Skip to content

SavvyDealerDataWarehouse Context Cache

Repo: https://github.com/RayNawara/SavvyDealerDataWarehouse · Path: C:/Users/adam/SavvyDealerDataWarehouse-context/ · Owner: Ray Nawara (Adam maintains the context cache) Status: Active dev · % Done: 10 · Last commit: 2026-04-08 (cache snapshot) Deployed: not deployed (prompt/spec package)

What it is

A local snapshot of the key files, schemas, and docs from Ray Nawara's SavvyDealerDataWarehouse Rails Multi AI Agent repo, cached so Adam's persona work integrates cleanly with Ray's production pipeline without guessing at interfaces.

Why it exists

Ray's repo is the SavvyDealer data warehouse of record — the unified PostgreSQL home for VinSolutions, Agile CRM, vAuto, CallRail, Drive Centric, GA4, and Google Ads data. Adam's analyst personas consume its ConsultationMetricsService payload and plug into its AiPromptTemplate GUI. Without a local cache, every persona decision would require cross-repo spelunking. The cache lets Adam verify table names, service shapes, and persona keys before shipping prompt changes.

How it works

Flat directory of ~27 files (~250 KB) copied out of the Rails app — schemas, services, jobs, models, seeds, and the multi-agent docs. Not an active codebase — reference only.

  • Rails 8.0.4 / Ruby 3.4.6 app at apps/savvy_warehouse, Kamal deployment, Solid Queue/Cache/Cable, Devise auth.
  • Multi-tenant PostgreSQL 18 (savvy_warehouse_production) with dealer isolation and JSONB storage.
  • Multi AI Agent system: four personas (persona_sales, persona_marketing, persona_bdc, persona_inventory) run in parallel via ConsultPersonaJob, synthesized by SynthesizeOpinionsJob.
  • Personas editable at /settings/ai_prompts via AiPromptTemplate with PaperTrail history and dealer > group > system scope.
  • Production: Hetzner cluster (2 web + 1 job + 1 DB + LB), staging at Netcup Nuremberg.

What's done

  • Full cache of CLAUDE.md, README.md, MULTI_AI_AGENT_SYSTEM.md, SESSION_PROGRESS.md, THURSDAY_MEETING_NOTES.md.
  • All schema SQL files (warehouse, customers, inventory, CallRail, vAuto) + schema_README.md.
  • Rails source: consultations_controller.rb, consult_persona_job.rb, start_consultation_job.rb, synthesize_opinions_job.rb, ai_prompt_template.rb, consultation.rb, opinion.rb, persona.rb, agent_runner.rb, consultation_metrics_service.rb.
  • Persona seed + migration: db/seeds/personas.rb, seed_persona_prompt_templates.rb.
  • Persona/agent docs: MULTI_PERSONA_ANALYSIS_PERFORMANCE.md, CEO_PERSONA_FEEDBACK.md (Adam's Jan 2026 feedback), MULTI_PERSONA_ANALYSIS_PLAN.md, AI_PROMPTS.md.
  • entities.json domain map.

What's next

  • Refresh cache when schema or persona pipeline changes upstream.
  • Contribute data dictionary + schema doc back to Ray's repo so both sides share one source of truth.
  • Add ConsultationMetricsService payload shape spec (what keys appear when, which dealers have spotty segments).
  • Track open items from Adam/Ray Thursday meetings as they affect persona contracts.

Where the code/content lives

  • C:/Users/adam/SavvyDealerDataWarehouse-context/CLAUDE.md — upstream project guide.
  • C:/Users/adam/SavvyDealerDataWarehouse-context/MULTI_AI_AGENT_SYSTEM.md — persona architecture.
  • C:/Users/adam/SavvyDealerDataWarehouse-context/apps_savvy_warehouse_app_services_consultation_metrics_service.rb — the payload builder persona prompts consume.
  • C:/Users/adam/SavvyDealerDataWarehouse-context/apps_savvy_warehouse_app_jobs_consult_persona_job.rb + synthesize_opinions_job.rb — the orchestration.
  • C:/Users/adam/SavvyDealerDataWarehouse-context/apps_savvy_warehouse_app_models_ai_prompt_template.rb — editable prompt storage.
  • C:/Users/adam/SavvyDealerDataWarehouse-context/schema_*.sql — warehouse tables.
  • C:/Users/adam/SavvyDealerDataWarehouse-context/docs_personas_CEO_PERSONA_FEEDBACK.md — the feedback that produced today's seeded prompts.

Integrations

Feeds ConsultationMetricsService payload to savvy-batch-analyst-prompts personas (v4). Downstream consumers: eventual Ray-team analytics products, dealer-facing batch reports, ai-crm financial summaries, dealer landing pages. Data sources on the warehouse side: VinSolutions API, Agile CRM (MySQL), Drive Centric (MySQL), CallRail API, GA4 via BigQuery→GCS, Google Ads CSV email reports, vAuto CSV.

Don't rebuild this — extend it

This cache is the onboarding bridge between Adam's persona work and Ray's warehouse repo — contribute schema docs and a data dictionary back upstream so both sides stay aligned instead of drifting.