platform-survey-2026-05-01

Bookstore OS / Indie Retail Platform — State Survey

Snapshot 2026-05-01. Sources cited inline by note filename. Vault is at docs/.

Platform Architecture

• Two production codebases plus one peripheral service:
  • Guild (guild/): Payload CMS 3.x + Next.js (App Router) + PostgreSQL + Redis. Source of truth for membership, loyalty, tiers, XP, check-ins. Live at dungeon.club. ([guild.md], [membership-platform.md])
  • Emporium (emporium/): Medusa v2 backend + Next.js storefront forked from Solace + Payload CMS for editorial + MeiliSearch + PostgreSQL 16 + Redis 7. Replaces the legacy Square Online (Weebly) shop. Phase 1 shipped 2026-04-29 as a 4-PR Graphite stack; 5 follow-on PRs in flight. ([emporium.md], [journal/2026-04-29.md], [journal/2026-04-30.md], [journal/2026-05-01.md])
  • Marty (marty/): Python 3.13 + FastAPI + Claude API + Hardcover (GraphQL) + Sinch SMS + Discord.py + PostgreSQL + Redis. SMS/Discord book-recommendation chatbot. “Development slowed.” ([references/marty.md], [references/projects.md])
• Hosting is uniform on Railway. Cloudflare for DNS/CDN. R2 for media.
• Multi-tenancy in Guild is shipped, not aspirational. Uses @payloadcms/plugin-multi-tenant with Organization as the tenant boundary (e.g. “Script Wizards” org owns “Dungeon Books JC” shop). Branch feat/multi-tenant-plugin, April 2026. Hierarchy: Organization → Shop → Member; Members are org-scoped, implicitly shop-scoped via tier. Per-tenant roles: admin vs staff; global super-admin vs user. Domain routing (hostname → shop → org) is planned, not implemented. Per-shop Square location IDs and per-shop Stripe price IDs (under one Stripe account per org) are configured. Points never transfer between shops by design. ([plans/guild-multitenancy.md])
• Cross-merchant identity layer is designed, not built. A schema in [references/cross-merchant-identity.md] proposes players, player_merchant_links, guilds, transactions, xp_events, achievements tables to unify Square’s per-location customer IDs under one player. This is the OutsideRPG network thesis — currently single-tenant in code.
• Agent layer: Marty is the only deployed AI agent; it is not in the auth chain. Authentik is the planned SSO hub (see Integrations). PostHog is wired for product analytics. ([plans/discord-integration.md], [references/guild-posthog-setup.md])
• Deployment topology: All services on Railway. Self-hosted infra (Cosmos box) is planned for Authentik, HedgeDoc, Kavita, Calibre Web, Nextcloud, Affine — exposed via Cloudflare Tunnel under *.dungeon.club. ([plans/discord-integration.md])
• Kiosk architecture: Currently a Next.js route group (kiosk)/ inside the Guild monolith. Decision logged to keep it in the monolith ([plans/guild-kiosk-architecture.md]). V2 design ([plans/guild-kiosk-v2.md]) decouples input/server/display via SSE + a PC/SC NFC daemon (Node nfc-pcsc on systemd) instead of HID-keyboard emulation. V2 is designed, partially implemented — triggered by an overnight Pi crash on 2026-04-09 that lost taps because the browser input field was unfocused.

