Skip to content

Rootphi-Core — Project Overview

What This Is

Brian's personal creative infrastructure platform — "CreativeOS," the operating system of a full-stack creative's life. Not public-facing yet. Designed to be flexible, scalable, self-maintaining, and robust enough to support rapid deployment of new apps and creative projects. Eventually productizable for other creative entrepreneurs.

Vision

CreativeOS is a creative launchpad — go from idea to deployed prototype to live product quickly. The distance between "I have an idea" and "it's running somewhere people can use it" should be as short as possible. Designed to match the pace and breadth of Brian's creative output.

Aspirational Goal

Productize the platform itself as "CreativeOS" — allowing other creative entrepreneurs to build their own version of this system. The culmination of everything built here could become a product in its own right.

Architecture (Decided 2026-03-27)

Architecture was determined via 7-provider LLM triangulation (Anthropic, OpenAI, Google, xAI, DeepSeek, Mistral, Together AI). Full synthesis document: rootphi-core/creativeos-architecture-synthesis_2026-03-27.md

Core Stack (Consensus)

  • Database: PostgreSQL (7/7 unanimous) via Supabase, with pgvector for embeddings
  • Frontend: Next.js + React + TypeScript (6/7)
  • Hosting: Vercel (frontend) + Supabase (DB/Auth/Realtime) + Cloudflare (edge/CDN) — managed-platform approach, 4/7 majority over raw AWS
  • Architecture Pattern: Event-driven modular monolith (Shopify pattern) — single deployable with strict module boundaries, extract to services later as needed
  • Auth: Start with Supabase Auth, migrate to Clerk or Auth0 for multi-tenancy
  • LLM Orchestration: LangGraph (5/7) for multi-model debate and agentic RAG
  • Knowledge Layer: Hybrid RAG — pgvector for semantic search + Neo4j for knowledge graphs (6/7)
  • Workflow Automation: n8n (6/7, already operational)
  • Canvas Interface: tldraw SDK (6/7) as React component with custom shapes API
  • CI/CD: GitHub Actions (6/7)
  • Multi-tenancy: PostgreSQL Row-Level Security (RLS) from day one, tenant_id on every table

Key Architectural Principles

  • "First tenant" mental model — Brian is user #1 of a platform, not building bespoke software
  • Configuration over code — workflows, dashboards, domain setups stored as config, not hardcoded
  • Plugin architecture — new creative domains added as plugins (shapes + KB config + workflows + dashboard cards)
  • Kernel/application separation — core platform (auth, storage, events, AI, canvas, dashboard) is the kernel; everything else is an app

The CreativeOS Stack Diagram

FRONTEND (Next.js + React)
  Canvas (tldraw SDK) | Dashboard (ADHD Command Center)
  Domain Plugin UIs   | Notification Triage

API LAYER (tRPC or REST, tenant-scoped)

APPLICATION / PLUGIN LAYER
  Diamond MMA | Art Critique | Patent Engine | ...

OS KERNEL SERVICES
  Auth (Supabase) | Events (NATS/pg_notify)
  Storage (S3/R2) | AI Orchestration (LangGraph)
  Knowledge (pgvector + Neo4j) | Workflows (n8n)

DATA LAYER (PostgreSQL)
  pgvector | RLS multi-tenancy | JSON documents
  Event log | Spatial state | Knowledge store

INFRASTRUCTURE
  Vercel | Supabase | Cloudflare | n8n Cloud
  Neo4j Aura | Modal (AI compute)

Infrastructure Readiness Roadmap (6 Phases)

These phases harden rootphi-core itself — the platform layer that everything else sits on. Brian's guiding principle: "I'm patient, and want to do this right the first time." The ordering is intentional: each phase depends on the one before it.

Phase 1: Ecosystem Audit & Architecture Design (current)

Map what exists, identify implicit dependencies and undocumented assumptions, design the target architecture. The infrastructure grew organically across a single productive day — this phase turns "works by coincidence" into "works by design." Includes: scan all skills, configs, scheduled tasks, scripts, mounted folders; build a complete dependency map; draft the architecture document that becomes the blueprint for everything else. - Status: Largely complete. Architecture triangulation done (7 providers, 32 API calls). Consensus stack decided. Phase 1 tech spec is next.

Phase 2: Security Hardening

Formalize key management, audit access patterns, secure the backup pipeline. Comes before QA intentionally — if the security model is wrong, fixing it later means retesting everything. Includes: formalize key rotation procedures (clipboard workflow established), audit what's exposed where, ensure backups don't leak secrets, review all MCP connector access scopes. - Status: Partially done. Clipboard-based key workflow established. OpenAI and Google keys rotated. Full audit pending.

