hippo-memory 1.72.0 → 2.1.0

package/dist/eval/eval-suite.js

@@ -1,6 +1,8 @@
-/** Self-contained feature eval suite: a synthetic corpus with known ground truth, per-feature metrics and regression detection against a saved baseline.
+/**
+ * Self-contained feature eval suite: a synthetic corpus with known ground truth, per-feature metrics and regression
+ * detection against a saved baseline.
  *  Zero API calls (no LLM judge, no embeddings), deterministic (fixed timestamps, content, IDs) and fast. */
-import { createMemory, Layer } from '../core/memory.js';
+import { createMemory, DAG_LEVEL, Layer } from '../core/memory.js';
 import { search } from '../search/bm25-search.js';
 import { multihopSearch } from '../search/multihop.js';
 import { mrr, recallAtK, ndcgAtK } from './eval.js';
@@ -16,9 +18,9 @@ function mem(id, content, opts = {}) {
         layer: opts.layer ?? Layer.Episodic,
         tags: opts.tags ?? [],
         baseHalfLifeDays: 30,
-        extracted_from: opts.extracted_from,
-        dag_level: opts.dag_level,
-        dag_parent_id: opts.dag_parent_id,
+        extractedFrom: opts.extracted_from,
+        dagLevel: opts.dag_level,
+        dagParentId: opts.dag_parent_id,
     });
     entry.id = id;
     if (opts.created)
@@ -26,169 +28,372 @@ function mem(id, content, opts = {}) {
     return entry;
 }
 /** Direct recall: basic keyword matching. */
+function directRecallEntries() {
+    return [
+        mem('dr-1', 'The PostgreSQL database migration failed because the users table had a NOT NULL constraint on the email column', {
+            created: dateOffset(1),
+            tags: ['topic:database'],
+        }),
+        mem('dr-2', 'React component rendering performance improved by 40% after memoizing the expensive computation in useMemo', {
+            created: dateOffset(2),
+            tags: ['topic:frontend'],
+        }),
+        mem('dr-3', 'The API rate limiter should use a sliding window algorithm instead of fixed window to prevent burst traffic', {
+            created: dateOffset(3),
+            tags: ['topic:api'],
+        }),
+        mem('dr-4', 'Docker container memory limits need to be set to 512MB for the worker service to prevent OOM kills', {
+            created: dateOffset(4), tags: ['topic:devops']
+        }),
+        mem('dr-5', 'The JWT token expiration was set to 24 hours but should be reduced to 1 hour for security compliance', {
+            created: dateOffset(5),
+            tags: ['topic:security'],
+        }),
+        mem('dr-6', 'Webpack bundle size grew to 2.3MB because lodash was imported as a whole instead of cherry-picking', {
+            created: dateOffset(6), tags: ['topic:frontend']
+        }),
+        mem('dr-7', 'The Redis cache TTL for user sessions should match the JWT expiration to prevent stale sessions', {
+            created: dateOffset(7), tags: ['topic:caching']
+        }),
+        mem('dr-8', 'GraphQL resolver for nested comments has N+1 query problem solved by DataLoader batching', {
+            created: dateOffset(8),
+            tags: ['topic:api'],
+        }),
+    ];
+}
+function directRecallCase(n, query, description) {
+    return { id: `dr-q${n}`, category: 'direct-recall', query, expectedIds: [`dr-${n}`], description };
+}
+function directRecallCases() {
+    return [
+        directRecallCase(1, 'PostgreSQL migration NOT NULL constraint', 'exact keyword match on DB migration'),
+        directRecallCase(2, 'React useMemo performance memoizing', 'React performance optimization'),
+        directRecallCase(3, 'sliding window rate limiter API', 'API rate limiting approach'),
+        directRecallCase(4, 'Docker OOM memory limits worker', 'Docker memory config'),
+        directRecallCase(5, 'JWT token expiration security', 'JWT security setting'),
+        directRecallCase(6, 'webpack bundle size lodash', 'bundle size issue'),
+        directRecallCase(7, 'Redis cache TTL session expiration', 'cache TTL config'),
+        directRecallCase(8, 'GraphQL N+1 DataLoader batching', 'GraphQL N+1 fix'),
+    ];
+}
 function addDirectRecall(entries, cases) {
-    entries.push(mem('dr-1', 'The PostgreSQL database migration failed because the users table had a NOT NULL constraint on the email column', {
-        created: dateOffset(1), tags: ['topic:database']
-    }), mem('dr-2', 'React component rendering performance improved by 40% after memoizing the expensive computation in useMemo', {
-        created: dateOffset(2), tags: ['topic:frontend']
-    }), mem('dr-3', 'The API rate limiter should use a sliding window algorithm instead of fixed window to prevent burst traffic', {