Integrations

• Square POS (live, bidirectional state sync, no inventory sync yet for new stack):
  • APIs in use across projects: Catalog, Inventory, Orders, Payments (Web Payments SDK), Locations, Customers, Customer Custom Attributes, Terminal API (planned). ([references/square-integration.md])
  • Custom attributes pushed: guild_tier, guild_points, guild_status, guild_member_id — namespaced to avoid clobbering staff notes. Triggered on Stripe webhook (tier change), Square webhook (purchase points), redemption.
  • Square Loyalty deliberately not used. Decision logged 2026-04-02 ([references/square-loyalty-decision.md]). Reasons: API is dashboard-only for program config, one program per seller (breaks multi-tenancy), tier model is spend-thresholds not paid subscriptions, redemption restricted to order discounts.
  • Auto-discount at register is a known gap. Square does not expose an API to auto-create/update discounts; staff manually applies tier discount based on Square customer attribute. ([plans/auto-discounts.md])
  • Sync architecture for Emporium ↔ Square: in-process Medusa module (no separate service); ISBN as shared key (Square gtin ↔ Medusa variant sku); Redis 30s TTL for sync-loop prevention; nightly reconciliation job; dead-letter sync_log table. Not yet implemented in the new emporium stack — Phase 3 briefs were superseded 2026-04-29 by extended-catalog-and-preorders.md which narrows the sync to in-stock SKUs only. ([references/rebuild-spec.md], [plans/extended-catalog-and-preorders.md])
• Edelweiss Omnibus (relationship live; integration is CSV + manual):
  • Procurement brain. Carrie owns the workflow. Dave Lucey contact noted in [journal/2026-04-03.md].
  • Plan ([plans/extended-catalog-and-preorders.md], approved by Carrie 2026-04-29): customers can pre-order titles not on shelf; orders captured in Medusa, batched into Omnibus by staff; shipment received into Square; Medusa orders flip to “ready to ship.” Ingestion = Omnibus CSV → Medusa import script (manual/weekly). API integration deferred to “later” (when Bookshop.org API ships or Omnibus exposes data).
• Medusa v2 (live):
  • Phase 1 shipped 2026-04-29: Solace storefront ported, [countryCode] segment removed (US-only), Stripe checkout end-to-end with 3DS verified (PR #5), MeiliSearch instant-search via server proxy at /api/search (PR #3, no master key in client), Strapi → Payload swap (PR #4) with custom Lexical → React renderer (~150 LOC).
  • Follow-on PRs #10–13: real Open Library cover ingestion, related-products on blog posts, native Next sitemap + branded 404, fully Payload-driven footer (Footer global with address/hours/social/affiliate links). Established rule: anything staff would change without a deploy lives in Payload; Medusa is pure ecommerce. ([journal/2026-04-30.md])
  • Pinned: React 19 forced via pnpm.overrides; Tailwind kept at 3.4 because Medusa UI Preset is a v3 preset; meilisearch@0.57.0 exports lowercase Meilisearch.
  • Failure modes encountered: Solace seed script is non-idempotent (region creation collides on re-run); new products silently invisible to storefront if landed on Default Sales Channel instead of Online Store; Open Library has near-zero coverage for indie RPGs / zines (Mörk Borg, Mausritter — structural gap, no ISBN). One-shot idempotent scripts under backend/src/scripts/ are the workaround pattern (add-product-thumbnails.ts).
• Payload CMS (live in both projects):
  • Guild: Members, Tiers, Purchases, LoyaltyTransactions, CheckIns, Shops, Organizations, ActiveBuffs, Kiosks (9+ collections per [roadmap.md]).
  • Emporium: Pages, BlogPosts, Homepage global, Footer global. Dual-validation pattern (schema-side + render-side sanitizeUrl for hrefs). Lexical → React renderer is hand-rolled.
  • Auth incident: payload-auth plugin was incompatible with Payload 3.80.0 admin panel (form-state server action re-rendered on every keystroke, resetting inputs). Reverted. Current architecture is dual auth: native Payload for admin/staff (Users), Better Auth standalone for members (Members). Discord OAuth via Better Auth shipped in PR #87. ([plans/payload-auth-next-session.md], [plans/discord-integration.md])
• Twenty CRM: Nothing in vault.
• MeiliSearch (live in Emporium): Product index with Medusa price enrichment at proxy time. 150ms debounce + AbortController per-keystroke cancel. Subscribers auto-index on product CRUD. ([journal/2026-04-29.md], [references/rebuild-spec.md])
• Hi.Events (live SaaS, custom integration):
  • SaaS at app.hi.events. Auth via JWT (no long-lived API keys; refresh ~7 days). Endpoints used: POST /auth/login, /auth/refresh, GET /organizers/{id}/events, /events/{id}/attendees, /events/{id}/check_in_stats. Pagination is Laravel-style. Webhook integration uses HMAC-signed checkin.created and order.created; matched to Payload Member by email; hiEventsAttendeeId populated on first link. Webhook testing via Tailscale funnel. ([references/hi-events-api.md], [references/hi-events-webhook-testing.md], [plans/guild-hi-events-integration.md])
• Stripe (live): Subscriptions for tiers (Bronze $10,Silver$25, Gold $50,Mithril$100). Webhook handler updates tier/status, then fans out to syncMemberToSquare() and (planned) syncMemberToAuthentik(). 3DS verified 2026-04-29 in Emporium.
• Authentik (planned, not deployed): Will be the OIDC provider for selfhosted apps (HedgeDoc as POC at write.dungeon.club, then Kavita, Calibre Web, Nextcloud, Affine). Tier → group mapping (guild-silver, guild-gold, guild-mithril). Guild pushes tier directly via API; Discord is not in the auth chain. ([plans/discord-integration.md])
• Discord (live as login): Better Auth Discord OAuth (PR #87). discordId/discordUsername on Member. Marty bot exists for chat; a Phase 3 cosmetic role-sync via Marty is planned, optional.
• Idempotency / sync patterns observed:
  • Square webhook → unique constraint on square_transaction_id ([cross-merchant-identity.md]).
  • Hi.Events webhook → HMAC signature verification (env HI_EVENTS_WEBHOOK_SECRET).
  • Sync-loop prevention: short-lived Redis key (ISBN, ts) TTL 30s ([rebuild-spec.md]).
  • Check-in dedup: 4-hour window, dedup key varies by method (member+shop for nfc/phone/manual/purchase; member+qrCodeId for QR; member+externalEventId for events). ([references/guild-checkin-cooldowns.md])
  • Stripe webhook → fire-and-forget downstream syncs with console.warn on failure; nightly reconciliation for inventory drift > 2 units, auto-correct on 1-unit drift.

