HO Demo Seed

The HO tenant (T0ADQ6UAVKR) is the primary sandbox for VERA development and smoke testing. A daily reseed keeps it in a known, clean state so each session starts from a predictable baseline.

What the daily reseed does

lib/demo-seed.js runs automatically at 8:50 AM ET every weekday (10 minutes before the 9 AM morning briefing fires). It:

1. Regenerates TSVs from generate-dummy-data/generate-v2.js (fresh random hours/budgets each run; names are stable).
2. Acquires a Postgres advisory lock to prevent concurrent runs.
3. Wipes all mutable tables for the HO tenant in FK-safe order: proposal_liabilities → proposal_roles → proposals (these must go first — proposal_liabilities.vendor_user_id hard-FKs users with no ON DELETE clause, so any surviving row blocks the users wipe and rolls back the entire seed), then conversations, time_entries, budgets, pending_pto, unresolved_items, liabilities, rate_history, project_tasks, projects (children then parents), users, tasks, clients, disciplines, company_calendar.
4. Re-inserts from TSVs in dependency order (disciplines → clients → tasks → users → projects → …).
5. Re-appends the three preserved real users with collision-safe IDs.
6. Seeds anchor data for the preserved users: Test Client, four anchor projects, time entries, envelopes, rate history, PTO requests, unresolved items, and a vendor scenario.

Scheduled by initDemoSeed(TENANTS) called at boot in index.js.

Manual reseed

DM VERA !test seed as an owner on the HO tenant. VERA replies with a summary of tables wiped and rows inserted.

If HO state gets messy mid-day, !test seed is the fastest recovery path.

---

Preserved real users

These three users are never wiped — their rows are read before the wipe and re-appended after.

Slack ID	Name	Role
U0AEJHN3MNV	Tom Hart	Owner
U0AKQ2EDWAU	MTH	Manager
U0AJTN12DPY	UTH	User

The generator produces a Tom Hart row with a fake Slack ID (U0000000001). That row is discarded during the merge; the real Tom Hart row (with the real Slack ID) takes its place.

---

The cast of characters

Generated users (15)

These users are produced by generate-v2.js on every reseed. Names are stable; budgets and hours vary.

ID	Name	Role	Type	Notes
1	Tom Hart	Owner	FTE	Placeholder — replaced by preserved Tom Hart
2	Rachel Kim	Owner	FTE	Second owner
3	Sarah Chen	Manager	FTE	PM archetype: consistent under-budget pace
4	Mike Johnson	Manager	FTE	PM archetype: aggressive pace, often over
5	Alex Kim	User	FTE
6	Nina Foster	User	FTE
7	Priya Patel	User	FTE
8	Lena Zhao	User	FTE
9	Chris Taylor	User	Contractor	20 h/wk
10	Marcus Webb	User	Contractor	20 h/wk
11	Jamie Park	User	Contractor	20 h/wk
12	David Okonkwo	User	Contractor	20 h/wk
13	Amy Walsh	User	Contractor	20 h/wk
14	Talia Rodriguez	User	Vendor	Sells deliverables; logs no time
15	Jordan Reed	User	Vendor	Sells deliverables; logs no time

Plus the three preserved real users (Tom Hart at real Slack ID, MTH, UTH) appended after the generated set.

Disciplines (6)

Six disciplines are seeded, each carrying a default_task_id that maps 1:1 to the corresponding task. This ensures every client time entry is categorised in the By-discipline segmentation view with zero Unmapped rows.

ID	Name	Default task
1	Design	Design (task 2)
2	Development	Development (task 3)
3	Strategy	Strategy (task 1)
4	Copywriting	Copywriting (task 4)
5	Project Management	Project Management (task 5)
6	QA & Review	QA & Review (task 6)

Task 7 (Admin) is internal-only and intentionally has no discipline — internal projects are excluded from segmentation.

