA,kFAoBC;AA5GD,0EAA4G;AAI5G,mFAAgF;AAsChF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA6CG;AACH,SAAgB,mCAAmC,CAIjD,IAAmD;IAEnD,MAAM,EACJ,OAAO,GAAG,0EAAsD,CAAC,OAAO,EACxE,cAAc,GAAG,0EAAsD,CAAC,QAAQ,EAChF,qBAAqB,GAAG,0EAAsD,CAAC,qBAAqB,EACpG,IAAI,GAAG,0EAAsD,CAAC,IAAI,KAEhE,IAAI,EADH,IAAI,UACL,IAAI,EANF,8DAML,CAAO,CAAC;IACT,OAAO,IAAA,6DAA6B,kCAC/B,IAAI,KACP,cAAc;QACd,OAAO;QACP,qBAAqB;QACrB,IAAI,IACJ,CAAC;AACL,CAAC"}

package/dist/src/llm/createToolSelectionEvaluator.d.ts

@@ -0,0 +1,64 @@
+import { CreateClassificationEvaluatorArgs } from "../types/evals";
+import { ClassificationEvaluator } from "./ClassificationEvaluator";
+export interface ToolSelectionEvaluatorArgs<RecordType extends Record<string, unknown> = ToolSelectionEvaluationRecord> extends Omit<CreateClassificationEvaluatorArgs<RecordType>, "promptTemplate" | "choices" | "optimizationDirection" | "name"> {
+    optimizationDirection?: CreateClassificationEvaluatorArgs<RecordType>["optimizationDirection"];
+    name?: CreateClassificationEvaluatorArgs<RecordType>["name"];
+    choices?: CreateClassificationEvaluatorArgs<RecordType>["choices"];
+    promptTemplate?: CreateClassificationEvaluatorArgs<RecordType>["promptTemplate"];
+}
+/**
+ * A record to be evaluated by the tool selection evaluator.
+ */
+export type ToolSelectionEvaluationRecord = {
+    /**
+     * The input query or conversation context.
+     */
+    input: string;
+    /**
+     * The available tools that the LLM could use.
+     */
+    availableTools: string;
+    /**
+     * The tool or tools selected by the LLM.
+     */
+    toolSelection: string;
+};
+/**
+ * Creates a tool selection evaluator function.
+ *
+ * This function returns an evaluator that determines whether the correct tool
+ * was selected for a given context. Unlike the tool invocation evaluator which
+ * checks if the tool was called correctly with proper arguments, this evaluator
+ * focuses on whether the right tool was chosen in the first place.
+ *
+ * The evaluator checks for:
+ * - Whether the LLM chose the best available tool for the user query
+ * - Whether the tool name exists in the available tools list
+ * - Whether the correct number of tools were selected for the task
+ * - Whether the tool selection is safe and appropriate
+ *
+ * @param args - The arguments for creating the tool selection evaluator.
+ * @param args.model - The model to use for classification.
+ * @param args.choices - The possible classification choices (defaults to correct/incorrect).
+ * @param args.promptTemplate - The prompt template to use (defaults to TOOL_SELECTION_TEMPLATE).
+ * @param args.telemetry - The telemetry to use for the evaluator.
+ *
+ * @returns An evaluator function that takes a {@link ToolSelectionEvaluationRecord} and returns
+ * a classification result indicating whether the tool selection is correct or incorrect.
+ *
+ * @example
+ * ```ts
+ * const evaluator = createToolSelectionEvaluator({ model: openai("gpt-4o-mini") });
+ *
+ * const result = await evaluator.evaluate({
+ *   input: "User: What is the weather in San Francisco?",
+ *   availableTools: `WeatherTool: Get the current weather for a location.
+ * NewsTool: Stay connected to global events with our up-to-date news around the world.
+ * MusicTool: Create playlists, search for music, and check the latest music trends.`,
+ *   toolSelection: "WeatherTool"
+ * });
+ * console.log(result.label); // "correct" or "incorrect"
+ * ```
+ */
+export declare function createToolSelectionEvaluator<RecordType extends Record<string, unknown> = ToolSelectionEvaluationRecord>(args: ToolSelectionEvaluatorArgs<RecordType>): ClassificationEvaluator<RecordType>;
+//# sourceMappingURL=createToolSelectionEvaluator.d.ts.map