AI/Agent Layer

• Marty (Python/FastAPI). “Martinus Trismegistus” persona. Claude API for generation, Hardcover GraphQL for book metadata, Bookshop.org for affiliate links, Sinch for SMS, Discord.py for Discord. Status: SMS fully functional; Discord bot implemented, needs token for production. Stack matches the [plans/manapool-seller.md] microservice. Square-API in-chat purchase, analytics — not built. ([references/marty.md])
• PostHog (live in Guild): client init via instrumentation-client.ts; server via posthog-node singleton in src/lib/posthog-server.ts; /ingest reverse proxy to dodge ad-blockers. 12 events instrumented including signup_submitted, login_succeeded, subscription_activated, purchase_points_earned, points_redeemed. Six dashboards/insights built (signup→sub funnel, subscription health, tier upgrade funnel, loyalty activity, DAU). PostHog project ID 370533. ([references/guild-posthog-setup.md])
• AI DB access pattern documented but not confirmed deployed: [references/ai-db-access.md] (2026-04-06) prescribes a ai_readonly Postgres role with 30s statement_timeout, 5-conn limit, RLS for sensitive rows, MCP Postgres server for Claude. Whether currently provisioned: unclear from notes.
• Cernel evaluation ([journal/2026-04-23.md]): Concluded that AI product-data ingestion is the wedge for GTIN-less long tail (indie RPGs, zines, small-press) — not trade books. Trade books are owned by ONIX/Ingram/Hardcover incumbents. Image-gen explicitly rejected as “worse-than-useless” because indie RPG/zine art is core identity and the community rejects AI art.
• Indie ingest tool: Phase 1 designed (supplier registry + manual entry + URL/PDF/ISBN ingest + Claude enrichment for system, tropes, safety tools, content warnings, shelf-talker copy + Medusa+Meili write-through). Designed, not built. New plan file plans/indie-ingest-tool.md (renamed from rpg-zine-ingest-tool.md) referenced in the journal but not yet present in vault.
• Coding agent infra: Vault contains a CLAUDE.md instructing Claude Code to use Obsidian rather than its own auto-memory directory. PostHog wizard left an “agent skill folder.” No evals, no formal prompt library captured in vault.
• Eval approaches: Nothing in vault.

