khoros-aurora-sdk 26.1.2 → 26.2.0

package/.github/CODEOWNERS

@@ -1 +1 @@
-* @lithiumtech/eng-khoros-community-dev
+* @lithiumtech/eng-khoros-community-admin

package/.github/workflows/npm-publish.yml

@@ -1,12 +1,11 @@
 # This workflow will publish a package to NPM Packages when a release is created
-# Uses OIDC Trusted Publishing - no npm token required
-# Setup: Configure trusted publisher at https://www.npmjs.com/package/khoros-aurora-sdk/access
 
 name: Node.js Package
 
 on:
   release:
     types: [created, edited]
+  workflow_dispatch:
 
 jobs:
   publish-npm:
@@ -18,14 +17,9 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-node@v4
         with:
-          node-version: '22.x'
+          node-version: '24.x'
           registry-url: https://registry.npmjs.org/
-      - name: Upgrade npm for OIDC support
-        run: npm install -g npm@latest
-      - name: Verify npm version
-        run: npm --version
       - name: Publish to npm
-        # Note: --provenance requires public repo, removed for internal/private repos
         run: |
           if [[ "${{ github.event.release.prerelease }}" == "true" ]]; then
             npm publish --access public --tag pre

package/CHANGELOG.md

@@ -1,12 +1,5 @@
 ### CHANGE LOG
 
-**26.1.0-pre.2**
-
-- Fixed ESLint resolver interface compatibility issue that caused build failures in CI environments
-  - Updated `eslint-import-resolver-typescript` from 3.6.3 to 3.8.1
-  - Resolves compatibility issue with `eslint-plugin-import@2.32.0`
-  - Fixes error: "EslintPluginImportResolveError: Error while loading rule 'import/no-cycle': typescript with invalid interface loaded as resolver"
-
 **25.4.0-rc.2**
 
 - Added the following bootstrap components:

package/build-hash.txt

@@ -1 +1 @@
-80958f1
+dfcfbdc

package/graphql/ai-moderation-dashboard.graphqls

@@ -0,0 +1,317 @@
+# AI Moderation Dashboard GraphQL Schema
+# Provides queries for the AI moderation insights dashboard
+
+extend type Query {
+    """
+    Get comprehensive AI moderation dashboard data for a time range.
+    Returns summary statistics, accuracy metrics, daily trends, and provider comparison.
+    """
+    aiModerationDashboard(
+        "Start time (epoch milliseconds)"
+        startTime: Long!
+
+        "End time (epoch milliseconds)"
+        endTime: Long!
+
+        "Optional: Filter by specific provider"
+        provider: String
+    ): AIModerationDashboard!
+
+    """
+    Get recommendations for improving AI moderation accuracy.
+    Analyzes recent data and suggests threshold adjustments, provider changes, etc.
+    """
+    aiModerationRecommendations(
+        "Lookback period in days (default: 30)"
+        lookbackDays: Int
+    ): AIModerationRecommendations!
+}
+
+extend type Mutation {
+    """
+    Manually trigger AI moderation metrics aggregation.
+    Runs the aggregation service immediately and resets the timer.
+    Admin-only operation.
+    """
+    refreshAIModerationMetrics: AIModerationMetricsRefreshResult!
+}
+
+"""
+Result of manual metrics refresh operation.
+"""
+type AIModerationMetricsRefreshResult {
+    "Whether the refresh was successful"
+    success: Boolean!
+
+    "Number of hour buckets processed"
+    hoursProcessed: Int!
+
+    "Timestamp when refresh completed (epoch ms)"
+    completedAt: Long!
+
+    "Message describing the result"
+    message: String!
+}
+
+"""
+Complete dashboard data including summary, metrics, trends, and comparisons.
+"""
+type AIModerationDashboard {
+    "Overall summary statistics for the time range"
+    summary: AIModerationDashboardSummary!
+
+    "Hourly accuracy metrics (precision, recall, F1, etc.)"
+    accuracyMetrics: [AIModerationAccuracyMetric!]!
+
+    "Daily aggregated statistics"
+    dailyStats: [AIModerationDailyStats!]!
+
+    "Provider performance comparison"
+    providerComparison: [AIModerationProviderStats!]!
+
+    "Time range for this data"
+    timeRange: TimeRange!
+}
+
+"""
+Summary statistics aggregated across the entire time range.
+"""
+type AIModerationDashboardSummary {
+    "Total AI moderation requests processed"
+    totalRequests: Int!
+
+    "Total human feedback records (ground truth)"
+    totalFeedback: Int!
+
+    "Overall accuracy (0.0 - 1.0)"
+    overallAccuracy: Float
+
+    "Average precision across all hours"
+    avgPrecision: Float
+
+    "Average recall across all hours"
+    avgRecall: Float
+
+    "Average F1 score"
+    avgF1Score: Float
+
+    "Average moderation latency (ms)"
+    avgLatencyMs: Int
+
+    "95th percentile latency (ms)"
+    p95LatencyMs: Int
+
+    "Total errors"
+    totalErrors: Int
+
+    "Error rate (0.0 - 1.0)"
+    errorRate: Float
+
+    "Total cost in USD (V1_5)"
+    totalCostUsd: Float
+
+    "Average cost per request in USD"
+    avgCostPerRequest: Float
+}
+
+"""
+Hourly accuracy metrics with confusion matrix and calculated scores.
+"""
+type AIModerationAccuracyMetric {
+    "Time bucket (ISO 8601 datetime)"
+    timeBucket: String!
+
+    "AI provider name"
+    provider: String!
+
+    "Total AI verdicts in this hour"
+    totalVerdicts: Int!
+
+    "Total human feedback received"
+    totalFeedback: Int!
+
+    "True positives (AI rejected, human confirmed)"
+    truePositives: Int!
+
+    "False positives (AI rejected, human approved)"
+    falsePositives: Int!
+
+    "True negatives (AI approved, human confirmed)"
+    trueNegatives: Int!
+
+    "False negatives (AI approved, human rejected)"
+    falseNegatives: Int!
+
+    "Precision score (0.0 - 1.0)"
+    precision: Float
+
+    "Recall score (0.0 - 1.0)"
+    recall: Float
+
+    "F1 score (0.0 - 1.0)"
+    f1Score: Float
+
+    "Accuracy score (0.0 - 1.0)"
+    accuracy: Float
+
+    "Average latency (ms)"
+    avgLatencyMs: Int
+
+    "Median latency (ms)"
+    p50LatencyMs: Int
+
+    "95th percentile latency (ms)"
+    p95LatencyMs: Int
+
+    "99th percentile latency (ms)"
+    p99LatencyMs: Int
+}
+
+"""
+Daily aggregated statistics for trend analysis.
+"""
+type AIModerationDailyStats {
+    "Date (YYYY-MM-DD)"
+    date: String!
+
+    "Total requests on this day"
+    totalRequests: Int!
+
+    "Number rejected"
+    rejectedCount: Int!
+
+    "Number accepted"
+    acceptedCount: Int!
+
+    "Average confidence score"
+    avgConfidence: Float
+
+    "Average latency (ms)"
+    avgLatencyMs: Int
+
+    "Total errors"
+    errorCount: Int
+}
+
+"""
+Provider performance comparison statistics.
+"""
+type AIModerationProviderStats {
+    "Provider name (OPENAI, ANTHROPIC, etc.)"
+    provider: String!
+
+    "Total requests processed"
+    totalRequests: Int!
+
+    "Total feedback received"
+    totalFeedback: Int!
+
+    "Overall accuracy (0.0 - 1.0)"
+    overallAccuracy: Float
+
+    "Overall precision (0.0 - 1.0)"
+    overallPrecision: Float
+
+    "Overall recall (0.0 - 1.0)"
+    overallRecall: Float
+
+    "Average latency (ms)"
+    avgLatencyMs: Int
+
+    "95th percentile latency (ms)"
+    p95LatencyMs: Int
+
+    "Error rate (0.0 - 1.0)"
+    errorRate: Float
+
+    "Total tokens consumed"
+    totalTokens: Long
+
+    "Average tokens per request"
+    avgTokensPerRequest: Int
+
+    "Total cost in USD (V1_5)"
+    totalCostUsd: Float
+
+    "Average cost per request in USD"
+    avgCostPerRequest: Float
+
+    "Cost per accurate verdict (V1_5)"
+    costPerAccurateVerdict: Float
+}
+
+# Note: Using existing TimeRange type from metricSchema.graphqls
+
+"""
+Recommendations for improving AI moderation.
+"""
+type AIModerationRecommendations {
+    "List of actionable recommendations"
+    recommendations: [AIModerationRecommendation!]!
+
+    "Analysis metadata"
+    analysis: RecommendationAnalysis!
+}
+
+"""
+Individual recommendation with priority and expected impact.
+"""
+type AIModerationRecommendation {
+    "Unique recommendation ID"
+    id: String!
+
+    "Recommendation type (THRESHOLD, PROVIDER, ERROR_INVESTIGATION, etc.)"
+    type: RecommendationType!
+
+    "Priority level"
+    priority: RecommendationPriority!
+
+    "Title (short description)"
+    title: String!
+
+    "Detailed description and rationale"
+    description: String!
+
+    "Expected impact description"
+    expectedImpact: String
+
+    "Supporting data/metrics"
+    supportingData: JSON
+}
+
+"""
+Analysis metadata for recommendations.
+"""
+type RecommendationAnalysis {
+    "Analysis timestamp (epoch ms)"
+    timestamp: Long!
+
+    "Lookback period (days)"
+    lookbackDays: Int!
+
+    "Sample size (total requests analyzed)"
+    sampleSize: Int!
+
+    "Confidence level (HIGH, MEDIUM, LOW)"
+    confidence: String!
+}
+
+"""
+Recommendation type enum.
+"""
+enum RecommendationType {
+    THRESHOLD_ADJUSTMENT
+    PROVIDER_SWITCH
+    ERROR_INVESTIGATION
+    PERFORMANCE_TUNING
+    COST_OPTIMIZATION
+}
+
+"""
+Recommendation priority enum.
+"""
+enum RecommendationPriority {
+    HIGH
+    MEDIUM
+    LOW
+}

