wulf-pulse/migrations/071_analyzer_aggregate_reports.sql

-- AI Ticket Analyzer — Phase 2.6 aggregate reports.
-- See docs/ticket-analyzer-phase2-spec.md → Section D.4.
--
-- Adds:
--   1. analyzer_aggregate_reports — one row per cross-ticket report
--   2. analyzer_stage_executions:
--        a. drop NOT NULL on analysis_id (a stage row may belong to a report instead)
--        b. add aggregate_report_id FK
--        c. add CHECK constraint enforcing exactly one of analysis_id or aggregate_report_id

CREATE TABLE IF NOT EXISTS analyzer_aggregate_reports (
  id                       UUID PRIMARY KEY DEFAULT gen_random_uuid(),
  generated_by_user_id     TEXT REFERENCES "user"(id) ON DELETE SET NULL,
  generated_at             TIMESTAMPTZ NOT NULL DEFAULT NOW(),

  -- Inputs
  filter_criteria          JSONB NOT NULL,
  analysis_ids             UUID[] NOT NULL,
  ticket_count             INT NOT NULL,
  include_itglue_context   BOOLEAN NOT NULL DEFAULT true,
  report_title             TEXT,

  -- SQL-derived (computed before LLM call)
  category_distribution        JSONB,
  client_distribution          JSONB,
  resolution_path_distribution JSONB,
  root_cause_distribution      JSONB,
  date_range_actual            JSONB,

  -- LLM-derived
  documentation_gaps     JSONB,
  process_gaps           JSONB,
  client_patterns        JSONB,
  recurrence_clusters    JSONB,
  systemic_observations  JSONB,
  recommended_actions    JSONB,
  narrative_summary      TEXT,
  executive_summary      TEXT,

  -- Metadata
  total_input_tokens     INT,
  total_output_tokens    INT,
  estimated_cost_usd     NUMERIC(10,4),
  model_used             TEXT,
  itglue_context_included BOOLEAN,
  status                 TEXT NOT NULL DEFAULT 'pending'
                           CHECK (status IN ('pending','running','complete','failed')),
  error_message          TEXT
);

CREATE INDEX IF NOT EXISTS idx_analyzer_aggregate_reports_user_date
  ON analyzer_aggregate_reports (generated_by_user_id, generated_at DESC);

CREATE INDEX IF NOT EXISTS idx_analyzer_aggregate_reports_analysis_ids
  ON analyzer_aggregate_reports USING gin (analysis_ids);

CREATE INDEX IF NOT EXISTS idx_analyzer_aggregate_reports_pending
  ON analyzer_aggregate_reports (status, generated_at)
  WHERE status IN ('pending','running');

COMMENT ON TABLE analyzer_aggregate_reports IS
  'Cross-ticket aggregate reports. Inputs (filter_criteria, analysis_ids) are immutable; outputs (LLM-derived fields) are written once when the runner completes.';

-- analyzer_stage_executions: relax analysis_id, add aggregate_report_id, enforce mutual exclusion.
ALTER TABLE analyzer_stage_executions
  ALTER COLUMN analysis_id DROP NOT NULL;

ALTER TABLE analyzer_stage_executions
  ADD COLUMN IF NOT EXISTS aggregate_report_id UUID
    REFERENCES analyzer_aggregate_reports(id) ON DELETE CASCADE;

DO $$
BEGIN
  IF NOT EXISTS (
    SELECT 1 FROM pg_constraint
     WHERE conname = 'analyzer_stage_executions_parent_check'
  ) THEN
    ALTER TABLE analyzer_stage_executions
      ADD CONSTRAINT analyzer_stage_executions_parent_check
      CHECK ((analysis_id IS NOT NULL) <> (aggregate_report_id IS NOT NULL));
  END IF;
END $$;

CREATE INDEX IF NOT EXISTS idx_analyzer_stage_executions_aggregate_report
  ON analyzer_stage_executions (aggregate_report_id, stage_order)
  WHERE aggregate_report_id IS NOT NULL;