@hazeljs/eval 0.8.1

package/LICENSE

@@ -0,0 +1,192 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   Copyright 2024 HazelJS Team
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
+
+---
+
+TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+1. Definitions.
+
+   "License" shall mean the terms and conditions for use, reproduction,
+   and distribution as defined by Sections 1 through 9 of this document.
+
+   "Licensor" shall mean the copyright owner or entity authorized by
+   the copyright owner that is granting the License.
+
+   "Legal Entity" shall mean the union of the acting entity and all
+   other entities that control, are controlled by, or are under common
+   control with that entity. For the purposes of this definition,
+   "control" means (i) the power, direct or indirect, to cause the
+   direction or management of such entity, whether by contract or
+   otherwise, or (ii) ownership of fifty percent (50%) or more of the
+   outstanding shares, or (iii) beneficial ownership of such entity.
+
+   "You" (or "Your") shall mean an individual or Legal Entity
+   exercising permissions granted by this License.
+
+   "Source" form shall mean the preferred form for making modifications,
+   including but not limited to software source code, documentation
+   source, and configuration files.
+
+   "Object" form shall mean any form resulting from mechanical
+   transformation or translation of a Source form, including but
+   not limited to compiled object code, generated documentation,
+   and conversions to other media types.
+
+   "Work" shall mean the work of authorship, whether in Source or
+   Object form, made available under the License, as indicated by a
+   copyright notice that is included in or attached to the work
+   (an example is provided in the Appendix below).
+
+   "Derivative Works" shall mean any work, whether in Source or Object
+   form, that is based on (or derived from) the Work and for which the
+   editorial revisions, annotations, elaborations, or other modifications
+   represent, as a whole, an original work of authorship. For the purposes
+   of this License, Derivative Works shall not include works that remain
+   separable from, or merely link (or bind by name) to the interfaces of,
+   the Work and Derivative Works thereof.
+
+   "Contribution" shall mean any work of authorship, including
+   the original version of the Work and any modifications or additions
+   to that Work or Derivative Works thereof, that is intentionally
+   submitted to Licensor for inclusion in the Work by the copyright owner
+   or by an individual or Legal Entity authorized to submit on behalf of
+   the copyright owner. For the purposes of this definition, "submitted"
+   means any form of electronic, verbal, or written communication sent
+   to the Licensor or its representatives, including but not limited to
+   communication on electronic mailing lists, source code control systems,
+   and issue tracking systems that are managed by, or on behalf of, the
+   Licensor for the purpose of discussing and improving the Work, but
+   excluding communication that is conspicuously marked or otherwise
+   designated in writing by the copyright owner as "Not a Contribution."
+
+   "Contributor" shall mean Licensor and any individual or Legal Entity
+   on behalf of whom a Contribution has been received by Licensor and
+   subsequently incorporated within the Work.
+
+2. Grant of Copyright License. Subject to the terms and conditions of
+   this License, each Contributor hereby grants to You a perpetual,
+   worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+   copyright license to reproduce, prepare Derivative Works of,
+   publicly display, publicly perform, sublicense, and distribute the
+   Work and such Derivative Works in Source or Object form.
+
+3. Grant of Patent License. Subject to the terms and conditions of
+   this License, each Contributor hereby grants to You a perpetual,
+   worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+   (except as stated in this section) patent license to make, have made,
+   use, offer to sell, sell, import, and otherwise transfer the Work,
+   where such license applies only to those patent claims licensable
+   by such Contributor that are necessarily infringed by their
+   Contribution(s) alone or by combination of their Contribution(s)
+   with the Work to which such Contribution(s) was submitted. If You
+   institute patent litigation against any entity (including a
+   cross-claim or counterclaim in a lawsuit) alleging that the Work
+   or a Contribution incorporated within the Work constitutes direct
+   or contributory patent infringement, then any patent licenses
+   granted to You under this License for that Work shall terminate
+   as of the date such litigation is filed.
+
+4. Redistribution. You may reproduce and distribute copies of the
+   Work or Derivative Works thereof in any medium, with or without
+   modifications, and in Source or Object form, provided that You
+   meet the following conditions:
+
+   (a) You must give any other recipients of the Work or
+   Derivative Works a copy of this License; and
+
+   (b) You must cause any modified files to carry prominent notices
+   stating that You changed the files; and
+
+   (c) You must retain, in the Source form of any Derivative Works
+   that You distribute, all copyright, patent, trademark, and
+   attribution notices from the Source form of the Work,
+   excluding those notices that do not pertain to any part of
+   the Derivative Works; and
+
+   (d) If the Work includes a "NOTICE" text file as part of its
+   distribution, then any Derivative Works that You distribute must
+   include a readable copy of the attribution notices contained
+   within such NOTICE file, excluding those notices that do not
+   pertain to any part of the Derivative Works, in at least one
+   of the following places: within a NOTICE text file distributed
+   as part of the Derivative Works; within the Source form or
+   documentation, if provided along with the Derivative Works; or,
+   within a display generated by the Derivative Works, if and
+   wherever such third-party notices normally appear. The contents
+   of the NOTICE file are for informational purposes only and
+   do not modify the License. You may add Your own attribution
+   notices within Derivative Works that You distribute, alongside
+   or as an addendum to the NOTICE text from the Work, provided
+   that such additional attribution notices cannot be construed
+   as modifying the License.
+
+   You may add Your own copyright statement to Your modifications and
+   may provide additional or different license terms and conditions
+   for use, reproduction, or distribution of Your modifications, or
+   for any such Derivative Works as a whole, provided Your use,
+   reproduction, and distribution of the Work otherwise complies with
+   the conditions stated in this License.
+
+5. Submission of Contributions. Unless You explicitly state otherwise,
+   any Contribution intentionally submitted for inclusion in the Work
+   by You to the Licensor shall be under the terms and conditions of
+   this License, without any additional terms or conditions.
+   Notwithstanding the above, nothing herein shall supersede or modify
+   the terms of any separate license agreement you may have executed
+   with Licensor regarding such Contributions.
+
+6. Trademarks. This License does not grant permission to use the trade
+   names, trademarks, service marks, or product names of the Licensor,
+   except as required for reasonable and customary use in describing the
+   origin of the Work and reproducing the content of the NOTICE file.
+
+7. Disclaimer of Warranty. Unless required by applicable law or
+   agreed to in writing, Licensor provides the Work (and each
+   Contributor provides its Contributions) on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+   implied, including, without limitation, any warranties or conditions
+   of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+   PARTICULAR PURPOSE. You are solely responsible for determining the
+   appropriateness of using or redistributing the Work and assume any
+   risks associated with Your exercise of permissions under this License.
+
+8. Limitation of Liability. In no event and under no legal theory,
+   whether in tort (including negligence), contract, or otherwise,
+   unless required by applicable law (such as deliberate and grossly
+   negligent acts) or agreed to in writing, shall any Contributor be
+   liable to You for damages, including any direct, indirect, special,
+   incidental, or consequential damages of any character arising as a
+   result of this License or out of the use or inability to use the
+   Work (including but not limited to damages for loss of goodwill,
+   work stoppage, computer failure or malfunction, or any and all
+   other commercial damages or losses), even if such Contributor
+   has been advised of the possibility of such damages.
+
+9. Accepting Warranty or Additional Liability. While redistributing
+   the Work or Derivative Works thereof, You may choose to offer,
+   and charge a fee for, acceptance of support, warranty, indemnity,
+   or other liability obligations and/or rights consistent with this
+   License. However, in accepting such obligations, You may act only
+   on Your own behalf and on Your sole responsibility, not on behalf
+   of any other Contributor, and only if You agree to indemnify,
+   defend, and hold each Contributor harmless for any liability
+   incurred by, or claims asserted against, such Contributor by reason
+   of your accepting any such warranty or additional liability.
+
+END OF TERMS AND CONDITIONS

