---
title: Data Architecture Summary — lfcs-pricing
type: claude-md-reference
parent: CLAUDE.md
created: 2026-05-08
status: live
---

# Data architecture summary

Three-base architecture per Stage 4 § 1. Brief reference; full schema in `lfcs-pricing-investigation/investigation-output/stage-4-data-architecture.md`.

## 1. LFCS-Operations base (`appE43UvTyARe5oJs`)

15 operational tables + 8 `_methodology_*` tables (unified-base fallback per T1-fail). Operational scope: per-job / per-client / per-bid.

**Keep + extended (5 existing extended in M0):** `Jobs`, `Drawing Elements`, `Suppliers`, `Bids`, `Bid Line Items`. Existing fields preserved; M0 added `job_category`, `complexity_factors`, `final_submitted_*`, `_internal_*` confidential fields.

**Keep as-is (6 existing):** `Drawings`, `Revisions`, `Specifications`, `RFIs`, `Variations`, `Rates`.

**New operational (5, M0):** `Bid Line Item Overrides` (`tblPp9hjyWvWdRYIK`), `Commercial Events` (`tbl6LXA7evO4CzZTD`), `_Internal_Audit_Trail` (`tblSjvdXg8imY3rUy`), `_Internal_Session_Logs` (`tblrkb9vRtgiSPNtC`), `_Internal_Sanitisation_Log` (`tblZ1syjOOeoaR8SP`).

## 2. `_methodology_*` tables (currently in same base; T1-fail fallback)

Cross-job reusable layer. Unified-base prefix per `_System/CLAUDE.md` "Airtable as canonical query/index layer" + Stage 5 milestone-0 § kill criteria. Migration to a separate `Construction-Methodology` base is a future Rocky-confirmed task (PAT scope upgrade required).

8 tables: `_methodology_Element_Taxonomy`, `_methodology_Reason_Code_Taxonomy`, `_methodology_Concept_Patterns`, `_methodology_Common_Misses_Patterns`, `_methodology_Pre_Pricing_Grill_Templates`, `_methodology_Productivity_Benchmarks`, `_methodology_Rate_Templates`, `_methodology_Exclusion_Library`. Counts post-Task-1: 100 records total.

## 3. RateRight Supabase

Project ref `memscjotxrzqnhrvnnkc`. Out of scope for pricing. Noted only because the strategic frame includes that the operator (Rocky) also runs RateRight — but pricing agent has no read or write to that data.

## 4. Vault structure

- `Operations/Jobs/{Upcoming|Ongoing Jobs|Completed Jobs|Lost Jobs}/{job}/` — per-job 11-section folder
- `construction-methodology/` — vault mirror of `_methodology_*` tables (canonical MDs)
- `Operations/lfcs-standards/` — operational LFCS-specific standards (rate-card, pricing-checklist, folder-spec, templates/)
- `Operations/skills/lfcs-*/SKILL.md` — invokable skills
- `_Internal/{audit-trail,prompt-logs,agent-confidence,sanitisation-log,spike-fixtures,spike-briefs}/` — confidential cross-job agent layer
- Per-job `_audit/` — confidential per-job

Cross-job vs per-job rule: cross-job source-of-truth lives at `_Internal/`. Per-job `_audit/` holds filtered views via mirror skill (one-way: cross-job → per-job, never reverse). Conflict rule: cross-job `_Internal/` wins.