Multi-Tenancy Readiness

Guild’s data model is multi-tenant. The gap is everywhere downstream of the schema:

• Configuration-driven, today: organization, shop, kiosk API key, Square location ID, Stripe price IDs per tier, NFC card UIDs per member, Hi.Events organizer ID. ([plans/guild-multitenancy.md])
• Top 5 places Dungeon-specific assumptions are baked in (gleaned across notes):
  1. Tier definitions and benefits. Bronze/Silver/Gold/Mithril, $10/$25/$50/$100, 1×/1.5×/2×/3× multipliers, 100 pts = $1, are presented as global product decisions in [membership-platform.md] rather than per-shop config. Per-shop tiers exist in the schema but the marketing/portal copy and the iron/bronze/silver/gold/mithril enum are hardcoded.
  2. Brand/visual direction. “Cassette futurism + 80s sword & sorcery” for Guild and “Indie Publisher” for Emporium are baked into Tailwind preset (preset/plugins/colors.js), font choices, and component copy. The 2026-04-30 footer PR established the rule that “shop voice” should live in Payload globals — but most copy is still hardcoded.
  3. Domain routing. Each shop is supposed to have a domain field with Next.js middleware doing hostname→shop→org. Not implemented. Today everything resolves to dungeon.club / dungeonbooks.com. ([plans/guild-multitenancy.md])
  4. Email/notification templates. Resend module pattern from Agilo Fashion is the porting target; templates in [rebuild-spec.md] reference Dungeon Books address, hours, footer literally.
  5. NFC card format and kiosk hardware assumptions. WalletMate II in HID mode typing 0450CF01635713↵; NTAG215 cards; KIOSK_API_KEY + KIOSK_SHOP_ID env per device. Lite Kiosk (phone-as-reader) is the partner-shop scaling pattern but not yet built ([plans/guild-kiosk-v2.md]).
• Onboarding a new org is documented but manual: create Org in admin, create Shop with Square location, create Tiers with Stripe price IDs, create User with tenants: [{ tenant: orgId, roles: ['admin'], shop: shopId }]. ([plans/guild-multitenancy.md])

Observability & Ops

• Product analytics: PostHog (Guild). 12 instrumented events; 6 dashboards. ([references/guild-posthog-setup.md])
• Logs: Pino structured logging mentioned for Square sync ([references/square-integration.md]). Nightly reconciliation job emits structured logs.
• Traces / APM: Nothing in vault.
• Dashboards: PostHog only. Stripe dashboard for payment monitoring. No Grafana/internal dashboards in vault.
• Alerts: Nightly reconciliation alerts on inventory drift > 2 units (email/Slack — destination not specified). No on-call rotation, paging, or SLO docs.
• Runbooks: [references/hi-events-webhook-testing.md] is the only formal runbook. Tailscale funnel pattern recurs.
• Incidents handled:
  • 2026-04-09 Pi overnight crash ([plans/guild-kiosk-v2.md]): WiFi didn’t autoreconnect, Tailscale didn’t autostart, browser input lost focus → all NFC taps dropped until staff intervention. Triggered the V2 decoupled-input architecture. Mitigations: NetworkManager headless, systemctl enable tailscaled, Chromium kiosk auto-launch, NFC daemon as systemd service with Restart=always, hardware/systemd watchdog under consideration.
  • payload-auth admin panel breakage (Payload 3.80.0): documented and reverted. ([plans/payload-auth-next-session.md])
  • WSL Docker port-forwarding fight during MeiliSearch wiring: resolved by moving Meili call server-side. ([journal/2026-04-29.md])
  • Dark Reader extension hydration mismatches in dev (false-positive bug attribution).
• Square webhook reliability is flagged as tier-dependent ([extended-catalog-and-preorders.md] open question), with polling fallback planned.