package/README.md

@@ -0,0 +1,184 @@
+# @hazeljs/eval
+
+Evaluation toolkit for HazelJS AI applications: golden datasets, classical IR metrics (precision/recall@k, MRR, NDCG), RAG-style heuristics, agent trajectory scoring, LLM-as-judge helpers, and CI-friendly reporting.
+
+[![npm version](https://img.shields.io/npm/v/@hazeljs/eval.svg)](https://www.npmjs.com/package/@hazeljs/eval)
+[![License: Apache-2.0](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://www.apache.org/licenses/LICENSE-2.0)
+
+## Installation
+
+```bash
+npm install @hazeljs/eval
+```
+
+Peer dependency:
+
+```bash
+npm install @hazeljs/core
+```
+
+## Features
+
+- **Golden datasets** — JSON format with cases (input, optional expected output, tool calls, retrieved IDs)
+- **Retrieval metrics** — `precisionAtK`, `recallAtK`, `meanReciprocalRank`, `ndcgAtK`
+- **RAG helpers** — `evaluateRetrieval`, `answerContextOverlap` (lightweight faithfulness proxy)
+- **Agent trajectories** — `trajectoryScore`, `toolCallAccuracy`
+- **LLM judge** — `parseJudgeScore`, `buildRelevanceJudgePrompt`, `buildFaithfulnessJudgePrompt` (wire to your model)
+- **Runner** — `runGoldenDataset` with configurable concurrency and thresholds
+- **CI** — `reportEvalForCi` with optional non-zero exit via `process.exitCode`
+
+## Quick start
+
+### Load a dataset and run evals
+
+```typescript
+import { loadGoldenDatasetFromJson, runGoldenDataset, reportEvalForCi } from '@hazeljs/eval';
+
+const dataset = loadGoldenDatasetFromJson('./eval/golden.json');
+
+const result = await runGoldenDataset(
+  dataset,
+  async ({ input }) => {
+    // Call your RAG, agent, or HazelAI pipeline here
+    return {
+      output: await myPipeline.answer(input),
+      toolCalls: ['search', 'summarize'],
+      retrievedIds: ['doc-1', 'doc-2'],
+    };
+  },
+  { concurrency: 2, minAverageScore: 0.75 }
+);
+
+reportEvalForCi(result, { exitOnFail: true });
+```
+
+### Retrieval metrics only
+
+```typescript
+import { evaluateRetrieval } from '@hazeljs/eval';
+
+const metrics = evaluateRetrieval({
+  query: 'What is HazelJS?',
+  retrievedIds: ['a', 'b', 'c'],
+  relevantIds: ['a', 'x'],
+  k: 5,
+});
+// metrics.precisionAtK, recallAtK, mrr, ndcgAtK
+```
+
+### Golden dataset JSON shape
+
+```json
+{
+  "name": "support-bot",
+  "version": "1.0.0",
+  "cases": [
+    {
+      "id": "refund-1",
+      "input": "How do I get a refund?",
+      "expectedOutput": "within 30 days",
+      "expectedToolCalls": ["lookup_policy"],
+      "expectedRetrievedIds": ["policy-refunds"]
+    }
+  ]
+}
+```
+
+## Real-life example: regression-test a support FAQ (RAG + CI)
+
+You ship a **customer support** app backed by **HazelAI** and **@hazeljs/rag**. Before each release, you run the same golden questions and fail the pipeline if quality drops. (Install `@hazeljs/ai` and `@hazeljs/rag` alongside `@hazeljs/eval` for this flow.)
+
+**1. Check in a golden dataset** — e.g. `eval/golden.support.json`:
+
+```json
+{
+  "name": "support-faq",
+  "version": "2025.04",
+  "cases": [
+    {
+      "id": "billing-cycle",
+      "input": "When am I charged each month?",
+      "expectedOutput": "first of the month"
+    },
+    {
+      "id": "api-rate-limit",
+      "input": "What happens if I exceed the API rate limit?",
+      "expectedOutput": "429"
+    }
+  ]
+}
+```
+
+**2. Add a small eval script** (run with `npx tsx scripts/run-support-eval.ts` or compile into your CLI):
+
+```typescript
+import * as path from 'path';
+import { HazelAI } from '@hazeljs/ai';
+import { loadGoldenDatasetFromJson, runGoldenDataset, reportEvalForCi } from '@hazeljs/eval';
+
+async function main() {
+  const datasetPath = path.join(process.cwd(), 'eval/golden.support.json');
+  const dataset = loadGoldenDatasetFromJson(datasetPath);
+
+  const ai = HazelAI.create({
+    defaultProvider: 'openai',
+    persistence: {
+      rag: {
+        vectorStore: 'qdrant',
+        connectionString: process.env.QDRANT_URL ?? 'http://127.0.0.1:6333',
+        indexName: 'support-docs',
+      },
+    },
+  });
+
+  const result = await runGoldenDataset(
+    dataset,
+    async ({ input }) => {
+      const rag = await ai.rag.ask(input, { topK: 8 });
+      const retrievedIds = rag.sources.map((s) => s.id);
+      return {
+        output: rag.answer,
+        retrievedIds,
+      };
+    },
+    { concurrency: 1, minAverageScore: 0.7 }
+  );
+
+  reportEvalForCi(result, { exitOnFail: process.env.CI === 'true' });
+}
+
+main().catch((e) => {
+  console.error(e);
+  process.exitCode = 1;
+});
+```
+
+**3. Wire CI** — in GitHub Actions (or similar), after indexes are built and env vars (`OPENAI_API_KEY`, `QDRANT_URL`, …) are set:
+
+```yaml
+- name: Run support golden evals
+  run: npx tsx scripts/run-support-eval.ts
+  env:
+    CI: true
+    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+    QDRANT_URL: ${{ secrets.QDRANT_URL }}
+```
+
+If any case scores below `minAverageScore` or the run errors, `reportEvalForCi` sets `process.exitCode = 1` and the job fails, blocking a bad deploy. Adjust the runner to return `toolCalls` too if you evaluate an **agent** (e.g. `ai.agent('support', input)`) instead of plain RAG.
+
+## Scripts
+
+| Command           | Description                                                                                                                                             |
+| ----------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `npm run build`   | Compile TypeScript to `dist/`                                                                                                                           |
+| `npm test`        | Jest with coverage thresholds (85%+ statements/lines/functions/branches on implementation files; `src/index.ts` is excluded as a pure re-export barrel) |
+| `npm run test:ci` | Jest with coverage reports; thresholds can be relaxed via `--no-coverage-threshold` when needed                                                         |
+
+## License
+
+Apache-2.0
+
+## Links
+
+- [HazelJS documentation](https://hazeljs.ai)
+- [Monorepo source](https://github.com/hazel-js/hazeljs) — `packages/eval`

