ELOSTATE — AI Executive Operating System

An AI Executive Operating System built under a constitutional discipline: the System must earn the right to speak before it asserts. Confident, fluent answers delivered before understanding is earned are the failure mode this project exists to defeat.

---

What ELOSTATE is

ELOSTATE is not a productivity tool and not an instant-answers dashboard. It is a system that holds open the diagnostic questions an executive needs to think about, surfaces evidence as it accumulates, and only ever offers suggestions in dialogue with the user — never as directives.

The governing document is the project constitution (private). Every architectural choice in this repo is enforcing one of its rules.

Core surfaces

Surface	What it does	Constitutional rule
Command Center	Surfaces today's open questions, uncertainties, and things worth noticing	§3.3 (guide, don't overtake) — questions, not actions
Operations	Task board + an "Awaiting evidence" panel that stays empty until the Understanding Gate clears	§3.2 (Understanding Gate is structural, not optional)
Team Intelligence	Workload + performance — scores are tagged demo when not derived from live signals	§3.4 (no fixed day-one behavior)
Finance / Marketing	Same shape as Operations — derived metrics, "Awaiting evidence" until the gate clears	§3.2
Conversation Dialogue	Paste a transcript, state YOUR read first, the System refines with explicit WHY	§3.3
Decision Dialogue	Four-phase conversation: Situation → Your read → System response → Decide & record	§3.3, §2 (explain WHY)

What it never does

• Assert a problem before signals support it (the Understanding Gate, encoded in the schema).
• Show a confident "AI diagnosis" before the user has stated their own read (the structural interrupt in every AI surface).
• Generate canned Safe/Balanced/Aggressive tiers that bypass the user's actual proposal.
• Recommend actions in the daily briefing — only surface questions and uncertainties.
• Loosen a rule for builder convenience without a ratified amendment.

---

Architecture

src/
  app/
    page.tsx                       # Landing
    login/                         # Real Supabase auth (demo-mode fallback)
    onboarding/                    # Writes companies + profiles
    dashboard/
      page.tsx                     # Command Center — Today's Open Questions
      operations/                  # Tasks + AwaitingEvidence panel
      team/                        # Per-member workload (demo-tagged)
      finance/                     # MRR/runway/expenses (demo-tagged)
      marketing/                   # Funnel/channels/campaigns (demo-tagged)
      conversations/               # 3-phase: Transcript → Your read → System
      decisions/                   # 4-phase: Situation → Read → System → Decide
      settings/
    api/
      ai/
        briefing/                  # → generateDailyQuestions (NOT a directive)
        conversation-dialogue/     # Guide-don't-overtake transcript analyzer
        decision-dialogue/         # Guide-don't-overtake decision flow
        analyze/                   # (legacy) — still violates §3.3 until propagated
        decision/                  # (legacy)
        conversation/              # (legacy)
        finance/, marketing/       # (legacy)
      decisions/                   # POST: persist dialogue + outcome
      seed/                        # POST: bootstrap demo data for a real company
  components/
    layout/                        # Sidebar (live user/company), TopBar
    ui/
      ScoreRing.tsx                # Honest — supports null + demo tag
      StatusBadge.tsx
      AwaitingEvidence.tsx         # The empty state that respects the Gate
  lib/
    claude.ts                      # Anthropic SDK wrappers — guide-don't-overtake
    mock-data.ts                   # Demo fixtures (clearly labeled in UI)
    supabase/                      # Browser + server clients, env-gate
    data/                          # Per-domain fetchers with mock fallback
  middleware.ts                    # Session refresh + route guard (no-op in demo)

supabase/migrations/
  0001_init.sql                    # Companies, profiles, tasks, decisions, RLS
  0002_understanding_gate.sql      # Signals, problems, gate trigger
  0003_decision_dialogues.sql      # Full dialogue persistence (immutable user fields)
  0004_events.sql                  # Append-only events — foundation under signals

docs/
  amendments/                      # Constitutional amendments (append-only audit trail)
  LOCAL_DEV.md                     # Windows/PowerShell/VS Code/Claude Code workflow
  UNDERSTANDING_GATE.md            # Design rationale for §3.2 enforcement
  GUIDE_DONT_OVERTAKE.md           # Implementation rule for §3.3 surfaces

scripts/
  check-setup.ps1                  # Local environment doctor
.vscode/                           # Recommended extensions + tasks + debug profiles

---

Two modes — demo and live

Mode	Trigger	Behavior
Demo	No Supabase env vars	Middleware short-circuits. Login/onboarding pass through. Dashboards read from mock-data.ts. ScoreRings show a demo tag. Banners are honest.
Live	NEXT_PUBLIC_SUPABASE_URL + NEXT_PUBLIC_SUPABASE_ANON_KEY set	Real auth, real onboarding writes, per-company data via RLS, Decision Dialogues persist with full WHY preserved.

Demo mode exists to let you preview UX without external services. Live mode is the real product — once configured, the Understanding Gate is enforced at the DB layer and cannot be bypassed by application code.

---

Local development

See docs/LOCAL_DEV.md for the full Windows / PowerShell / VS Code / Claude Code workflow. Quick start:

git clone https://github.com/<you>/TeamPilot.git
cd TeamPilot
npm install
Copy-Item .env.example .env.local
npm run setup:check              # color-checks env, ports, toolchain
npm run dev                      # http://localhost:4321

The dev server uses port 4321 (not 3000) to avoid collisions.

Useful scripts

npm run dev            # dev server with hot reload
npm run typecheck      # tsc --noEmit
npm run lint           # ESLint
npm run check          # typecheck + lint
npm run setup:check    # PowerShell env doctor

---

Going live (Supabase)

1. Create a free project at supabase.com.
2. In the SQL Editor, run each migration in order:
   • supabase/migrations/0001_init.sql
   • supabase/migrations/0002_understanding_gate.sql
   • supabase/migrations/0003_decision_dialogues.sql
   • supabase/migrations/0004_events.sql
3. Settings → API: copy URL + anon key into .env.local:

   NEXT_PUBLIC_SUPABASE_URL=https://xxxx.supabase.co
   NEXT_PUBLIC_SUPABASE_ANON_KEY=eyJ...
   

4. Settings → Authentication → Email: turn off "Confirm email" for local testing.
5. npm run dev and sign up — onboarding writes the first companies + profiles rows.

For AI features: also add ANTHROPIC_API_KEY=sk-ant-... from console.anthropic.com.

---

Deploying

Vercel: import the repo, add the same env vars in Project → Settings → Environment Variables, deploy. The dev port (4321) only applies locally; Vercel routes its own HTTPS port.

---

Constitutional discipline

Every change to this codebase is subject to the rules in the project constitution (private). Changes to those rules themselves go through the amendment process. Ratified amendments so far:

• AMD-001 — Establish the amendment process
• AMD-002 — Ratify Understanding Gate default thresholds (3 / 2 / 80)

If a rule conflicts with shipping faster, the rule wins. Speed that skips understanding is the failure mode this project was built to defeat.

---

License

MIT