Operator Persona & UX

• Operator persona (Carrie + retail staff): not engineers, but tech-comfortable. Carrie corrects architecture mid-conversation (caught the Omnibus-already-does-batching error in extended-catalog plan). Staff is expected to:
  • Use Square POS for walk-in sales and discount entry (manual tier discount lookup via guild_tier custom attribute on customer profile).
  • Use Payload admin to edit Footer, BlogPosts, Pages, Members, manually adjust points/tiers.
  • Approve member↔Square linking requests (anti-abuse) per [plans/guild-onboarding-rethink.md].
  • Plan Omnibus batches by scanning the “awaiting distributor” Medusa worklist.
  • Tap NFC cards into Payload admin’s nfcCardUid text field for v1 card registration.
• Day-to-day surfaces:
  • NFC kiosk (Pi + WalletMate II + touchscreen at 115 Brunswick St): idle/welcome/duplicate/not-found states. Welcome screen shows character sheet — name, tier, class, XP bar with tick-up animation, level, points + dollar value. 5s auto-return.
  • Member portal at dungeon.club ([plans/guild-portal-ui-implementation.md]): redesigned to RPG character sheet pattern — Character Sheet (/dashboard), Adventure Log (/dashboard/activity), Quest Log (/dashboard/quests), Rewards (/dashboard/rewards), Guild Hall (/dashboard/tiers). Sidebar nav with shadcn sidebar primitives.
  • Emporium storefront at dungeonbooks.com: shop, blog, staff picks (deferred), reading lists, events.
  • SMS / Discord chatbot Marty ([references/marty.md]): persona-driven, lowercase casual style; book recs.
  • Manapool seller microservice ([plans/manapool-seller.md]): personal MTG card sales — Gmail polling, TCGTracking labels, Discord webhook to #orders.
  • Staff dashboard ([references/guild-staff-dashboard.md]): three-layer design (Glance / Conversation / Admin) — designed, partially built. Glance card on check-in surfaces tier, points-as-dollars, subscription status, active buff, Square customer link.
  • Discord as community surface; Authentik will gate the selfhosted app suite (HedgeDoc/Kavita/CWA/Nextcloud/Affine) when deployed.
• POS gap is the kiosk’s reason for being. Square’s native register has no way to inject loyalty UI; the kiosk is the second-screen workaround. ([references/square-loyalty-decision.md])

Distribution & Community

• Jersey City canvassing plan ([plans/jc-canvassing-plan.md]): walk every event-hosting Square-using small business in JC. Target ≥3 of ~30 to pilot. Pitch is “real-world MMORPG with your shop as a quest hub.” Demo on phone, NFC card live tap. Known leads: Victory Point (warm), Sarap Shop (warm), Pixel Labs, Call Me Mia.
• Phil at Victory Point Cafe ([people/phil-victory-point.md] referenced; [plans/platform-pricing.md]): top pilot candidate. Software engineer, failed prior annual membership program. Pricing signal: would pay $100/mominimumforfullsuite,$500–1k/yr for allocation coordination alone, said $1k/yr is too low if value’s real (Bookmanager comp).
• Memory: project_phil-pilot-candidate, project_guild-traction (12 launch members, NJ pilot for 3 shops).
• Withfriends sunset (2026-04-23) ([journal/2026-04-23.md]): Carrie chose dungeon.club over Ko-fi migration. Withfriends being decommissioned within 2 days; existing paid terms honored. Public landing + tier cards + Stripe replacing invite-gating. Special migrations for Sean Buckley ($25\to$20 with $5 credit) and Allan (100% off coupon removal, 26th-of-month anchor).
• Facebook group of Square bookstore owners: not mentioned in vault.
• Edelweiss / Omnibus relationship: Dave Lucey/Omnibus call 2026-04-03 ([journal/2026-04-03.md]). No formal partnership doc.
• Indie RPG / zine supply thesis ([journal/2026-04-23.md]): Faire structurally can’t serve indie RPG/zine; Exalted Funeral is a retailer not a distributor. Bookcon validated: suppliers split between Faire and direct wholesale. Dungeon Books positioned to be the supply layer between them. Phase 3 trigger = Guild pilot shops asking to source from us → regional East Coast wholesale.
• Bookstore-developer peers in references: Kunal Gupta (Withfriends → pivoting toward Guild’s model per memory), Leonard Scheidemantel, Maggie Tang (Loyalist, fine dining vertical, $10M raised), Jason at Diesel Books.
• Press shipped ([references/expansion-strategy.md]): Shelf Awareness, Patch.com, JCity Times, NewPages, NJ.com / Jersey Journal, IndieBound Featured Bookstore (2026-02-18).
• Talks given: None captured in vault.
• Blog posts: Emporium blog system shipped 2026-04-29; “first blog post” listed as a Phase 4 launch task. No prior posts captured.