package/dist/agent-trajectory.d.ts

@@ -0,0 +1,9 @@
+/**
+ * Agent trajectory evaluation — expected tool sequence vs actual.
+ */
+export interface AgentTrajectory {
+    toolCalls: string[];
+}
+export declare function trajectoryScore(expected: string[], actual: string[]): number;
+export declare function toolCallAccuracy(expected: string[], actual: string[]): number;
+//# sourceMappingURL=agent-trajectory.d.ts.map

package/dist/agent-trajectory.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"agent-trajectory.d.ts","sourceRoot":"","sources":["../src/agent-trajectory.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,MAAM,WAAW,eAAe;IAC9B,SAAS,EAAE,MAAM,EAAE,CAAC;CACrB;AAED,wBAAgB,eAAe,CAAC,QAAQ,EAAE,MAAM,EAAE,EAAE,MAAM,EAAE,MAAM,EAAE,GAAG,MAAM,CAY5E;AAED,wBAAgB,gBAAgB,CAAC,QAAQ,EAAE,MAAM,EAAE,EAAE,MAAM,EAAE,MAAM,EAAE,GAAG,MAAM,CAS7E"}

package/dist/agent-trajectory.js

@@ -0,0 +1,33 @@
+"use strict";
+/**
+ * Agent trajectory evaluation — expected tool sequence vs actual.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.trajectoryScore = trajectoryScore;
+exports.toolCallAccuracy = toolCallAccuracy;
+function trajectoryScore(expected, actual) {
+    if (expected.length === 0)
+        return 1;
+    const exp = [...expected];
+    let matched = 0;
+    let j = 0;
+    for (let i = 0; i < actual.length && j < exp.length; i++) {
+        if (actual[i] === exp[j]) {
+            matched++;
+            j++;
+        }
+    }
+    return matched / exp.length;
+}
+function toolCallAccuracy(expected, actual) {
+    const expSet = new Set(expected);
+    if (expSet.size === 0)
+        return 1;
+    const actSet = new Set(actual);
+    let hit = 0;
+    for (const e of expSet) {
+        if (actSet.has(e))
+            hit++;
+    }
+    return hit / expSet.size;
+}

package/dist/ci-reporter.d.ts

@@ -0,0 +1,10 @@
+import type { EvalRunResult } from './types';
+export interface CiReporterOptions {
+    /** When true, exit process with code 1 if eval failed */
+    exitOnFail?: boolean;
+}
+/**
+ * Print a human-readable summary and optionally exit with failure for CI.
+ */
+export declare function reportEvalForCi(result: EvalRunResult, options?: CiReporterOptions): void;
+//# sourceMappingURL=ci-reporter.d.ts.map

package/dist/ci-reporter.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"ci-reporter.d.ts","sourceRoot":"","sources":["../src/ci-reporter.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,SAAS,CAAC;AAE7C,MAAM,WAAW,iBAAiB;IAChC,yDAAyD;IACzD,UAAU,CAAC,EAAE,OAAO,CAAC;CACtB;AAED;;GAEG;AACH,wBAAgB,eAAe,CAAC,MAAM,EAAE,aAAa,EAAE,OAAO,CAAC,EAAE,iBAAiB,GAAG,IAAI,CAgBxF"}