package/dist/src/llm/createToolSelectionEvaluator.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"createToolSelectionEvaluator.d.ts","sourceRoot":"","sources":["../../../src/llm/createToolSelectionEvaluator.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,iCAAiC,EAAE,MAAM,gBAAgB,CAAC;AAEnE,OAAO,EAAE,uBAAuB,EAAE,MAAM,2BAA2B,CAAC;AAGpE,MAAM,WAAW,0BAA0B,CACzC,UAAU,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,6BAA6B,CAC1E,SAAQ,IAAI,CACZ,iCAAiC,CAAC,UAAU,CAAC,EAC7C,gBAAgB,GAAG,SAAS,GAAG,uBAAuB,GAAG,MAAM,CAChE;IACC,qBAAqB,CAAC,EAAE,iCAAiC,CAAC,UAAU,CAAC,CAAC,uBAAuB,CAAC,CAAC;IAC/F,IAAI,CAAC,EAAE,iCAAiC,CAAC,UAAU,CAAC,CAAC,MAAM,CAAC,CAAC;IAC7D,OAAO,CAAC,EAAE,iCAAiC,CAAC,UAAU,CAAC,CAAC,SAAS,CAAC,CAAC;IACnE,cAAc,CAAC,EAAE,iCAAiC,CAAC,UAAU,CAAC,CAAC,gBAAgB,CAAC,CAAC;CAClF;AAED;;GAEG;AACH,MAAM,MAAM,6BAA6B,GAAG;IAC1C;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,cAAc,EAAE,MAAM,CAAC;IACvB;;OAEG;IACH,aAAa,EAAE,MAAM,CAAC;CACvB,CAAC;AAEF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAoCG;AACH,wBAAgB,4BAA4B,CAC1C,UAAU,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,6BAA6B,EAE1E,IAAI,EAAE,0BAA0B,CAAC,UAAU,CAAC,GAC3C,uBAAuB,CAAC,UAAU,CAAC,CAerC"}