-        created: dateOffset(3), tags: ['topic:api']
-    }), mem('dr-4', 'Docker container memory limits need to be set to 512MB for the worker service to prevent OOM kills', {
-        created: dateOffset(4), tags: ['topic:devops']
-    }), mem('dr-5', 'The JWT token expiration was set to 24 hours but should be reduced to 1 hour for security compliance', {
-        created: dateOffset(5), tags: ['topic:security']
-    }), mem('dr-6', 'Webpack bundle size grew to 2.3MB because lodash was imported as a whole instead of cherry-picking', {
-        created: dateOffset(6), tags: ['topic:frontend']
-    }), mem('dr-7', 'The Redis cache TTL for user sessions should match the JWT expiration to prevent stale sessions', {
-        created: dateOffset(7), tags: ['topic:caching']
-    }), mem('dr-8', 'GraphQL resolver for nested comments has N+1 query problem solved by DataLoader batching', { created: dateOffset(8), tags: ['topic:api'] }));
-    cases.push({
-        id: 'dr-q1', category: 'direct-recall', query: 'PostgreSQL migration NOT NULL constraint', expectedIds: ['dr-1'],
-        description: 'exact keyword match on DB migration'
-    }, {
-        id: 'dr-q2', category: 'direct-recall', query: 'React useMemo performance memoizing', expectedIds: ['dr-2'],
-        description: 'React performance optimization'
-    }, { id: 'dr-q3', category: 'direct-recall', query: 'sliding window rate limiter API', expectedIds: ['dr-3'], description: 'API rate limiting approach' }, { id: 'dr-q4', category: 'direct-recall', query: 'Docker OOM memory limits worker', expectedIds: ['dr-4'], description: 'Docker memory config' }, { id: 'dr-q5', category: 'direct-recall', query: 'JWT token expiration security', expectedIds: ['dr-5'], description: 'JWT security setting' }, { id: 'dr-q6', category: 'direct-recall', query: 'webpack bundle size lodash', expectedIds: ['dr-6'], description: 'bundle size issue' }, { id: 'dr-q7', category: 'direct-recall', query: 'Redis cache TTL session expiration', expectedIds: ['dr-7'], description: 'cache TTL config' }, { id: 'dr-q8', category: 'direct-recall', query: 'GraphQL N+1 DataLoader batching', expectedIds: ['dr-8'], description: 'GraphQL N+1 fix' });
+    entries.push(...directRecallEntries());
+    cases.push(...directRecallCases());
 }
 /** Extraction preference: extracted facts should rank above raw source. */
+function infrastructureMeetingEntries() {
+    return [
+        mem('ep-src-1', 'speaker:Alice: So we had this big meeting yesterday about the deployment pipeline and Bob mentioned that the ' +
+            'staging environment is ' +
+            'using Kubernetes 1.28 and we should upgrade to 1.30 before the end of Q2 because of the security patches', {
+            created: dateOffset(10),
+            tags: ['speaker:Alice', 'topic:infrastructure', 'session:meeting-1'],
+        }),
+        mem('ep-ext-1', 'The staging environment runs Kubernetes 1.28 and needs to be upgraded to 1.30 before end of Q2 for security ' +
+            'patches', {
+            created: dateOffset(10),
+            layer: Layer.Semantic,
+            tags: ['speaker:Alice', 'topic:infrastructure', 'extracted'],
+            extracted_from: 'ep-src-1',
+        }),
+    ];
+}
+function monitoringMeetingEntries() {
+    return [
+        mem('ep-src-2', 'speaker:Bob: Yeah and the thing about the monitoring is that we switched from Datadog to Grafana last month ' +
+            'and the alerting rules ' +
+            'still need to be migrated, Carol was supposed to handle that but she has been busy with the frontend rewrite', {
+            created: dateOffset(11),
+            tags: ['speaker:Bob', 'topic:monitoring', 'session:meeting-1'],
+        }),
+        mem('ep-ext-2', 'The team switched from Datadog to Grafana last month but alerting rules have not been migrated yet. Carol is ' +
+            'responsible but blocked ' +
+            'by the frontend rewrite.', {
+            created: dateOffset(11),
+            layer: Layer.Semantic,
+            tags: ['speaker:Bob', 'topic:monitoring', 'extracted'],
+            extracted_from: 'ep-src-2',
+        }),
+    ];
+}
+function authMeetingEntries() {
+    return [
+        mem('ep-src-3', 'speaker:Carol: The login page redesign is almost done, I just need to wire up the OAuth2 PKCE flow with the ' +
+            'new identity provider and ' +
+            'write the integration tests', {
+            created: dateOffset(12),
+            tags: ['speaker:Carol', 'topic:auth', 'session:meeting-2'],
+        }),