package/dist/ci-reporter.js

@@ -0,0 +1,19 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.reportEvalForCi = reportEvalForCi;
+/**
+ * Print a human-readable summary and optionally exit with failure for CI.
+ */
+function reportEvalForCi(result, options) {
+    // eslint-disable-next-line no-console
+    console.log(`[hazeljs/eval] ${result.datasetName}@${result.datasetVersion} ` +
+        `avg=${result.averageScore.toFixed(3)} passed=${result.passed}`);
+    for (const c of result.caseResults) {
+        // eslint-disable-next-line no-console
+        console.log(`  - ${c.caseId}: score=${c.score.toFixed(3)} passed=${c.passed}` +
+            (c.error ? ` error=${c.error}` : ''));
+    }
+    if (options?.exitOnFail && !result.passed) {
+        process.exitCode = 1;
+    }
+}

package/dist/dataset-loader.d.ts

@@ -0,0 +1,3 @@
+import type { GoldenDataset } from './types';
+export declare function loadGoldenDatasetFromJson(path: string): GoldenDataset;
+//# sourceMappingURL=dataset-loader.d.ts.map

package/dist/dataset-loader.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"dataset-loader.d.ts","sourceRoot":"","sources":["../src/dataset-loader.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,SAAS,CAAC;AAE7C,wBAAgB,yBAAyB,CAAC,IAAI,EAAE,MAAM,GAAG,aAAa,CAOrE"}

package/dist/dataset-loader.js

@@ -0,0 +1,45 @@
+"use strict";
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.loadGoldenDatasetFromJson = loadGoldenDatasetFromJson;
+const fs = __importStar(require("fs"));
+function loadGoldenDatasetFromJson(path) {
+    const raw = fs.readFileSync(path, 'utf8');
+    const data = JSON.parse(raw);
+    if (!data.name || !data.version || !Array.isArray(data.cases)) {
+        throw new Error(`Invalid golden dataset JSON at ${path}`);
+    }
+    return data;
+}

package/dist/golden-runner.d.ts

@@ -0,0 +1,16 @@
+import type { EvalRunOptions, EvalRunResult, GoldenDataset } from './types';
+export type CaseRunner = (input: {
+    id: string;
+    input: string;
+    /** @deprecated use metadata */
+    metadata?: Record<string, unknown>;
+}) => Promise<{
+    output: string;
+    toolCalls?: string[];
+    retrievedIds?: string[];
+}>;
+/**
+ * Run a golden dataset against an async runner (your app, agent, or RAG pipeline).
+ */
+export declare function runGoldenDataset(dataset: GoldenDataset, runner: CaseRunner, options?: EvalRunOptions): Promise<EvalRunResult>;
+//# sourceMappingURL=golden-runner.d.ts.map

package/dist/golden-runner.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"golden-runner.d.ts","sourceRoot":"","sources":["../src/golden-runner.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,aAAa,EAAc,aAAa,EAAE,MAAM,SAAS,CAAC;AAIxF,MAAM,MAAM,UAAU,GAAG,CAAC,KAAK,EAAE;IAC/B,EAAE,EAAE,MAAM,CAAC;IACX,KAAK,EAAE,MAAM,CAAC;IACd,+BAA+B;IAC/B,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACpC,KAAK,OAAO,CAAC;IAAE,MAAM,EAAE,MAAM,CAAC;IAAC,SAAS,CAAC,EAAE,MAAM,EAAE,CAAC;IAAC,YAAY,CAAC,EAAE,MAAM,EAAE,CAAA;CAAE,CAAC,CAAC;AAEjF;;GAEG;AACH,wBAAsB,gBAAgB,CACpC,OAAO,EAAE,aAAa,EACtB,MAAM,EAAE,UAAU,EAClB,OAAO,CAAC,EAAE,cAAc,GACvB,OAAO,CAAC,aAAa,CAAC,CAyFxB"}

package/dist/golden-runner.js

@@ -0,0 +1,82 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.runGoldenDataset = runGoldenDataset;
+const agent_trajectory_1 = require("./agent-trajectory");
+const retrieval_metrics_1 = require("./retrieval-metrics");
+/**
+ * Run a golden dataset against an async runner (your app, agent, or RAG pipeline).
+ */
+async function runGoldenDataset(dataset, runner, options) {
+    const caseResults = [];
+    const concurrency = Math.max(1, options?.concurrency ?? 1);
+    const chunks = [];
+    for (let i = 0; i < dataset.cases.length; i += concurrency) {
+        chunks.push(dataset.cases.slice(i, i + concurrency));
+    }
+    for (const batch of chunks) {
+        const batchOut = await Promise.all(batch.map(async (c) => {
+            try {
+                const result = await runner({
+                    id: c.id,
+                    input: c.input,
+                    metadata: c.metadata,
+                });
+                let score = 0;
+                const details = {};
+                if (c.expectedOutput !== undefined) {
+                    const out = result.output ?? '';
+                    const exp = c.expectedOutput;
+                    const passed = out.includes(exp) || out.trim().toLowerCase() === exp.trim().toLowerCase();
+                    score = passed ? 1 : 0;
+                    details.outputMatch = passed;
+                }
+                if (c.expectedToolCalls && c.expectedToolCalls.length > 0) {
+                    const traj = (0, agent_trajectory_1.trajectoryScore)(c.expectedToolCalls, result.toolCalls ?? []);
+                    score = score === 0 ? traj : (score + traj) / 2;
+                    details.trajectoryScore = traj;
+                }
+                if (c.expectedRetrievedIds && c.expectedRetrievedIds.length > 0) {
+                    const rel = new Set(c.expectedRetrievedIds);
+                    const p = (0, retrieval_metrics_1.precisionAtK)(result.retrievedIds ?? [], rel, 5);
+                    score = score === 0 ? p : (score + p) / 2;
+                    details.precisionAt5 = p;
+                }
+                if (c.expectedOutput === undefined &&
+                    (!c.expectedToolCalls || c.expectedToolCalls.length === 0) &&
+                    (!c.expectedRetrievedIds || c.expectedRetrievedIds.length === 0)) {
+                    score = 1;
+                }
+                const threshold = options?.minAverageScore ?? 0.7;
+                return {
+                    caseId: c.id,
+                    passed: score >= threshold,
+                    score,
+                    details,
+                };
+            }
+            catch (e) {
+                return {
+                    caseId: c.id,
+                    passed: false,
+                    score: 0,
+                    error: e instanceof Error ? e.message : String(e),
+                };
+            }
+        }));
+        caseResults.push(...batchOut);
+    }
+    const averageScore = caseResults.length === 0
+        ? 0
+        : caseResults.reduce((s, r) => s + r.score, 0) / caseResults.length;
+    const minAvg = options?.minAverageScore ?? 0;
+    const passed = averageScore >= minAvg &&
+        caseResults.every((r) => r.passed) &&
+        !caseResults.some((r) => r.error);
+    return {
+        datasetName: dataset.name,
+        datasetVersion: dataset.version,
+        caseResults,
+        averageScore,
+        passed,
+    };
+}