package/dist/src/llm/createToolSelectionEvaluator.js

@@ -0,0 +1,61 @@
+"use strict";
+var __rest = (this && this.__rest) || function (s, e) {
+    var t = {};
+    for (var p in s) if (Object.prototype.hasOwnProperty.call(s, p) && e.indexOf(p) < 0)
+        t[p] = s[p];
+    if (s != null && typeof Object.getOwnPropertySymbols === "function")
+        for (var i = 0, p = Object.getOwnPropertySymbols(s); i < p.length; i++) {
+            if (e.indexOf(p[i]) < 0 && Object.prototype.propertyIsEnumerable.call(s, p[i]))
+                t[p[i]] = s[p[i]];
+        }
+    return t;
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.createToolSelectionEvaluator = createToolSelectionEvaluator;
+const default_templates_1 = require("../__generated__/default_templates");
+const createClassificationEvaluator_1 = require("./createClassificationEvaluator");
+/**
+ * Creates a tool selection evaluator function.
+ *
+ * This function returns an evaluator that determines whether the correct tool
+ * was selected for a given context. Unlike the tool invocation evaluator which
+ * checks if the tool was called correctly with proper arguments, this evaluator
+ * focuses on whether the right tool was chosen in the first place.
+ *
+ * The evaluator checks for:
+ * - Whether the LLM chose the best available tool for the user query
+ * - Whether the tool name exists in the available tools list
+ * - Whether the correct number of tools were selected for the task
+ * - Whether the tool selection is safe and appropriate
+ *
+ * @param args - The arguments for creating the tool selection evaluator.
+ * @param args.model - The model to use for classification.
+ * @param args.choices - The possible classification choices (defaults to correct/incorrect).
+ * @param args.promptTemplate - The prompt template to use (defaults to TOOL_SELECTION_TEMPLATE).
+ * @param args.telemetry - The telemetry to use for the evaluator.
+ *
+ * @returns An evaluator function that takes a {@link ToolSelectionEvaluationRecord} and returns
+ * a classification result indicating whether the tool selection is correct or incorrect.
+ *
+ * @example
+ * ```ts
+ * const evaluator = createToolSelectionEvaluator({ model: openai("gpt-4o-mini") });
+ *
+ * const result = await evaluator.evaluate({
+ *   input: "User: What is the weather in San Francisco?",
+ *   availableTools: `WeatherTool: Get the current weather for a location.
+ * NewsTool: Stay connected to global events with our up-to-date news around the world.
+ * MusicTool: Create playlists, search for music, and check the latest music trends.`,
+ *   toolSelection: "WeatherTool"
+ * });
+ * console.log(result.label); // "correct" or "incorrect"
+ * ```
+ */
+function createToolSelectionEvaluator(args) {
+    const { choices = default_templates_1.TOOL_SELECTION_CLASSIFICATION_EVALUATOR_CONFIG.choices, promptTemplate = default_templates_1.TOOL_SELECTION_CLASSIFICATION_EVALUATOR_CONFIG.template, optimizationDirection = default_templates_1.TOOL_SELECTION_CLASSIFICATION_EVALUATOR_CONFIG.optimizationDirection, name = default_templates_1.TOOL_SELECTION_CLASSIFICATION_EVALUATOR_CONFIG.name } = args, rest = __rest(args, ["choices", "promptTemplate", "optimizationDirection", "name"]);
+    return (0, createClassificationEvaluator_1.createClassificationEvaluator)(Object.assign(Object.assign({}, rest), { promptTemplate,
+        choices,
+        optimizationDirection,
+        name }));
+}
+//# sourceMappingURL=createToolSelectionEvaluator.js.map

package/dist/src/llm/createToolSelectionEvaluator.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"createToolSelectionEvaluator.js","sourceRoot":"","sources":["../../../src/llm/createToolSelectionEvaluator.ts"],"names":[],"mappings":";;;;;;;;;;;;;AAyEA,oEAmBC;AA5FD,0EAAoG;AAIpG,mFAAgF;AAgChF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAoCG;AACH,SAAgB,4BAA4B,CAG1C,IAA4C;IAE5C,MAAM,EACJ,OAAO,GAAG,kEAA8C,CAAC,OAAO,EAChE,cAAc,GAAG,kEAA8C,CAAC,QAAQ,EACxE,qBAAqB,GAAG,kEAA8C,CAAC,qBAAqB,EAC5F,IAAI,GAAG,kEAA8C,CAAC,IAAI,KAExD,IAAI,EADH,IAAI,UACL,IAAI,EANF,8DAML,CAAO,CAAC;IACT,OAAO,IAAA,6DAA6B,kCAC/B,IAAI,KACP,cAAc;QACd,OAAO;QACP,qBAAqB;QACrB,IAAI,IACJ,CAAC;AACL,CAAC"}

package/dist/src/llm/index.d.ts

@@ -1,9 +1,13 @@
-export * from "./generateClassification";
-export * from "./createClassifierFn";
+export * from "./ClassificationEvaluator";
 export * from "./createClassificationEvaluator";
-export * from "./createHallucinationEvaluator";
-export * from "./createDocumentRelevanceEvaluator";
+export * from "./createClassifierFn";
 export * from "./createCorrectnessEvaluator";
-export * from "./ClassificationEvaluator";
+export * from "./createDocumentRelevanceEvaluator";
+export * from "./createFaithfulnessEvaluator";
+export * from "./createHallucinationEvaluator";
+export * from "./createToolInvocationEvaluator";
+export * from "./createToolResponseHandlingEvaluator";
+export * from "./createToolSelectionEvaluator";
+export * from "./generateClassification";
 export * from "./LLMEvaluator";
 //# sourceMappingURL=index.d.ts.map

