MAR 20, 2026 — SESSION 1
Foundation
AWS RDS PostgreSQL setup · MySQL→PostgreSQL conversion (1,731 lines) · 21 tables imported · 700 update_records · bfi_property_id discovered as universal identifier · database_architecture.html created
MAR 22, 2026 — SESSION 2
ETL Scripts + First Insights
3 ETL scripts built · comp_prices already had 6,963 rows · Both active hotels priced BELOW market ($110 vs $138–$143 comp avg) · Quality Inn Saturday: $120 vs competitors $243 — $123/room/night gap found
MAR 23, 2026 — SESSIONS 3 & 3b
OAuth2 + First Live ETL
Google org policy permanently blocked service accounts → switched to OAuth2 · All 3 ETL scripts updated · OAuth2 live with auto-refresh · property_master: 32 rows · comp_prices: 12,154 rows · History: 57,891 total · Scale target set: 200+ properties · pricing_intelligence_schema.sql created
MAR 24–30, 2026 — SESSION 4
Automation + CoStar STR
macOS launchd automation — initially failed, fixed Mar 30 (Full Disk Access) · Pipeline: 17,931 rows · comp_prices_history: 75,562 · CoStar STR parser built for 7 hotels · 4 STR tables: 1,043 rows total · Monthly STAR data: Occ/ADR/RevPAR + MPI/ARI/RGI
MAR 30, 2026 — SESSION 5
Full Architecture Design
6 data streams · 7-layer system architecture · FT-Transformer neural network selected · Per-property comparison engine · Human-in-the-loop (ACCEPT/OVERRIDE/SKIP) · AI reasoning layer (Claude API) · Cost estimate: $95/mo at 25 properties → $848/mo at 500
MAR 31, 2026 — SESSION 6
Comp Mapping Pipeline
Comp mapping end-to-end: Compset_mapping tab → DB → live Google Sheet (3 tabs) · Smart client detection · Pricing engine DB tables created (5 tables) · 106 mapping rows across 18 properties
APR 1, 2026 — SESSION 7
Pricing Engine Model + Room Categories
7-step pricing model designed (rank waterfall → base price → heat → multipliers → floor/ceiling → booking curve) · Room category classification engine: 21 categories, hierarchy-based, word-boundary matching · 3 LaunchAgents installed
APR 2, 2026 — SESSION 8
Active Properties Framework
26 active properties · active_properties_clean = source of truth · raw→clean→history pipeline · CLIENT/COMPETITOR detection fixed · Property Master column names lowercased
APR 6, 2026 — SESSION 9
ETL Reliability
25 active properties · 21 with comp prices · ETL timeouts added (15s connect, 2min query) · caffeinate wrappers · etl_catchup.py for missed runs · 6 launchd agents · system_flow.html live table counts
APR 6, 2026 — SESSION 10
Room Category Matrix (Bugs Found)
Room matrix generated but 3 bugs found: rateplan_id garbage (only values 2,4,5,7) · hotel name matching too aggressive · category filter too restrictive · Solutions identified, fixes queued
APR 13–17, 2026 — SESSION 11
Device Independence
GitHub repo created (LarryBFIS/rms_pipeline) · All code desensitized · ETL migrated to GitHub Actions (cloud) · 5 GitHub Secrets · utils/db.py NaN bug fixed · AWS security group opened · Live dashboard published · Room category diagnostic confirmed all 3 bug fixes · JOIN KEY: agg_id = bdc_hotel_id (153 matches)
APR 19, 2026 — SESSION 12
AI Pricing Showcase + Vision
AI Pricing Showcase published · Pricing model corrected: room-category competitive intelligence (not multipliers) · FT-Transformer: predicts revenue impact · Risk capacity + market heat designed · Engineer data audited (frozen since Mar 20) · PROJECT_MEMORY auto-update · Company vision: first AI-based RMS for US hospitality
NEXT — SESSION 13
Pricing Rule Engine (Build)
Build steps 1–6 in Python · Fix room category matrix with agg_id join · Replace showcase example with live data · Draft engineer spec for PMS feed · 60-sec video script
FUTURE
FT-Transformer + AI Layer
Neural network needs 60–90 days of rule engine output · Claude API reasoning layer · Owner communication system · Full company platform (5 services)