package/dist/index.d.ts

@@ -0,0 +1,12 @@
+/**
+ * @hazeljs/eval — Evaluation toolkit for AI-native HazelJS apps
+ */
+export type { GoldenCase, GoldenDataset, EvalRunOptions, CaseResult, EvalRunResult } from './types';
+export { precisionAtK, recallAtK, meanReciprocalRank, ndcgAtK } from './retrieval-metrics';
+export { evaluateRetrieval, answerContextOverlap, type RagRetrievalEvalInput, type RagRetrievalEvalResult, } from './rag-metrics';
+export { trajectoryScore, toolCallAccuracy, type AgentTrajectory } from './agent-trajectory';
+export { parseJudgeScore, buildRelevanceJudgePrompt, buildFaithfulnessJudgePrompt, type JudgeScore, type LLMJudgeFn, } from './llm-judge';
+export { runGoldenDataset, type CaseRunner } from './golden-runner';
+export { reportEvalForCi, type CiReporterOptions } from './ci-reporter';
+export { loadGoldenDatasetFromJson } from './dataset-loader';
+//# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,YAAY,EAAE,UAAU,EAAE,aAAa,EAAE,cAAc,EAAE,UAAU,EAAE,aAAa,EAAE,MAAM,SAAS,CAAC;AAEpG,OAAO,EAAE,YAAY,EAAE,SAAS,EAAE,kBAAkB,EAAE,OAAO,EAAE,MAAM,qBAAqB,CAAC;AAE3F,OAAO,EACL,iBAAiB,EACjB,oBAAoB,EACpB,KAAK,qBAAqB,EAC1B,KAAK,sBAAsB,GAC5B,MAAM,eAAe,CAAC;AAEvB,OAAO,EAAE,eAAe,EAAE,gBAAgB,EAAE,KAAK,eAAe,EAAE,MAAM,oBAAoB,CAAC;AAE7F,OAAO,EACL,eAAe,EACf,yBAAyB,EACzB,4BAA4B,EAC5B,KAAK,UAAU,EACf,KAAK,UAAU,GAChB,MAAM,aAAa,CAAC;AAErB,OAAO,EAAE,gBAAgB,EAAE,KAAK,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAEpE,OAAO,EAAE,eAAe,EAAE,KAAK,iBAAiB,EAAE,MAAM,eAAe,CAAC;AAExE,OAAO,EAAE,yBAAyB,EAAE,MAAM,kBAAkB,CAAC"}

package/dist/index.js

@@ -0,0 +1,27 @@
+"use strict";
+/**
+ * @hazeljs/eval — Evaluation toolkit for AI-native HazelJS apps
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.loadGoldenDatasetFromJson = exports.reportEvalForCi = exports.runGoldenDataset = exports.buildFaithfulnessJudgePrompt = exports.buildRelevanceJudgePrompt = exports.parseJudgeScore = exports.toolCallAccuracy = exports.trajectoryScore = exports.answerContextOverlap = exports.evaluateRetrieval = exports.ndcgAtK = exports.meanReciprocalRank = exports.recallAtK = exports.precisionAtK = void 0;
+var retrieval_metrics_1 = require("./retrieval-metrics");
+Object.defineProperty(exports, "precisionAtK", { enumerable: true, get: function () { return retrieval_metrics_1.precisionAtK; } });
+Object.defineProperty(exports, "recallAtK", { enumerable: true, get: function () { return retrieval_metrics_1.recallAtK; } });
+Object.defineProperty(exports, "meanReciprocalRank", { enumerable: true, get: function () { return retrieval_metrics_1.meanReciprocalRank; } });
+Object.defineProperty(exports, "ndcgAtK", { enumerable: true, get: function () { return retrieval_metrics_1.ndcgAtK; } });
+var rag_metrics_1 = require("./rag-metrics");
+Object.defineProperty(exports, "evaluateRetrieval", { enumerable: true, get: function () { return rag_metrics_1.evaluateRetrieval; } });
+Object.defineProperty(exports, "answerContextOverlap", { enumerable: true, get: function () { return rag_metrics_1.answerContextOverlap; } });
+var agent_trajectory_1 = require("./agent-trajectory");
+Object.defineProperty(exports, "trajectoryScore", { enumerable: true, get: function () { return agent_trajectory_1.trajectoryScore; } });
+Object.defineProperty(exports, "toolCallAccuracy", { enumerable: true, get: function () { return agent_trajectory_1.toolCallAccuracy; } });
+var llm_judge_1 = require("./llm-judge");
+Object.defineProperty(exports, "parseJudgeScore", { enumerable: true, get: function () { return llm_judge_1.parseJudgeScore; } });
+Object.defineProperty(exports, "buildRelevanceJudgePrompt", { enumerable: true, get: function () { return llm_judge_1.buildRelevanceJudgePrompt; } });
+Object.defineProperty(exports, "buildFaithfulnessJudgePrompt", { enumerable: true, get: function () { return llm_judge_1.buildFaithfulnessJudgePrompt; } });
+var golden_runner_1 = require("./golden-runner");
+Object.defineProperty(exports, "runGoldenDataset", { enumerable: true, get: function () { return golden_runner_1.runGoldenDataset; } });
+var ci_reporter_1 = require("./ci-reporter");
+Object.defineProperty(exports, "reportEvalForCi", { enumerable: true, get: function () { return ci_reporter_1.reportEvalForCi; } });
+var dataset_loader_1 = require("./dataset-loader");
+Object.defineProperty(exports, "loadGoldenDatasetFromJson", { enumerable: true, get: function () { return dataset_loader_1.loadGoldenDatasetFromJson; } });