package/dist/src/llm/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/llm/index.ts"],"names":[],"mappings":"AAAA,cAAc,0BAA0B,CAAC;AACzC,cAAc,sBAAsB,CAAC;AACrC,cAAc,iCAAiC,CAAC;AAChD,cAAc,gCAAgC,CAAC;AAC/C,cAAc,oCAAoC,CAAC;AACnD,cAAc,8BAA8B,CAAC;AAC7C,cAAc,2BAA2B,CAAC;AAC1C,cAAc,gBAAgB,CAAC"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/llm/index.ts"],"names":[],"mappings":"AAAA,cAAc,2BAA2B,CAAC;AAC1C,cAAc,iCAAiC,CAAC;AAChD,cAAc,sBAAsB,CAAC;AACrC,cAAc,8BAA8B,CAAC;AAC7C,cAAc,oCAAoC,CAAC;AACnD,cAAc,+BAA+B,CAAC;AAC9C,cAAc,gCAAgC,CAAC;AAC/C,cAAc,iCAAiC,CAAC;AAChD,cAAc,uCAAuC,CAAC;AACtD,cAAc,gCAAgC,CAAC;AAC/C,cAAc,0BAA0B,CAAC;AACzC,cAAc,gBAAgB,CAAC"}

package/dist/src/llm/index.js

@@ -14,12 +14,16 @@ var __exportStar = (this && this.__exportStar) || function(m, exports) {
     for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-__exportStar(require("./generateClassification"), exports);
-__exportStar(require("./createClassifierFn"), exports);
+__exportStar(require("./ClassificationEvaluator"), exports);
 __exportStar(require("./createClassificationEvaluator"), exports);
-__exportStar(require("./createHallucinationEvaluator"), exports);
-__exportStar(require("./createDocumentRelevanceEvaluator"), exports);
+__exportStar(require("./createClassifierFn"), exports);
 __exportStar(require("./createCorrectnessEvaluator"), exports);
-__exportStar(require("./ClassificationEvaluator"), exports);
+__exportStar(require("./createDocumentRelevanceEvaluator"), exports);
+__exportStar(require("./createFaithfulnessEvaluator"), exports);
+__exportStar(require("./createHallucinationEvaluator"), exports); // Deprecated: use createFaithfulnessEvaluator
+__exportStar(require("./createToolInvocationEvaluator"), exports);
+__exportStar(require("./createToolResponseHandlingEvaluator"), exports);
+__exportStar(require("./createToolSelectionEvaluator"), exports);
+__exportStar(require("./generateClassification"), exports);
 __exportStar(require("./LLMEvaluator"), exports);
 //# sourceMappingURL=index.js.map

package/dist/src/llm/index.js.map

@@ -1 +1 @@
-{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/llm/index.ts"],"names":[],"mappings":";;;;;;;;;;;;;;;;AAAA,2DAAyC;AACzC,uDAAqC;AACrC,kEAAgD;AAChD,iEAA+C;AAC/C,qEAAmD;AACnD,+DAA6C;AAC7C,4DAA0C;AAC1C,iDAA+B"}
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/llm/index.ts"],"names":[],"mappings":";;;;;;;;;;;;;;;;AAAA,4DAA0C;AAC1C,kEAAgD;AAChD,uDAAqC;AACrC,+DAA6C;AAC7C,qEAAmD;AACnD,gEAA8C;AAC9C,iEAA+C,CAAC,8CAA8C;AAC9F,kEAAgD;AAChD,wEAAsD;AACtD,iEAA+C;AAC/C,2DAAyC;AACzC,iDAA+B"}