package/graphql/aiAnswerLogSchema.graphqls

@@ -0,0 +1,144 @@
+extend type Query {
+  """
+  Fetches a paginated list of AI Answer logs for transparency and debugging.
+  Logs all /clmsearch endpoint invocations.
+  """
+  aiAnswerLogs(
+    "Constraints to filter the AI answer logs."
+    constraints: AiAnswerLogConstraints
+
+    "The number of results to return (first N)."
+    first: Int
+
+    "The cursor to return results after."
+    after: String
+
+    "The number of results to return (last N)."
+    last: Int
+
+    "The cursor to return results before."
+    before: String
+  ): AiAnswerLogConnection!
+}
+
+"""
+Defines an AI Answer log entry from the /clmsearch endpoint.
+"""
+type AiAnswerLog {
+  "Unique identifier for this log entry."
+  id: ID!
+
+  "The original search query text."
+  queryText: String!
+
+  "UID of the original community message that triggered the search."
+  messageUid: Int
+
+  "Subject/title of the original message."
+  messageSubject: String
+
+  "Board display ID for URL construction."
+  boardId: String
+
+  "UID of the AI-generated reply message."
+  answerMessageUid: Int
+
+  "Whether the query was classified as a valid question."
+  isQuery: Boolean!
+
+  "Whether an AI answer was successfully generated."
+  aiAnswerGenerated: Boolean!
+
+  "Whether an Expert Recommendation Card was generated."
+  expertCardGenerated: Boolean!
+
+  "The calculated answer confidence score (0.0-1.0)."
+  confidenceScore: Float
+
+  "The confidence threshold setting at the time of query (0.0-1.0)."
+  confidenceThreshold: Float
+
+  "Score distribution factor (0.0-1.0) - measures consistency across top-K results (40% weight)."
+  scoreDistribution: Float
+
+  "Author authority factor (0.0-1.0) - measures expertise of citation authors (25% weight)."
+  authorAuthority: Float
+
+  "Temporal recency factor (0.0-1.0) - measures age of cited content (15% weight)."
+  temporalRecency: Float
+
+  "Answer quality factor (0.0-1.0) - measures citation usage and verification (20% weight)."
+  answerQuality: Float
+
+  "Number of citations provided to the LLM for answer generation."
+  citationsProvided: Int
+
+  "Number of citations actually referenced in the generated answer."
+  citationsUsed: Int
+
+  "Citation coverage ratio (0.0-1.0) - citations used divided by citations provided."
+  citationCoverage: Float
+
+  "Reasons why AI answer was not generated (can be multiple)."
+  noAnswerReasons: [String!]
+
+  "When this query was processed."
+  processedAt: DateTime!
+
+  "Embedding model used for query vectorization (format: provider|model)."
+  embeddingModel: String
+
+  "Generation model used for answer creation (format: provider|model)."
+  generationModel: String
+}
+
+"""
+Connection object for paginated AI Answer log results.
+"""
+type AiAnswerLogConnection {
+  "Total count of results across all pages."
+  totalCount: Int!
+
+  "Pagination information."
+  pageInfo: PageInfo!
+
+  "The log entries."
+  edges: [AiAnswerLogEdge!]!
+}
+
+"""
+Edge type for AI Answer log results with cursor.
+"""
+type AiAnswerLogEdge implements Edge {
+  "Cursor for pagination."
+  cursor: String
+
+  "The AI Answer log entry."
+  node: AiAnswerLog!
+}
+
+"""
+Constraints for filtering AI Answer logs.
+"""
+input AiAnswerLogConstraints {
+  "Filter by date range."
+  dateRange: DateTimeGtLtConstraint
+
+  "Filter by whether query was classified as a question."
+  isQuery: Boolean
+
+  "Filter by whether AI answer was generated."
+  aiAnswerGenerated: Boolean
+
+  "Filter by whether expert card was generated."
+  expertCardGenerated: Boolean
+
+  "Filter by specific no-answer reasons (matches any in the list)."
+  noAnswerReasons: [String!]
+
+  "Filter by minimum confidence score (0.0-1.0)."
+  confidenceScoreMin: Float
+
+  "Filter by maximum confidence score (0.0-1.0)."
+  confidenceScoreMax: Float
+}

package/graphql/baseSchema.graphqls

@@ -360,6 +360,7 @@ enum ModerationStatus {
   MARKED_UNDECIDED
   MARKED_APPROVED
   MARKED_REJECTED
+  APPEALED
 }
 
 """