package/dist/llm-judge.d.ts

@@ -0,0 +1,16 @@
+/**
+ * LLM-as-judge — pluggable interface; wire to @hazeljs/ai or any HTTP API.
+ */
+export type JudgeScore = {
+    score: number;
+    reasoning?: string;
+};
+export type LLMJudgeFn = (prompt: string) => Promise<string>;
+/**
+ * Ask a judge model to score 0–1 given a rubric prompt.
+ * Expects JSON in the response: `{ "score": number, "reasoning"?: string }`
+ */
+export declare function parseJudgeScore(raw: string): Promise<JudgeScore>;
+export declare function buildRelevanceJudgePrompt(query: string, answer: string): string;
+export declare function buildFaithfulnessJudgePrompt(context: string, answer: string): string;
+//# sourceMappingURL=llm-judge.d.ts.map

package/dist/llm-judge.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"llm-judge.d.ts","sourceRoot":"","sources":["../src/llm-judge.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,MAAM,MAAM,UAAU,GAAG;IACvB,KAAK,EAAE,MAAM,CAAC;IACd,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB,CAAC;AAEF,MAAM,MAAM,UAAU,GAAG,CAAC,MAAM,EAAE,MAAM,KAAK,OAAO,CAAC,MAAM,CAAC,CAAC;AAE7D;;;GAGG;AACH,wBAAsB,eAAe,CAAC,GAAG,EAAE,MAAM,GAAG,OAAO,CAAC,UAAU,CAAC,CAQtE;AAED,wBAAgB,yBAAyB,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,GAAG,MAAM,CAO/E;AAED,wBAAgB,4BAA4B,CAAC,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,GAAG,MAAM,CASpF"}

package/dist/llm-judge.js

@@ -0,0 +1,39 @@
+"use strict";
+/**
+ * LLM-as-judge — pluggable interface; wire to @hazeljs/ai or any HTTP API.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.parseJudgeScore = parseJudgeScore;
+exports.buildRelevanceJudgePrompt = buildRelevanceJudgePrompt;
+exports.buildFaithfulnessJudgePrompt = buildFaithfulnessJudgePrompt;
+/**
+ * Ask a judge model to score 0–1 given a rubric prompt.
+ * Expects JSON in the response: `{ "score": number, "reasoning"?: string }`
+ */
+async function parseJudgeScore(raw) {
+    const cleaned = raw
+        .replace(/^```(?:json)?\s*/i, '')
+        .replace(/\s*```$/i, '')
+        .trim();
+    const parsed = JSON.parse(cleaned);
+    const score = typeof parsed.score === 'number' ? Math.max(0, Math.min(1, parsed.score)) : 0;
+    return { score, reasoning: parsed.reasoning };
+}
+function buildRelevanceJudgePrompt(query, answer) {
+    return `Rate how relevant the answer is to the query on a scale of 0 to 1.
+Respond with JSON only: {"score": number between 0 and 1, "reasoning": string}
+
+Query: ${query}
+
+Answer: ${answer}`;
+}
+function buildFaithfulnessJudgePrompt(context, answer) {
+    return `Rate whether the answer is grounded in the given context (0 = hallucinated, 1 = fully supported).
+Respond with JSON only: {"score": number between 0 and 1, "reasoning": string}
+
+Context:
+${context}
+
+Answer:
+${answer}`;
+}

package/dist/rag-metrics.d.ts

@@ -0,0 +1,19 @@
+/**
+ * RAG-oriented evaluation helpers (Ragas-style heuristics without external deps).
+ */
+export interface RagRetrievalEvalInput {
+    query: string;
+    retrievedIds: string[];
+    relevantIds: string[];
+    k?: number;
+}
+export interface RagRetrievalEvalResult {
+    precisionAtK: number;
+    recallAtK: number;
+    mrr: number;
+    ndcgAtK: number;
+}
+export declare function evaluateRetrieval(input: RagRetrievalEvalInput): RagRetrievalEvalResult;
+/** Token overlap between answer and context (cheap faithfulness proxy) */
+export declare function answerContextOverlap(answer: string, context: string): number;
+//# sourceMappingURL=rag-metrics.d.ts.map

package/dist/rag-metrics.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"rag-metrics.d.ts","sourceRoot":"","sources":["../src/rag-metrics.ts"],"names":[],"mappings":"AAAA;;GAEG;AAIH,MAAM,WAAW,qBAAqB;IACpC,KAAK,EAAE,MAAM,CAAC;IACd,YAAY,EAAE,MAAM,EAAE,CAAC;IACvB,WAAW,EAAE,MAAM,EAAE,CAAC;IACtB,CAAC,CAAC,EAAE,MAAM,CAAC;CACZ;AAED,MAAM,WAAW,sBAAsB;IACrC,YAAY,EAAE,MAAM,CAAC;IACrB,SAAS,EAAE,MAAM,CAAC;IAClB,GAAG,EAAE,MAAM,CAAC;IACZ,OAAO,EAAE,MAAM,CAAC;CACjB;AAED,wBAAgB,iBAAiB,CAAC,KAAK,EAAE,qBAAqB,GAAG,sBAAsB,CAStF;AAED,0EAA0E;AAC1E,wBAAgB,oBAAoB,CAAC,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,GAAG,MAAM,CAqB5E"}