Phase 3: QA & Testing

Validate everything. Live-test the triangulator, verify scheduled tasks actually fire, confirm backups are recoverable, stress-test the context-manager lifecycle. Also: define what "testing a skill" means — no test framework exists yet. - Status: Triangulator live-tested via n8n webhook (28 calls successful). Scheduled tasks and backup verification pending.

Phase 4: Stabilize & Standardize

Fix whatever Phase 3 reveals. Package all skills into final .skill files. Standardize how skills are structured, how configs are shared, how errors are handled. Turn "it works on my machine" into "it works reliably." - Status: Context-manager and workspace-health packaged. Triangulator needs repackaging (stale key + model config).

Phase 5: Documentation & Portability

Capture everything in documentation that's useful to Brian — not boilerplate, but "the stuff that lets you rebuild at 2 AM." Architecture decisions, skill development standards, disaster recovery guide (started), dependency map, dev standards. - Status: Disaster recovery guide written. Architecture synthesis document complete. Full documentation pass pending.

Phase 6: Extend

Only now add new capabilities: image/video critique skill, RAG/knowledge base skill, external integrations. The foundation is solid, tested, documented, and secure — new things built on it won't inherit hidden fragility. - Status: Not started. Concepts documented (Image/Video Critique, Expert Knowledge Base/RAG, Creative Data Room, Moment Maker, Patent Cross-Referencing Engine).

CreativeOS Application Build Phases

These phases build the actual CreativeOS application on top of the hardened infrastructure. Decided via 7-provider triangulation (2026-03-27).

App Phase 1: Foundation (Weeks 1-4)

  1. PostgreSQL + Supabase — database, auth, realtime
  2. Next.js app shell — API routes, basic dashboard, plugin architecture skeleton
  3. n8n integration — already working, formalize as workflow layer
  4. Basic RAG with pgvector — knowledge base v1 for one domain
  5. Tenant isolation (RLS) — bake in from day one

App Phase 2: Canvas + Intelligence (Weeks 5-8)

  1. tldraw canvas — custom shapes, spatial state persistence, AI integration
  2. Priority engine — ADHD dashboard ("what should I do now?")
  3. LangGraph orchestration — formalize multi-LLM debate pipeline
  4. First domain plugin — Diamond MMA or art critique

App Phase 3: Knowledge + Scale (Weeks 9-12)

  1. Neo4j knowledge graph — relationship-heavy domains
  2. Hybrid search — semantic + keyword/BM25
  3. Multi-modal embeddings — images for art, PDFs for patents
  4. Auto-population pipelines — self-maintaining knowledge bases

App Phase 4: Productization (Weeks 13-20)

  1. Plugin marketplace architecture
  2. Onboarding flow — second user (Tyler), then early adopters
  3. Self-hosted option — Docker + Coolify

Skill Ecosystem

  • context-manager: Persistent memory across sessions
  • workspace-health: Structural integrity checking
  • skill-creator: Meta-skill for building/evaluating skills
  • llm-triangulator: Multi-LLM synthesis (n8n backend, 7 providers)
  • (Future skills inherit shared infrastructure)

Connected Integrations

Gmail, Google Calendar, Webflow, HuggingFace, Clay, DocuSign, n8n, JSTOR, Nightfall/Otterly, Gamma, Eraser, DiveDB, Mermaid, Chrome Extension, MCP Registry

Projects Built on This Platform

  • Diamond MMA: Sales automation, influencer outreach, growth systems
  • Creative Data Room: Visual file sharing with watermarking (concept)
  • Moment Maker: Relationship CRM / gifting concierge (concept)
  • Patent Cross-Referencing Engine: IP discovery pipeline (concept)
  • Deep Research Tools: PE-grade analysis and synthesis (concept)
  • Relationship Intelligence Engine: Contact classification and attention surfacing (concept)

Current Status (2026-03-27)

  • Architecture decided via 7-provider triangulation
  • 4 custom skills built and operational
  • n8n workflow operational (LLM Triangulator, 5-node, all 7 providers confirmed working)
  • Google API upgraded to paid plan (Gemini 2.5 Pro)
  • Together AI key rotated
  • Context system fully deployed with backups
  • Phase 1 build ready to begin

Key Reference Documents

  • creativeos-architecture-synthesis_2026-03-27.md — Full triangulation results
  • creativeos-triangulation-brief_2026-03-27.md — Original questions and context brief
  • skills-source/llm-triangulator/ — Updated skill source (SKILL.md + config)
  • 2026-03-27: building-context-skill
  • 2026-03-27: triangulator-skill-build
  • 2026-03-27: context-manager-infrastructure
  • 2026-03-27: api-key-rotation
  • 2026-03-27: creativeos-architecture-triangulation
  • 2026-03-28: creativeos-phase1-spec (Phase 1 tech spec, GitHub + Supabase provisioned, API keys captured)