+        mem('ep-ext-3', 'Carol is nearly done with the login page redesign. Remaining work: wire up OAuth2 PKCE flow with the new ' +
+            'identity provider and write ' +
+            'integration tests.', {
+            created: dateOffset(12),
+            layer: Layer.Semantic,
+            tags: ['speaker:Carol', 'topic:auth', 'extracted'],
+            extracted_from: 'ep-src-3',
+        }),
+    ];
+}
+function extractionPreferenceEntries() {
+    return [...infrastructureMeetingEntries(), ...monitoringMeetingEntries(), ...authMeetingEntries()];
+}
+function extractionPreferenceCases() {
+    return [
+        {
+            id: 'ep-q1',
+            category: 'extraction-preference',
+            query: 'Kubernetes upgrade staging environment',
+            expectedIds: ['ep-ext-1'],
+            description: 'extracted fact about K8s upgrade should rank above raw utterance',
+        },
+        {
+            id: 'ep-q2',
+            category: 'extraction-preference',
+            query: 'Datadog Grafana alerting migration',
+            expectedIds: ['ep-ext-2'],
+            description: 'extracted fact about monitoring switch',
+        },
+        {
+            id: 'ep-q3',
+            category: 'extraction-preference',
+            query: 'OAuth2 PKCE login page redesign',
+            expectedIds: ['ep-ext-3'],
+            description: 'extracted fact about auth work',
+        },
+    ];
+}
 function addExtractionPreference(entries, cases) {
-    entries.push(mem('ep-src-1', 'speaker:Alice: So we had this big meeting yesterday about the deployment pipeline and Bob mentioned that the staging environment is ' +
-        'using Kubernetes 1.28 and we should upgrade to 1.30 before the end of Q2 because of the security patches', {
-        created: dateOffset(10), tags: ['speaker:Alice', 'topic:infrastructure', 'session:meeting-1'],
-    }), mem('ep-ext-1', 'The staging environment runs Kubernetes 1.28 and needs to be upgraded to 1.30 before end of Q2 for security patches', {
-        created: dateOffset(10), layer: Layer.Semantic, tags: ['speaker:Alice', 'topic:infrastructure', 'extracted'],
-        extracted_from: 'ep-src-1',
-    }), mem('ep-src-2', 'speaker:Bob: Yeah and the thing about the monitoring is that we switched from Datadog to Grafana last month and the alerting rules ' +
-        'still need to be migrated, Carol was supposed to handle that but she has been busy with the frontend rewrite', {
-        created: dateOffset(11), tags: ['speaker:Bob', 'topic:monitoring', 'session:meeting-1'],
-    }), mem('ep-ext-2', 'The team switched from Datadog to Grafana last month but alerting rules have not been migrated yet. Carol is responsible but blocked ' +
-        'by the frontend rewrite.', {
-        created: dateOffset(11), layer: Layer.Semantic, tags: ['speaker:Bob', 'topic:monitoring', 'extracted'],
-        extracted_from: 'ep-src-2',
-    }), mem('ep-src-3', 'speaker:Carol: The login page redesign is almost done, I just need to wire up the OAuth2 PKCE flow with the new identity provider and ' +
-        'write the integration tests', {
-        created: dateOffset(12), tags: ['speaker:Carol', 'topic:auth', 'session:meeting-2'],
-    }), mem('ep-ext-3', 'Carol is nearly done with the login page redesign. Remaining work: wire up OAuth2 PKCE flow with the new identity provider and write ' +
-        'integration tests.', {
-        created: dateOffset(12), layer: Layer.Semantic, tags: ['speaker:Carol', 'topic:auth', 'extracted'],
-        extracted_from: 'ep-src-3',
-    }));
-    cases.push({
-        id: 'ep-q1', category: 'extraction-preference', query: 'Kubernetes upgrade staging environment', expectedIds: ['ep-ext-1'],
-        description: 'extracted fact about K8s upgrade should rank above raw utterance'
-    }, {
-        id: 'ep-q2', category: 'extraction-preference', query: 'Datadog Grafana alerting migration', expectedIds: ['ep-ext-2'],
-        description: 'extracted fact about monitoring switch'
-    }, {
-        id: 'ep-q3', category: 'extraction-preference', query: 'OAuth2 PKCE login page redesign', expectedIds: ['ep-ext-3'],
-        description: 'extracted fact about auth work'
-    });
+    entries.push(...extractionPreferenceEntries());
+    cases.push(...extractionPreferenceCases());
 }
 /** Dag drill-down: summary nodes should surface children. */
-function addDagDrilldown(entries, cases) {
-    entries.push(mem('dag-child-1', 'The API response time for /users endpoint degraded from 50ms to 300ms after adding the permissions check', {
-        created: dateOffset(15), layer: Layer.Semantic, tags: ['topic:api-performance', 'extracted'],