package/dist/src/types/evals.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"evals.d.ts","sourceRoot":"","sources":["../../../src/types/evals.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,aAAa,EAAE,MAAM,QAAQ,CAAC;AACvC,OAAO,EAAE,aAAa,EAAE,MAAM,QAAQ,CAAC;AACvC,OAAO,EAAE,cAAc,EAAE,MAAM,cAAc,CAAC;AAE9C,OAAO,EAAE,aAAa,EAAE,MAAM,IAAI,CAAC;AAEnC;;GAEG;AACH,MAAM,WAAW,aAAa,CAAC,UAAU,EAAE,SAAS;IAClD,MAAM,EAAE,UAAU,CAAC;IACnB,QAAQ,CAAC,EAAE,UAAU,CAAC;IACtB,KAAK,CAAC,EAAE,SAAS,CAAC;IAClB,CAAC,GAAG,EAAE,MAAM,GAAG,OAAO,CAAC;CACxB;AAED,MAAM,WAAW,OAAO;IACtB,KAAK,EAAE,aAAa,CAAC;CACtB;AAGD,MAAM,WAAW,iBAAkB,SAAQ,OAAO;CAAG;AAErD;;GAEG;AACH,MAAM,WAAW,gBAAgB;IAC/B;;;OAGG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;;OAGG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;;OAGG;IACH,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED;;GAEG;AACH,MAAM,WAAW,oBAAoB;IACnC,KAAK,EAAE,MAAM,CAAC;IACd,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED;;GAEG;AACH,MAAM,WAAW,oBAAoB;IACnC,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,MAAM,CAAC;CACf;AAED;;GAEG;AACH,MAAM,MAAM,wBAAwB,GAAG,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;AAE9D;;GAEG;AACH,MAAM,WAAW,oBAAqB,SAAQ,aAAa;IAIzD,KAAK,EAAE,aAAa,CAAC;IACrB;;;OAGG;IACH,OAAO,EAAE,wBAAwB,CAAC;IAClC;;OAEG;IACH,cAAc,EAAE,cAAc,CAAC;CAChC;AAED,MAAM,WAAW,mBAAmB,CAClC,WAAW,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CACrE,SAAQ,aAAa;IACrB;;;OAGG;IACH,IAAI,EAAE,MAAM,CAAC;IACb;;OAEG;IACH,IAAI,EAAE,cAAc,CAAC;IACrB;;;OAGG;IACH,qBAAqB,CAAC,EAAE,qBAAqB,CAAC;IAC9C;;OAEG;IACH,YAAY,CAAC,EAAE,aAAa,CAAC,WAAW,CAAC,CAAC;CAC3C;AAED,MAAM,MAAM,sBAAsB,CAAC,UAAU,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,IAC3E,IAAI,CAAC,mBAAmB,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC;AAEhD,MAAM,WAAW,iCAAiC,CAChD,UAAU,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAC1C,SAAQ,oBAAoB,EAC1B,sBAAsB,CAAC,UAAU,CAAC;IACpC;;OAEG;IACH,cAAc,EAAE,cAAc,CAAC;CAChC;AAED,MAAM,MAAM,WAAW,CAAC,WAAW,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,IAAI,CACrE,IAAI,EAAE,WAAW,KACd,OAAO,CAAC,gBAAgB,CAAC,CAAC;AAE/B;;GAEG;AACH,MAAM,MAAM,cAAc,GAAG,KAAK,GAAG,MAAM,CAAC;AAE5C;;;GAGG;AACH,MAAM,MAAM,qBAAqB,GAAG,UAAU,GAAG,UAAU,CAAC;AAE5D;;GAEG;AACH,UAAU,oBAAoB;IAC5B;;OAEG;IACH,IAAI,EAAE,MAAM,CAAC;IACb;;OAEG;IACH,IAAI,EAAE,cAAc,CAAC;IACrB;;;OAGG;IACH,qBAAqB,CAAC,EAAE,qBAAqB,CAAC;CAC/C;AAED;;;GAGG;AACH,MAAM,WAAW,kBAAkB,CAAC,WAAW,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAC7E,SAAQ,oBAAoB;IAC5B;;OAEG;IACH,QAAQ,EAAE,WAAW,CAAC,WAAW,CAAC,CAAC;CACpC"}