Open Questions & Tech Debt

Explicit “figure out later” / “won’t scale past N” flags in vault:

• Domain routing for multi-tenant. Designed, not built. ([plans/guild-multitenancy.md])
• SSO across dungeonbooks.com ↔ dungeon.club. Open. ([references/ecosystem-architecture.md])
• Tier discount automation at Square POS. Manual today; “breaks at scale (~40+ members).” ([plans/auto-discounts.md])
• Free-tier vs paid-tier point conversion. Free members can earn but not convert to store credit; carrot logic. ([plans/guild-onboarding-rethink.md])
• Auth identity vs store identity decoupling. Open: do we self-serve Square link with verification, or always staff-approve? ([plans/guild-onboarding-rethink.md])
• SSE single-process limit. “Fine for current scale. If we move to multiple server instances later, swap the in-memory map for Redis pub/sub.” ([plans/guild-kiosk-v2.md])
• Sitemap pagination. Current limit: 1000 per collection. Add when any of {pages, posts, products, categories} crosses ~500. ([journal/2026-05-01.md])
• Vercel SSE long-lived connection limits. Open. ([plans/guild-kiosk-v2.md])
• Apple Wallet / Google Wallet pass provisioning. Designed; pipeline TBD. ([plans/guild-kiosk-v2.md])
• Hi.Events SaaS limitations. Auth requires storing login credentials and refreshing JWT every ~7 days; no long-lived API key. ([references/hi-events-api.md]) Self-hosting flagged as “future option” in [references/ecosystem-architecture.md].
• Square inventory webhook reliability depends on Square subscription tier; polling fallback planned. ([plans/extended-catalog-and-preorders.md])
• Pre-existing typecheck errors from Solace port were cleared in commit 4217469 on 2026-05-01 — flag that “don’t merge a green PR on top of a red baseline.”
• Indie RPG cover-image gap. Open Library has near-zero coverage for itch/Free League/etc. Plan: drop covers in backend/seed-assets/covers/<handle>.jpg or upload via admin. ([journal/2026-04-30.md])
• Seed script non-idempotent. Region creation collides on re-run. Workaround: one-shot scripts under backend/src/scripts/ (template: add-product-thumbnails.ts).
• E2E in CI deferred. Three explicit thresholds for adding it: pre-launch (1 happy-path checkout test), second contributor (5-test smoke), or a regression that escaped. ([journal/2026-04-29.md], [journal/2026-05-01.md])
• Catch-all /[slug]/page.tsx for arbitrary Payload Page slugs is unsafe-by-omission today (sitemap advertises slugs that 404).
• Auth incident debt: payload-auth plugin removed; native Payload + Better Auth dual stack is the resolved-not-yet-stabilized architecture. ([plans/payload-auth-next-session.md])
• Contradictions in vault:
  • [references/projects.md] still lists store stack as “Saleor 3.22 · Payload 3.x · Next.js · Stripe” — outdated. The 2026-03-12 platform pivot in [references/rebuild-spec.md] and live work in [emporium.md] are on Medusa v2. projects.md needs a refresh.
  • [plans/membership-platform.md] non-goals say “Multi-location support… is a product concern for later,” but [plans/guild-multitenancy.md] (April 2026) shipped multi-tenant with org/shop/member. Non-goals doc is stale.
  • [plans/membership-platform.md] presents tiers (Bronze/Silver/Gold/Mithril, fixed prices) as global; [plans/guild-multitenancy.md] places Tiers under Shop. Reconciliation unclear.
  • [references/projects.md] lists webshop as Drizzle ORM; webshop is being sunset. [journal/2026-04-23.md] says the Withfriends sunset, not webshop, was happening that week — different system. Worth confirming webshop’s actual sunset date.

