TB-09a — Historical Intelligence Tools

Summary

Build 10 API endpoints at /history/* that turn 85 GB of historical data across 5 archive sources into actionable intelligence. Endpoints: precedent finder, pricing intelligence, repeat client detection, reconciliation watchdog, morning context loader, production day estimator, permit timeline predictor, client communication search, invoice forensics, and jurisdiction expert.

What it produced

• history-service — unified archive search with in-memory cache (900+ lines, 4 sources: anomaly/blkhse/dropbox/xero), fuzzy address matching, branch name normalization
• 10 /history/* endpoints wired into both friday-ai and Kulu

Connections

• depends on: Supabase — projects, SOW lines for current data
• depends on: intake-enrichment-service — reuses jurisdiction lookup and certifier routing
• depends on: cert-generator-v3 — jurisdiction form mapping
• produced: history-service — 10 intelligence endpoints
• feeds into: friday-ai — historical intelligence tools available via tool_use
• feeds into: Kulu — historical-archive skill file