package/dist/rag-metrics.js

@@ -0,0 +1,39 @@
+"use strict";
+/**
+ * RAG-oriented evaluation helpers (Ragas-style heuristics without external deps).
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.evaluateRetrieval = evaluateRetrieval;
+exports.answerContextOverlap = answerContextOverlap;
+const retrieval_metrics_1 = require("./retrieval-metrics");
+function evaluateRetrieval(input) {
+    const k = input.k ?? 5;
+    const rel = new Set(input.relevantIds);
+    return {
+        precisionAtK: (0, retrieval_metrics_1.precisionAtK)(input.retrievedIds, rel, k),
+        recallAtK: (0, retrieval_metrics_1.recallAtK)(input.retrievedIds, rel, k),
+        mrr: (0, retrieval_metrics_1.meanReciprocalRank)(input.retrievedIds, rel),
+        ndcgAtK: (0, retrieval_metrics_1.ndcgAtK)(input.retrievedIds, rel, k),
+    };
+}
+/** Token overlap between answer and context (cheap faithfulness proxy) */
+function answerContextOverlap(answer, context) {
+    const a = new Set(answer
+        .toLowerCase()
+        .replace(/[^\p{L}\p{N}\s]/gu, ' ')
+        .split(/\s+/)
+        .filter((w) => w.length > 2));
+    const c = new Set(context
+        .toLowerCase()
+        .replace(/[^\p{L}\p{N}\s]/gu, ' ')
+        .split(/\s+/)
+        .filter((w) => w.length > 2));
+    if (a.size === 0)
+        return 0;
+    let inter = 0;
+    for (const w of a) {
+        if (c.has(w))
+            inter++;
+    }
+    return inter / a.size;
+}

package/dist/retrieval-metrics.d.ts

@@ -0,0 +1,12 @@
+/**
+ * Classic IR metrics for retrieval evaluation (no LLM required).
+ */
+/** Precision@k: fraction of top-k that are relevant */
+export declare function precisionAtK(retrievedIds: string[], relevantIds: Set<string>, k: number): number;
+/** Recall@k: fraction of all relevant docs found in top-k */
+export declare function recallAtK(retrievedIds: string[], relevantIds: Set<string>, k: number): number;
+/** Mean Reciprocal Rank — first relevant position */
+export declare function meanReciprocalRank(retrievedIds: string[], relevantIds: Set<string>): number;
+/** Normalized Discounted Cumulative Gain (binary relevance) */
+export declare function ndcgAtK(retrievedIds: string[], relevantIds: Set<string>, k: number): number;
+//# sourceMappingURL=retrieval-metrics.d.ts.map

package/dist/retrieval-metrics.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"retrieval-metrics.d.ts","sourceRoot":"","sources":["../src/retrieval-metrics.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,uDAAuD;AACvD,wBAAgB,YAAY,CAAC,YAAY,EAAE,MAAM,EAAE,EAAE,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,EAAE,CAAC,EAAE,MAAM,GAAG,MAAM,CAKhG;AAED,6DAA6D;AAC7D,wBAAgB,SAAS,CAAC,YAAY,EAAE,MAAM,EAAE,EAAE,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,EAAE,CAAC,EAAE,MAAM,GAAG,MAAM,CAQ7F;AAED,qDAAqD;AACrD,wBAAgB,kBAAkB,CAAC,YAAY,EAAE,MAAM,EAAE,EAAE,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,GAAG,MAAM,CAO3F;AAED,+DAA+D;AAC/D,wBAAgB,OAAO,CAAC,YAAY,EAAE,MAAM,EAAE,EAAE,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,EAAE,CAAC,EAAE,MAAM,GAAG,MAAM,CAY3F"}

package/dist/retrieval-metrics.js

@@ -0,0 +1,52 @@
+"use strict";
+/**
+ * Classic IR metrics for retrieval evaluation (no LLM required).
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.precisionAtK = precisionAtK;
+exports.recallAtK = recallAtK;
+exports.meanReciprocalRank = meanReciprocalRank;
+exports.ndcgAtK = ndcgAtK;
+/** Precision@k: fraction of top-k that are relevant */
+function precisionAtK(retrievedIds, relevantIds, k) {
+    if (k <= 0)
+        return 0;
+    const top = retrievedIds.slice(0, k);
+    const hits = top.filter((id) => relevantIds.has(id)).length;
+    return hits / Math.min(k, top.length || 1);
+}
+/** Recall@k: fraction of all relevant docs found in top-k */
+function recallAtK(retrievedIds, relevantIds, k) {
+    if (relevantIds.size === 0)
+        return 1;
+    const top = new Set(retrievedIds.slice(0, k));
+    let hit = 0;
+    for (const id of relevantIds) {
+        if (top.has(id))
+            hit++;
+    }
+    return hit / relevantIds.size;
+}
+/** Mean Reciprocal Rank — first relevant position */
+function meanReciprocalRank(retrievedIds, relevantIds) {
+    for (let i = 0; i < retrievedIds.length; i++) {
+        if (relevantIds.has(retrievedIds[i])) {
+            return 1 / (i + 1);
+        }
+    }
+    return 0;
+}
+/** Normalized Discounted Cumulative Gain (binary relevance) */
+function ndcgAtK(retrievedIds, relevantIds, k) {
+    const top = retrievedIds.slice(0, k);
+    const dcg = top.reduce((acc, id, idx) => {
+        const rel = relevantIds.has(id) ? 1 : 0;
+        return acc + (rel === 0 ? 0 : rel / Math.log2(idx + 2));
+    }, 0);
+    const idealHits = Math.min(relevantIds.size, k);
+    let idcg = 0;
+    for (let i = 0; i < idealHits; i++) {
+        idcg += 1 / Math.log2(i + 2);
+    }
+    return idcg === 0 ? 0 : dcg / idcg;
+}