Clients (11)

Name	Notes
Internal	Required infrastructure — always id=1
Meridian Health	Retainer client
Harbor Logistics	Retainer client
Bloom Retail	Fixed Price client
Pulse Analytics	T&M client
FreshFarm	Fixed Price client
CityCore	T&M client
Nexus Media	Fixed Price client
SkyBridge	T&M + Fixed Price client
Apex Fintech	Retainer client
Future Forward Fund	Pro Bono client

Generated projects (canonical shapes)

Name	Type	PM	Notes
Admin / Internal Meetings / New Business Development	Internal	Tom (id 1)	Overhead
PTO / Sick / Leave	Internal	—	Protected time-off projects
Meridian Health Support Retainer	Retainer	Sarah Chen	13+ completed periods + active Apr 2026
Harbor Logistics Platform Retainer	Retainer	Mike Johnson	Pacing over budget (demo alert trigger)
Apex Fintech Creative Retainer	Retainer	Sarah Chen
Bloom Retail Brand Refresh	Fixed Price	Sarah Chen
FreshFarm Brand Identity	Fixed Price	Sarah Chen
Nexus Media Campaign Suite	Fixed Price	Mike Johnson
SkyBridge Annual Report	Fixed Price	Mike Johnson
Future Forward Fund Impact	Pro Bono	Sarah Chen
CityCore App Development	T&M	Sarah Chen	Vendor liability (Talia Rodriguez, Design) — drives the By-discipline make-vs-buy face-off
Pulse Analytics Dashboard	T&M	Mike Johnson	Vendor liability (Jordan Reed, Copywriting) — drives the By-discipline make-vs-buy face-off
SkyBridge Website Redesign	T&M	Mike Johnson
Meridian Health Digital Support	T&M	Sarah Chen	Backbone engagement — spans the full 12-month window so the Monthly P&L never falls to $0 (the frozen retainers are stripped at seed)
Harbor Logistics Platform Support	T&M	Mike Johnson	Backbone engagement — full-window billable base, neutral delivery position

The three retainers above are generated but stripped at seed time (Retainer/Agile creation is frozen, #962), so they do not appear in the live demo. Their continuous monthly revenue is replaced by the two full-window T&M backbone engagements, whose logged hours are spread evenly across the window (not front-loaded) so every month shows believable, non-zero revenue.

Preserved-user anchor projects

Created by seedPreservedUserData after the bulk insert. These always exist post-reseed regardless of what the generator produces.

Name	Type	Purpose
Test Client	Client record	Parent for anchor projects
Test – Acme Retainer	Retainer	MTH PM; three burn states (Owner <50%, User healthy)
Test – Blue Website (FP)	Fixed Price	MTH PM; User over budget; vendor scenario
Test – Agile Sprint	Agile	MTH PM; User at warning burn
Test – Pro Bono Clinic	Pro Bono	Zero-cost project
Test – Internal Overhead	Internal	Shared overhead project for all preserved users

---

Verifying a reseed worked

Run these in Supabase SQL Editor after !test seed:

-- Confirm test cruft (Foo* projects) is gone
SELECT id, name FROM projects
WHERE tenant_id = 'e0c22d94-f990-5cd3-99b7-a4e2a2228147'
  AND name LIKE 'Foo%';
-- Expect 0 rows

-- Confirm cast is restored
SELECT COUNT(*) AS user_count FROM users
WHERE tenant_id = 'e0c22d94-f990-5cd3-99b7-a4e2a2228147'
  AND status = 'active';
-- Expect ≥ 18 (15 generated + 3 preserved)

-- Confirm anchor projects exist
SELECT id, name FROM projects
WHERE tenant_id = 'e0c22d94-f990-5cd3-99b7-a4e2a2228147'
  AND name LIKE 'Test – %';
-- Expect 4–5 rows (Acme Retainer, Blue Website FP, Agile Sprint, Pro Bono Clinic, Internal Overhead)

---

Safeguards

• seedDemoTenant returns { success: false } immediately if tenant.id !== 'T0ADQ6UAVKR'.
• !test seed in index.js checks tenant.id !== DEMO_TENANT_ID before calling the seed function.
• Advisory lock prevents concurrent runs (cron + manual).
• Every DELETE includes WHERE tenant_id = <HO UUID> as belt-and-suspenders beyond RLS.

---

Monitoring

Every successful seed run records a seed.daily event, visible on /status as "Demo seed (HO)". It follows the same weekday-only cadence as the morning/evening reminder checks — an amber reading early Monday morning (before the 8:50 AM ET run) is expected and self-heals once that run completes, not a failure to investigate.

---

Recovery if reseed fails

If !test seed fails mid-run, the database may be in a partially-wiped state. Re-run !test seed immediately — the wipe+insert is atomic (single transaction), so a failed run leaves the previous state intact or rolls back cleanly.

If state is unrecoverable, escalate to Tom for direct SQL intervention.