-
Phase 0 — Prep & Initial Setup
Completed $250 (completed)- Project structure & stack finalized
- Backend API + basic Excel parsing
- Normalization cache scaffolding
- Docker Compose baseline
-
Phase 1 — POS Integration & Data Foundation
~3 weeks (part‑time) 50 hrs × $50 = $2,500 (+ DO droplet)- PostgreSQL schema + pgvector
- Toast API integration (auth, pulls)
- Square/Fintech connectors
- Universal POS data mapper & scheduler
-
Phase 2 — LLM‑Driven Normalization
~2 weeks (part‑time) 40 hrs × $50 = $2,000 (+ ~$100–200/mo LLM)- Normalization microservice (DeepSeek/OpenAI‑compatible)
- Confidence scoring & advanced caching
- Human‑in‑the‑loop review UI
-
Phase 3 — Frontend & Client Management
~2 weeks (part‑time) 40 hrs × $50 = $1,500- POS connection management UI
- Client onboarding workflow
- Dashboards: data quality, sync health, KPIs
- Responsive UI (mobile-friendly)
-
Phase 4 — Graph & Advanced Analytics
Future expansion TBD (as needed)- Apache AGE property graph (inside Postgres)
- Cross‑POS analytics, predictive insights
- Anomaly detection & alerts
Stack (Open & Low‑Cost)
- • PostgreSQL + pgvector (LLM embeddings)
- • Apache AGE (property graph in Postgres)
- • FastAPI / Elysia for services
- • Metabase / Looker Studio for dashboards
- • DeepSeek / OpenAI‑compatible models with caching
Immediate Next Step
Migrate current Excel sheets to a PostgreSQL database under your control, with managed support. This unlocks consistent reporting immediately while we layer on normalization and graph analytics.
Budget Snapshot
- • Total dev budget target: $6,000
- • Ongoing infra: $12–24/mo (DO + Metabase)
- • LLM usage: $100–200/mo (variable)