package/dist/types.d.ts

@@ -0,0 +1,40 @@
+/**
+ * Core types for @hazeljs/eval
+ */
+export interface GoldenCase {
+    id: string;
+    input: string;
+    /** Expected final answer (substring or exact match depending on matcher) */
+    expectedOutput?: string;
+    /** Expected tool / function names in order (agent eval) */
+    expectedToolCalls?: string[];
+    /** Relevant document IDs or content snippets retrieval should surface */
+    expectedRetrievedIds?: string[];
+    metadata?: Record<string, unknown>;
+}
+export interface GoldenDataset {
+    name: string;
+    version: string;
+    cases: GoldenCase[];
+}
+export interface EvalRunOptions {
+    /** Max parallel cases (default 1) */
+    concurrency?: number;
+    /** Fail CI when average score below threshold */
+    minAverageScore?: number;
+}
+export interface CaseResult {
+    caseId: string;
+    passed: boolean;
+    score: number;
+    details?: Record<string, unknown>;
+    error?: string;
+}
+export interface EvalRunResult {
+    datasetName: string;
+    datasetVersion: string;
+    caseResults: CaseResult[];
+    averageScore: number;
+    passed: boolean;
+}
+//# sourceMappingURL=types.d.ts.map

package/dist/types.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,MAAM,WAAW,UAAU;IACzB,EAAE,EAAE,MAAM,CAAC;IACX,KAAK,EAAE,MAAM,CAAC;IACd,4EAA4E;IAC5E,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,2DAA2D;IAC3D,iBAAiB,CAAC,EAAE,MAAM,EAAE,CAAC;IAC7B,yEAAyE;IACzE,oBAAoB,CAAC,EAAE,MAAM,EAAE,CAAC;IAChC,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACpC;AAED,MAAM,WAAW,aAAa;IAC5B,IAAI,EAAE,MAAM,CAAC;IACb,OAAO,EAAE,MAAM,CAAC;IAChB,KAAK,EAAE,UAAU,EAAE,CAAC;CACrB;AAED,MAAM,WAAW,cAAc;IAC7B,qCAAqC;IACrC,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,iDAAiD;IACjD,eAAe,CAAC,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,UAAU;IACzB,MAAM,EAAE,MAAM,CAAC;IACf,MAAM,EAAE,OAAO,CAAC;IAChB,KAAK,EAAE,MAAM,CAAC;IACd,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAClC,KAAK,CAAC,EAAE,MAAM,CAAC;CAChB;AAED,MAAM,WAAW,aAAa;IAC5B,WAAW,EAAE,MAAM,CAAC;IACpB,cAAc,EAAE,MAAM,CAAC;IACvB,WAAW,EAAE,UAAU,EAAE,CAAC;IAC1B,YAAY,EAAE,MAAM,CAAC;IACrB,MAAM,EAAE,OAAO,CAAC;CACjB"}

package/dist/types.js

@@ -0,0 +1,5 @@
+"use strict";
+/**
+ * Core types for @hazeljs/eval
+ */
+Object.defineProperty(exports, "__esModule", { value: true });

package/package.json

@@ -0,0 +1,57 @@
+{
+  "name": "@hazeljs/eval",
+  "version": "0.8.1",
+  "description": "Evaluation toolkit for HazelJS AI apps — golden datasets, RAG metrics, agent trajectories, LLM-as-judge, CI reports",
+  "main": "dist/index.js",
+  "types": "dist/index.d.ts",
+  "files": [
+    "dist"
+  ],
+  "scripts": {
+    "build": "tsc",
+    "test": "jest --coverage --passWithNoTests",
+    "test:ci": "jest --coverage --coverageReporters=text --coverageReporters=lcov --no-coverage-threshold",
+    "lint": "eslint \"src/**/*.ts\" --ignore-pattern \"**/*.d.ts\"",
+    "lint:fix": "eslint \"src/**/*.ts\" --ignore-pattern \"**/*.d.ts\" --fix",
+    "clean": "rm -rf dist && find src -type f \\( -name '*.js' -o -name '*.d.ts' -o -name '*.d.ts.map' \\) -delete 2>/dev/null || true"
+  },
+  "dependencies": {
+    "zod": "^3.25.76"
+  },
+  "peerDependencies": {
+    "@hazeljs/core": ">=0.2.0-beta.0"
+  },
+  "devDependencies": {
+    "@hazeljs/core": "^0.8.1",
+    "@types/jest": "^29.5.14",
+    "@types/node": "^20.19.39",
+    "@typescript-eslint/eslint-plugin": "^8.58.0",
+    "@typescript-eslint/parser": "^8.58.0",
+    "eslint": "^8.57.1",
+    "eslint-config-prettier": "^9.1.2",
+    "eslint-plugin-prettier": "^5.5.5",
+    "jest": "^29.7.0",
+    "prettier": "^3.8.1",
+    "ts-jest": "^29.4.9",
+    "typescript": "^5.9.3"
+  },
+  "publishConfig": {
+    "access": "public"
+  },
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/hazel-js/hazeljs.git",
+    "directory": "packages/eval"
+  },
+  "keywords": [
+    "hazeljs",
+    "eval",
+    "llm",
+    "rag",
+    "agent",
+    "metrics"
+  ],
+  "author": "Muhammad Arslan <muhammad.arslan@hazeljs.ai>",
+  "license": "Apache-2.0",
+  "gitHead": "8b7685d1250c4622f25d83992f58e13a59bb3dba"
+}