-        dag_level: 1,
-    }), mem('dag-child-2', 'The /orders endpoint latency spiked to 500ms because of a missing index on the orders.created_at column', {
-        created: dateOffset(15), layer: Layer.Semantic, tags: ['topic:api-performance', 'extracted'],
-        dag_level: 1,
-    }), mem('dag-child-3', 'Batch API endpoint /reports/generate takes 8 seconds because it runs synchronously instead of using a job queue', {
-        created: dateOffset(15), layer: Layer.Semantic, tags: ['topic:api-performance', 'extracted'],
-        dag_level: 1,
-    }), mem('dag-summary-1', 'API performance issues: /users degraded to 300ms (permissions check), /orders spiked to 500ms (missing index), /reports takes 8s ' +
+function dagChildEntries() {
+    return [
+        mem('dag-child-1', 'The API response time for /users endpoint degraded from 50ms to 300ms after adding the permissions check', {
+            created: dateOffset(15),
+            layer: Layer.Semantic,
+            tags: ['topic:api-performance', 'extracted'],
+            dag_level: DAG_LEVEL.extracted,
+        }),
+        mem('dag-child-2', 'The /orders endpoint latency spiked to 500ms because of a missing index on the orders.created_at column', {
+            created: dateOffset(15),
+            layer: Layer.Semantic,
+            tags: ['topic:api-performance', 'extracted'],
+            dag_level: DAG_LEVEL.extracted,
+        }),
+        mem('dag-child-3', 'Batch API endpoint /reports/generate takes 8 seconds because it runs synchronously instead of using a job queue', {
+            created: dateOffset(15),
+            layer: Layer.Semantic,
+            tags: ['topic:api-performance', 'extracted'],
+            dag_level: DAG_LEVEL.extracted,
+        }),
+    ];
+}
+function dagSummaryEntry() {
+    return mem('dag-summary-1', 'API performance issues: /users degraded to 300ms (permissions check), /orders spiked to 500ms (missing ' +
+        'index), ' +
+        '/reports takes 8s ' +
         '(needs job queue)', {
-        created: dateOffset(16), layer: Layer.Semantic, tags: ['topic:api-performance'],
-        dag_level: 2, dag_parent_id: undefined,
-    }));
-    // Link children to parent
-    for (const child of [entries.find(e => e.id === 'dag-child-1'), entries.find(e => e.id === 'dag-child-2'), entries.find(e => e.id === 'dag-child-3')]) {
-        child.dag_parent_id = 'dag-summary-1';
-    }
-    cases.push({
-        id: 'dag-q1', category: 'dag-drilldown', query: 'API performance problems latency',
-        expectedIds: ['dag-child-1', 'dag-child-2', 'dag-child-3', 'dag-summary-1'], description: 'summary should drill down to all children'
-    }, {
-        id: 'dag-q2', category: 'dag-drilldown', query: 'endpoint response time degradation',
-        expectedIds: ['dag-child-1', 'dag-child-2', 'dag-summary-1'], description: 'query matching summary should surface relevant children'
+        created: dateOffset(16),
+        layer: Layer.Semantic,
+        tags: ['topic:api-performance'],
+        dag_level: DAG_LEVEL.topic,
+        dag_parent_id: undefined,
     });
 }
+function dagEntries() {
+    const children = dagChildEntries();
+    const summary = dagSummaryEntry();
+    for (const child of children)
+        child.dag_parent_id = summary.id;
+    return [...children, summary];
+}
+function dagCases() {
+    return [
+        {
+            id: 'dag-q1',
+            category: 'dag-drilldown',
+            query: 'API performance problems latency',
+            expectedIds: ['dag-child-1', 'dag-child-2', 'dag-child-3', 'dag-summary-1'],
+            description: 'summary should drill down to all children',
+        },
+        {
+            id: 'dag-q2',
+            category: 'dag-drilldown',
+            query: 'endpoint response time degradation',
+            expectedIds: ['dag-child-1', 'dag-child-2', 'dag-summary-1'],
+            description: 'query matching summary should surface relevant children',
+        },
+    ];
+}
+function addDagDrilldown(entries, cases) {
+    entries.push(...dagEntries());
+    cases.push(...dagCases());
+}
 /** Temporal: recency/oldest cues should affect ranking. */
+function temporalEntries() {
+    return [
+        mem('tmp-1', 'The team decided to use TypeScript for the new service', {
+            created: dateOffset(-30),
+            tags: ['topic:architecture'],
+        }),
+        mem('tmp-2', 'The team evaluated Rust as an alternative language for the service', {
+            created: dateOffset(-20),
+            tags: ['topic:architecture'],
+        }),
