Production fixes:
- Cross-business join safety: all queries join on (review_id, business_id)
- Timestamp normalization: iso_z() for all output timestamps
- Score formula alignment: matches PERIOD_SCORES_QUERY for consistency
- Invariant check: fails if scores.overall != comparisons.current
- primary_run_id: uses max(created_at) in time_window mode
- Language normalization: auto/auto-detect -> unknown
- Review language: majority voting over spans per review
Executive summary guardrails:
- Weakness priority: negative driver > qualifying dip > none
- Dip qualification: within 90 days AND review_count >= 3
- Most recent dip selection when multiple qualify
- No contradiction: "dip" cannot pair with "no major issues"
- Action grounding: must tie to cited weakness or top positive driver
CLI options:
- --no-summary: disable executive summary
- --require-summary: exit code 2 if LLM fails
- --summary-model: configurable model (default gpt-4o-mini)
Includes unit test suite (16 tests) for narrative guardrails.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Previously hardcoded "Last 12 months" which was misleading when data
spanned multiple years. Now calculates the actual period from the
earliest to latest review dates in the dataset.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Changed the business name query to ORDER BY COUNT(*) DESC instead of
arbitrary LIMIT 1, ensuring the correct business is identified even
when trace amounts of other business data leak into a job.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Change action_matrix reference from old 'actions' field.
Add issues_identified and health_score to stage result data.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Transform synthesis stage from consultant memo to productized report:
- New 6-section structure: Executive Summary, Risk Scorecard,
Critical Issues, Strengths, Action Matrix, 90-Day Tracking
- Add chart data aggregation (7 charts: gauge, pies, trends)
- Evidence-grounded LLM prompt requiring quote citations
- Hyper-specific solution generation with WHAT/WHO/WHEN/WHY
- Taxonomy-guided solutions contextualized to business type
- Staff name extraction from quotes for leverage actions
- Success metrics tied to actual complaint keywords
Report sections now include:
- Health score (1-100) with revenue-at-risk estimates
- Risk indicators with color-coded scores and trends
- Critical issues with evidence, root cause, and specific solutions
- Strengths with named staff and specific marketing actions
- Action matrix with effort/impact quadrants and deadlines
- 90-day KPIs with 30/60/90 day targets
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add Stage5Synthesizer class that generates AI narratives and action plans
- Add generate() method to LLMClient for synthesis generation
- Integrate Stage 5 into pipeline runner after route stage
- Add synthesis JSONB column to pipeline.executions table
- Update reviewiq_analytics API to return synthesis data
- Synthesis includes: executive narrative, sentiment/category/timeline insights,
action plan, marketing angles, and priority recommendations
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Frontend:
- Add Synthesis type with action plan, insights, annotations
- ExecutiveSummary: Accept synthesis prop for AI narrative
- SentimentPie: Accept insight prop for contextual explanation
- IntensityHeatmap: Accept insight + highlightDomain props
- TimelineChart: Accept insight + annotations props
- All components gracefully degrade when synthesis is null
Backend:
- Add Stage 4: Synthesize for generating AI narratives
- Gathers context from classified spans
- Generates executive narrative, section insights, action plan
- Produces timeline annotations and marketing angles
- Stores synthesis in pipeline.executions table
Components show AI insights with purple gradient styling when available,
fall back to existing behavior when synthesis is not yet generated.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit implements a plugin-like pipeline architecture with:
Pipeline Core Package (packages/pipeline-core/):
- BasePipeline abstract class all pipelines implement
- PipelineRegistry for database-backed discovery/management
- PipelineRunner for execution with status tracking
- DashboardConfig contracts for dynamic widget definitions
Database Migration (006_pipeline_registry.sql):
- pipeline.registry table for registered pipelines
- pipeline.executions table for execution history
- Views for execution stats and monitoring
ReviewIQ Pipeline Refactor:
- Implements BasePipeline interface
- Adds get_dashboard_config() with widget definitions
- Adds get_widget_data() methods for all dashboard widgets
- Maintains backward compatibility with Pipeline alias
Generic Pipeline API (api/routes/pipelines.py):
- GET /api/pipelines - List all registered pipelines
- GET /api/pipelines/{id} - Pipeline details
- POST /api/pipelines/{id}/execute - Execute pipeline
- GET /api/pipelines/{id}/dashboard - Dashboard config
- GET /api/pipelines/{id}/widgets/{w} - Widget data
- GET /api/pipelines/{id}/executions - Execution history
Frontend Dynamic Dashboard System:
- DynamicDashboard component renders from config
- WidgetRegistry maps types to components
- Widget components: StatCard, LineChart, BarChart,
PieChart, DataTable, Heatmap
- Pipeline API client library
Frontend Pipeline Pages:
- /pipelines - List all registered pipelines
- /pipelines/[id] - Dynamic dashboard for pipeline
- /pipelines/[id]/executions - Execution history
- Pipelines nav item in Sidebar
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add ScraperV1Adapter to transform scraped reviews into pipeline format
- Handles relative timestamps (centerDate)
- Generates deterministic IDs for DOM-sourced reviews
- Filters out empty (rating-only) reviews
- Add sample barbershop reviews (79 reviews, 46 with text)
- Real data from Las Palmas barbershop
- Multi-language: Spanish, English, German, Norwegian, Italian
- Add test_pipeline_real_data.py for E2E testing with real data
- Uses mock classifier based on keywords and rating
- Full pipeline flow: raw -> enriched -> spans -> issues -> facts
Test results with real data:
- 46 reviews processed
- 6 languages detected (es: 35, en: 7, de: 1, no: 1, it: 1, ca: 1)
- 3 issues identified from negative reviews
- 29 fact records aggregated across date range 2017-2025
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