+{"version":3,"file":"evals.d.ts","sourceRoot":"","sources":["../../../src/types/evals.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,aAAa,EAAE,MAAM,QAAQ,CAAC;AACvC,OAAO,EAAE,aAAa,EAAE,MAAM,QAAQ,CAAC;AACvC,OAAO,EAAE,cAAc,EAAE,MAAM,cAAc,CAAC;AAE9C,OAAO,EAAE,aAAa,EAAE,MAAM,IAAI,CAAC;AAEnC;;GAEG;AACH,MAAM,WAAW,aAAa,CAAC,UAAU,EAAE,SAAS;IAClD,MAAM,EAAE,UAAU,CAAC;IACnB,QAAQ,CAAC,EAAE,UAAU,CAAC;IACtB,KAAK,CAAC,EAAE,SAAS,CAAC;IAClB,CAAC,GAAG,EAAE,MAAM,GAAG,OAAO,CAAC;CACxB;AAED,MAAM,WAAW,OAAO;IACtB,KAAK,EAAE,aAAa,CAAC;CACtB;AAGD,MAAM,WAAW,iBAAkB,SAAQ,OAAO;CAAG;AAErD;;GAEG;AACH,MAAM,WAAW,gBAAgB;IAC/B;;;OAGG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;;OAGG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;;OAGG;IACH,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED;;GAEG;AACH,MAAM,WAAW,oBAAoB;IACnC,KAAK,EAAE,MAAM,CAAC;IACd,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED;;GAEG;AACH,MAAM,WAAW,oBAAoB;IACnC,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,MAAM,CAAC;CACf;AAED;;GAEG;AACH,MAAM,MAAM,wBAAwB,GAAG,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;AAE9D;;GAEG;AACH,MAAM,WAAW,oBAAqB,SAAQ,aAAa;IAIzD,KAAK,EAAE,aAAa,CAAC;IACrB;;;OAGG;IACH,OAAO,EAAE,wBAAwB,CAAC;IAClC;;OAEG;IACH,cAAc,EAAE,cAAc,CAAC;CAChC;AAED,MAAM,WAAW,mBAAmB,CAClC,WAAW,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CACrE,SAAQ,aAAa;IACrB;;;OAGG;IACH,IAAI,EAAE,MAAM,CAAC;IACb;;OAEG;IACH,IAAI,EAAE,cAAc,CAAC;IACrB;;;OAGG;IACH,qBAAqB,CAAC,EAAE,qBAAqB,CAAC;IAC9C;;OAEG;IACH,YAAY,CAAC,EAAE,aAAa,CAAC,WAAW,CAAC,CAAC;CAC3C;AAED,MAAM,MAAM,sBAAsB,CAAC,UAAU,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,IAC3E,IAAI,CAAC,mBAAmB,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC;AAEhD,MAAM,WAAW,iCAAiC,CAChD,UAAU,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAE1C,SAAQ,oBAAoB,EAAE,sBAAsB,CAAC,UAAU,CAAC;IAChE;;OAEG;IACH,cAAc,EAAE,cAAc,CAAC;CAChC;AAED,MAAM,MAAM,WAAW,CAAC,WAAW,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,IAAI,CACrE,IAAI,EAAE,WAAW,KACd,OAAO,CAAC,gBAAgB,CAAC,CAAC;AAE/B;;GAEG;AACH,MAAM,MAAM,cAAc,GAAG,KAAK,GAAG,MAAM,CAAC;AAE5C;;;GAGG;AACH,MAAM,MAAM,qBAAqB,GAAG,UAAU,GAAG,UAAU,CAAC;AAE5D;;GAEG;AACH,UAAU,oBAAoB;IAC5B;;OAEG;IACH,IAAI,EAAE,MAAM,CAAC;IACb;;OAEG;IACH,IAAI,EAAE,cAAc,CAAC;IACrB;;;OAGG;IACH,qBAAqB,CAAC,EAAE,qBAAqB,CAAC;CAC/C;AAED;;;GAGG;AACH,MAAM,WAAW,kBAAkB,CACjC,WAAW,SAAS,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAC3C,SAAQ,oBAAoB;IAC5B;;OAEG;IACH,QAAQ,EAAE,WAAW,CAAC,WAAW,CAAC,CAAC;CACpC"}