+        mem('tmp-3', 'The team added Go as a candidate language for the service rewrite', {
+            created: dateOffset(-10),
+            tags: ['topic:architecture'],
+        }),
+        mem('tmp-4', 'The team finalized the language choice as Go for the service rewrite', {
+            created: dateOffset(-1),
+            tags: ['topic:architecture'],
+        }),
+    ];
+}
+function temporalCases() {
+    return [
+        {
+            id: 'tmp-q1',
+            category: 'temporal',
+            query: 'what did the team recently decide about the service language',
+            expectedIds: ['tmp-4'],
+            description: 'recent cue should boost newest entry',
+        },
+        {
+            id: 'tmp-q2',
+            category: 'temporal',
+            query: 'what was the first language choice for the service',
+            expectedIds: ['tmp-1'],
+            description: 'oldest cue should boost earliest entry',
+        },
+        {
+            id: 'tmp-q3',
+            category: 'temporal',
+            query: 'latest update on the service rewrite language',
+            expectedIds: ['tmp-4'],
+            description: 'latest should boost most recent',
+        },
+        {
+            id: 'tmp-q4',
+            category: 'temporal',
+            query: 'original architecture decision for the service',
+            expectedIds: ['tmp-1'],
+            description: 'original should boost earliest',
+        },
+    ];
+}
 function addTemporal(entries, cases) {
-    entries.push(mem('tmp-1', 'The team decided to use TypeScript for the new service', { created: dateOffset(-30), tags: ['topic:architecture'] }), mem('tmp-2', 'The team evaluated Rust as an alternative language for the service', { created: dateOffset(-20), tags: ['topic:architecture'] }), mem('tmp-3', 'The team added Go as a candidate language for the service rewrite', { created: dateOffset(-10), tags: ['topic:architecture'] }), mem('tmp-4', 'The team finalized the language choice as Go for the service rewrite', { created: dateOffset(-1), tags: ['topic:architecture'] }));
-    cases.push({
-        id: 'tmp-q1', category: 'temporal', query: 'what did the team recently decide about the service language', expectedIds: ['tmp-4'],
-        description: 'recent cue should boost newest entry'
-    }, {
-        id: 'tmp-q2', category: 'temporal', query: 'what was the first language choice for the service', expectedIds: ['tmp-1'],
-        description: 'oldest cue should boost earliest entry'
-    }, {
-        id: 'tmp-q3', category: 'temporal', query: 'latest update on the service rewrite language', expectedIds: ['tmp-4'],
-        description: 'latest should boost most recent'
-    }, {
-        id: 'tmp-q4', category: 'temporal', query: 'original architecture decision for the service', expectedIds: ['tmp-1'],
-        description: 'original should boost earliest'
-    });
+    entries.push(...temporalEntries());
+    cases.push(...temporalCases());
 }
-/** Noise resistance: relevant memories found despite noise. */
-function addNoiseResistance(entries, cases) {
-    // Add 30 noise entries
-    const noiseTopics = [
-        'breakfast meeting catering ordered sandwiches', 'office temperature thermostat adjusted',
-        'printer paper refill third floor supply', 'parking lot gate code changed 4521',
-        'fire drill scheduled next Thursday morning', 'coffee machine broken maintenance called',
-        'desk booking system new policy hybrid', 'meeting room projector HDMI adapter missing',
-        'birthday celebration for Dave next Friday', 'recycling bins moved to kitchen area',
-        'badge access updated for new hires', 'elevator maintenance scheduled weekend',
-        'lunch order from Italian place confirmed', 'desk plants watering schedule posted',
-        'air conditioning unit serviced last week', 'office carpet cleaning Friday evening',
-        'new microwave installed in kitchen', 'visitor parking available spots three',
-        'holiday schedule posted on intranet', 'team photo session Thursday afternoon',
-        'chair ergonomics assessment signup sheet', 'standing desk adjustment instructions',
-        'kitchen fridge cleanup policy reminder', 'window blinds replaced on south side',
-        'bicycle rack installed in parking garage', 'team lunch budget increased quarterly',
-        'noise canceling headphones approved expense', 'monitor arm request form updated',
-        'desk drawer key replacement procedure', 'building security hours extended',
+// Noise resistance: relevant memories must be found despite these.
+const NOISE_TOPICS = [
+    'breakfast meeting catering ordered sandwiches', 'office temperature thermostat adjusted',
+    'printer paper refill third floor supply', 'parking lot gate code changed 4521',
+    'fire drill scheduled next Thursday morning', 'coffee machine broken maintenance called',
+    'desk booking system new policy hybrid', 'meeting room projector HDMI adapter missing',
+    'birthday celebration for Dave next Friday', 'recycling bins moved to kitchen area',
+    'badge access updated for new hires', 'elevator maintenance scheduled weekend',
+    'lunch order from Italian place confirmed', 'desk plants watering schedule posted',
+    'air conditioning unit serviced last week', 'office carpet cleaning Friday evening',
+    'new microwave installed in kitchen', 'visitor parking available spots three',
+    'holiday schedule posted on intranet', 'team photo session Thursday afternoon',
+    'chair ergonomics assessment signup sheet', 'standing desk adjustment instructions',
+    'kitchen fridge cleanup policy reminder', 'window blinds replaced on south side',
+    'bicycle rack installed in parking garage', 'team lunch budget increased quarterly',
+    'noise canceling headphones approved expense', 'monitor arm request form updated',
+    'desk drawer key replacement procedure', 'building security hours extended',
+];
+function noiseEntries() {
+    return NOISE_TOPICS.map((topic, i) => mem(`noise-${i}`, topic, { created: dateOffset(i), tags: ['topic:office'] }));
+}
+function noiseTargetEntries() {
+    return [
+        mem('nr-1', 'The database connection pool was exhausted because max_connections was set to 10 but the application had 25 ' +
+            'concurrent requests', {
+            created: dateOffset(20),
+            tags: ['topic:database', 'topic:performance'],
+        }),
+        mem('nr-2', 'The S3 bucket policy was misconfigured allowing public read access to customer data uploads', {
+            created: dateOffset(21), tags: ['topic:security', 'topic:s3'],
+        }),
+    ];
+}
+function noiseCases() {
+    return [
+        {
+            id: 'nr-q1',
+            category: 'noise-resistance',
+            query: 'database connection pool exhausted max connections',
+            expectedIds: ['nr-1'],
+            description: 'find DB issue despite 30 noise entries',
+        },
+        {
+            id: 'nr-q2',
+            category: 'noise-resistance',
+            query: 'S3 bucket public access security misconfiguration',
+            expectedIds: ['nr-2'],
+            description: 'find S3 issue despite noise',
+        },
+        {
+            id: 'nr-q3', category: 'noise-resistance', query: 'PostgreSQL migration constraint', expectedIds: ['dr-1'],
+            description: 'find earlier DB entry despite noise'
+        },
+        {
+            id: 'nr-q4',
+            category: 'noise-resistance',
+            query: 'JWT token security expiration',
+            expectedIds: ['dr-5'],
+            description: 'find security entry despite noise',
+        },
     ];
-    for (let i = 0; i < noiseTopics.length; i++) {
-        entries.push(mem(`noise-${i}`, noiseTopics[i], { created: dateOffset(i), tags: ['topic:office'] }));
-    }
-    entries.push(mem('nr-1', 'The database connection pool was exhausted because max_connections was set to 10 but the application had 25 concurrent requests', {
-        created: dateOffset(20), tags: ['topic:database', 'topic:performance'],
-    }), mem('nr-2', 'The S3 bucket policy was misconfigured allowing public read access to customer data uploads', {
-        created: dateOffset(21), tags: ['topic:security', 'topic:s3'],
-    }));
-    cases.push({
-        id: 'nr-q1', category: 'noise-resistance', query: 'database connection pool exhausted max connections', expectedIds: ['nr-1'],
-        description: 'find DB issue despite 30 noise entries'
-    }, {
-        id: 'nr-q2', category: 'noise-resistance', query: 'S3 bucket public access security misconfiguration', expectedIds: ['nr-2'],
-        description: 'find S3 issue despite noise'
-    }, {
-        id: 'nr-q3', category: 'noise-resistance', query: 'PostgreSQL migration constraint', expectedIds: ['dr-1'],
-        description: 'find earlier DB entry despite noise'
-    }, {
-        id: 'nr-q4', category: 'noise-resistance', query: 'JWT token security expiration', expectedIds: ['dr-5'], description: 'find security entry despite noise'
-    });
+}
+function addNoiseResistance(entries, cases) {
+    entries.push(...noiseEntries(), ...noiseTargetEntries());
+    cases.push(...noiseCases());
 }
 /** Multi-hop: entity chaining across sessions. */
+function multiHopEntries() {
+    return [
+        mem('mh-1', 'speaker:Alice works on the payment gateway integration with Stripe. She found a webhook signature validation ' +
+            'bug.', {
+            created: dateOffset(25),
+            tags: ['speaker:Alice', 'topic:payments'],
+        }),
+        mem('mh-2', 'speaker:Bob reviewed the Stripe webhook code and confirmed the signature validation uses the wrong secret key ' +
+            'from the test environment.', {
+            created: dateOffset(26),
+            tags: ['speaker:Bob', 'topic:payments'],
+        }),
+        mem('mh-3', 'speaker:Alice fixed the Stripe webhook by switching to the production secret key. Payment confirmations now ' +
+            'arrive within 2 seconds.', {
+            created: dateOffset(27),
+            tags: ['speaker:Alice', 'topic:payments'],
+        }),
+        mem('mh-4', 'speaker:Carol reported that the billing dashboard shows incorrect revenue numbers because it reads from the ' +
+            'payments_raw table instead of ' +
+            'payments_reconciled.', {
+            created: dateOffset(28),
+            tags: ['speaker:Carol', 'topic:billing'],
+        }),
+    ];
+}
+function multiHopCases() {
+    return [
+        {
+            id: 'mh-q1', category: 'multi-hop', query: 'Who fixed the Stripe webhook bug and what was the root cause?',
+            expectedIds: ['mh-1', 'mh-2', 'mh-3'], description: 'chain Alice -> Stripe -> Bob -> fix'
+        },
+        {
+            id: 'mh-q2', category: 'multi-hop', query: 'What are all the payment-related issues the team discussed?',
+            expectedIds: ['mh-1', 'mh-2', 'mh-3', 'mh-4'], description: 'find all payment topics across speakers'
+        },
+    ];
+}
 function addMultiHop(entries, cases) {
-    entries.push(mem('mh-1', 'speaker:Alice works on the payment gateway integration with Stripe. She found a webhook signature validation bug.', {
-        created: dateOffset(25), tags: ['speaker:Alice', 'topic:payments'],
-    }), mem('mh-2', 'speaker:Bob reviewed the Stripe webhook code and confirmed the signature validation uses the wrong secret key from the test environment.', {
-        created: dateOffset(26), tags: ['speaker:Bob', 'topic:payments'],
-    }), mem('mh-3', 'speaker:Alice fixed the Stripe webhook by switching to the production secret key. Payment confirmations now arrive within 2 seconds.', {
-        created: dateOffset(27), tags: ['speaker:Alice', 'topic:payments'],
-    }), mem('mh-4', 'speaker:Carol reported that the billing dashboard shows incorrect revenue numbers because it reads from the payments_raw table instead of ' +
-        'payments_reconciled.', {
-        created: dateOffset(28), tags: ['speaker:Carol', 'topic:billing'],
-    }));
-    cases.push({
-        id: 'mh-q1', category: 'multi-hop', query: 'Who fixed the Stripe webhook bug and what was the root cause?',
-        expectedIds: ['mh-1', 'mh-2', 'mh-3'], description: 'chain Alice -> Stripe -> Bob -> fix'
-    }, {
-        id: 'mh-q2', category: 'multi-hop', query: 'What are all the payment-related issues the team discussed?',
-        expectedIds: ['mh-1', 'mh-2', 'mh-3', 'mh-4'], description: 'find all payment topics across speakers'
-    });
+    entries.push(...multiHopEntries());
+    cases.push(...multiHopCases());
 }
 export function buildSyntheticCorpus() {
     const entries = [];
@@ -203,9 +408,9 @@ export function buildSyntheticCorpus() {
 }
 /** Each category's mean scores, in the order the corpus first names the category. */
 function featureAverages(cases, caseResults) {
-    const categories = [...new Set(cases.map(c => c.category))];
-    return categories.map(cat => {
-        const catCases = caseResults.filter(r => r.case.category === cat);
+    const categories = [...new Set(cases.map((c) => c.category))];
+    return categories.map((cat) => {
+        const catCases = caseResults.filter((r) => r.case.category === cat);
         const n = catCases.length;
         const avgMrr = catCases.reduce((s, r) => s + r.mrrVal, 0) / n;
         const avgR5 = catCases.reduce((s, r) => s + r.r5, 0) / n;
@@ -282,7 +487,8 @@ export function detectRegressions(baseline, current) {
     };
 }
 export function resultToBaseline(result) {
-    // SAFETY: the loop below sets every FeatureCategory key from result.features, which holds every category in buildSyntheticCorpus's cases.
+    // SAFETY: the loop below sets every FeatureCategory key from result.features, which holds every category in
+    // buildSyntheticCorpus's cases.
     const features = {};
     for (const f of result.features) {
         features[f.category] = { mrr: f.mrr, recallAt5: f.recallAt5, ndcgAt5: f.ndcgAt5 };
@@ -301,48 +507,64 @@ function deltaStr(d) {
     const sign = d >= 0 ? '+' : '';
     return sign + pct(d);
 }
+function formatFeatureRow(feat, base) {
+    let vsBaseline = 'NEW';
+    let verdict = 'baseline';
+    if (base) {
+        const delta = feat.ndcgAt5 - base.ndcgAt5;
+        vsBaseline = deltaStr(delta);
+        verdict = delta < -REGRESSION_THRESHOLD ? 'REGRESS' : 'PASS';
+    }
+    const cells = [
+        pad(feat.category, 24),
+        pad(pct(feat.mrr), 8),
+        pad(pct(feat.recallAt5), 8),
+        pad(pct(feat.ndcgAt5), 8),
+        pad(vsBaseline, 12),
+        verdict,
+    ];
+    return cells.join(' | ');
+}
+function formatReportLines(report) {
+    const lines = [];
+    if (report.verdict === 'REGRESSION') {
+        lines.push('');
+        lines.push(`REGRESSIONS DETECTED (>${pct(REGRESSION_THRESHOLD)} drop):`);
+        for (const r of report.regressions) {
+            lines.push(`  ${r.category}.${r.metric}: ${pct(r.baseline)} -> ${pct(r.current)} (${deltaStr(r.delta)})`);
+        }
+    }
+    if (report.improvements.length > 0) {
+        lines.push('');
+        lines.push('Improvements:');
+        for (const imp of report.improvements) {
+            lines.push(`  ${imp.category}.${imp.metric}: ${pct(imp.baseline)} -> ${pct(imp.current)} (${deltaStr(imp.delta)})`);
+        }
+    }
+    lines.push('');
+    lines.push(`Verdict: ${report.verdict}${report.regressions.length > 0 ? ` (${report.regressions.length} regressions)` : ''}`);
+    return lines;
+}
 export function formatResult(result, baseline) {
     const lines = [];
     lines.push(`Hippo Eval v${result.version} — ${result.totalCases} queries, ${result.durationMs}ms`);
     lines.push('');
-    lines.push(`${pad('Feature', 24)} | ${pad('MRR', 8)} | ${pad('R@5', 8)} | ${pad('NDCG@5', 8)} | ${pad('vs baseline', 12)} | verdict`);
-    lines.push(`${'─'.repeat(24)}-|-${'─'.repeat(8)}-|-${'─'.repeat(8)}-|-${'─'.repeat(8)}-|-${'─'.repeat(12)}-|${'─'.repeat(8)}`);
+    lines.push(`${pad('Feature', 24)} | ${pad('MRR', 8)} | ${pad('R@5', 8)} | ${pad('NDCG@5', 8)} | ${pad('vs baseline', 12)} | ` +
+        `verdict`);
+    lines.push(`${'─'.repeat(24)}-|-${'─'.repeat(8)}-|-${'─'.repeat(8)}-|-` +
+        `${'─'.repeat(8)}-|-${'─'.repeat(12)}-|${'─'.repeat(8)}`);
     let report;
     if (baseline) {
         report = detectRegressions(baseline, result);
     }
     for (const feat of result.features) {
-        const base = baseline?.features[feat.category];
-        let vsBaseline = 'NEW';
-        let verdict = 'baseline';
-        if (base) {
-            const delta = feat.ndcgAt5 - base.ndcgAt5;
-            vsBaseline = deltaStr(delta);
-            verdict = delta < -REGRESSION_THRESHOLD ? 'REGRESS' : 'PASS';
-        }
-        const cells = [pad(feat.category, 24), pad(pct(feat.mrr), 8), pad(pct(feat.recallAt5), 8), pad(pct(feat.ndcgAt5), 8), pad(vsBaseline, 12), verdict];
-        lines.push(cells.join(' | '));
+        lines.push(formatFeatureRow(feat, baseline?.features[feat.category]));
     }
     lines.push('');
-    lines.push(`Overall: MRR ${pct(result.overall.mrr)} | R@5 ${pct(result.overall.recallAt5)} | NDCG@5 ${pct(result.overall.ndcgAt5)}`);
-    if (report) {
-        if (report.verdict === 'REGRESSION') {
-            lines.push('');
-            lines.push(`REGRESSIONS DETECTED (>${pct(REGRESSION_THRESHOLD)} drop):`);
-            for (const r of report.regressions) {
-                lines.push(`  ${r.category}.${r.metric}: ${pct(r.baseline)} -> ${pct(r.current)} (${deltaStr(r.delta)})`);
-            }
-        }
-        if (report.improvements.length > 0) {
-            lines.push('');
-            lines.push('Improvements:');
-            for (const imp of report.improvements) {
-                lines.push(`  ${imp.category}.${imp.metric}: ${pct(imp.baseline)} -> ${pct(imp.current)} (${deltaStr(imp.delta)})`);
-            }
-        }
-        lines.push('');
-        lines.push(`Verdict: ${report.verdict}${report.regressions.length > 0 ? ` (${report.regressions.length} regressions)` : ''}`);
-    }
+    lines.push(`Overall: MRR ${pct(result.overall.mrr)} | R@5 ${pct(result.overall.recallAt5)} | NDCG@5 ` +
+        `${pct(result.overall.ndcgAt5)}`);
+    if (report)
+        lines.push(...formatReportLines(report));
     return lines.join('\n');
 }
 //# sourceMappingURL=eval-suite.js.map