Business Model State

• Consumer-facing tiers: Bronze $10/Silver$25 / Gold $50/Mithril$100/mo. Iron is the internal label for non-member Square customers (no Payload record). Welcome gift, store discounts (0/10/20/25%), event discounts (10/10/15/50%), free monthly ARC at Silver+, yearly tote at Silver+, private event annually at Mithril, etc. Loyalty multipliers 1×/1.5×/2×/3×. Points = 1/$base,100pts=$1 store credit. Points never expire (no dark patterns, no breakage by design). ([plans/membership-platform.md])
• B2B platform pricing for partner shops ([plans/platform-pricing.md], current recommendation):
  • Small (<$250Krev):$99/mo
  • Medium ($250K–$750K): $179/mo
  • Large ($750K+):$299/mo
  • Comp: Bookmanager ~$7Kupfront/$2K/yr ongoing / 1% online transaction fee. Medium tier ($2,148/yr) is “right in that range.”
  • Switch to variable pricing when Dungeon Books becomes the tier-2 distributor (margin on inventory replaces subscription).
  • Napkin: 30 shops at $170/mo=$61.2k/yr (≈ one salary + infra); 100 shops = $216k/yr;500=$1.14M.
• Customer pricing experiments in journal:
  • 2026-04-23 Withfriends sunset: invite-gating removed; price drop from old $25Silvertonew$20 (Sean Buckley individually credited).
  • Allan: 100% off coupon removed, billing cycle anchor reset to 26th, $10 charge on 26th.
• Phil’s pricing signal ([plans/platform-pricing.md]): would pay $100/mominimumforfullsuite,$500–1k/yr for allocation alone, $1k/yr “too low if value is real.”
• Customer conversations / SOWs:
  • Wednesday call with Kunal Gupta 2026-04-01 — exit from WithFriends (CTO offer turned down per memory). [plans/wednesday-prep.md], [journal/2026-04-01.md].
  • JC canvassing: known warm leads above.
  • No formal SOWs / retainers in vault.
• Unit economics modeling: Napkin only. ~$300K/yrgamestoreatSmalltier=$1,188/yr; comparable to Bookmanager.
• Memory items relevant: project_brian-buyout (Brian leaving Leica mid-May 2026 — buyout vs cofounder pending, retail + game producer background), project_guild-priority (Guild is Panat’s founder bet, chosen over WithFriends CTO), user_panat-context (E-2 visa cannot lead nonprofit; long-term: vertical integration for indie bookstore industry).
• Nonprofit / capital strategy ([references/expansion-strategy.md]): Lease ends July 2027 (16-month deadline). Dungeon Arts 501(c)(3) targeted Q2 2026 (Carrie leads). Funding landscape mapped: JCACTF (up to $30koperating,$17.5k program), Hudson County LAP, NJ State Council GOS/GPS (3-year operating), NJEDA A.R.T. (Phase II disbursed; Phase III TBD), NJ Cultural Trust. Most require 501(c)(3) status.
• Research lab framing ([plans/research-lab.md]): The platform work is positioned as the commercial application of network-coordination-dynamics research. “5 people with agents operate like 30.” Phase 0 is now (lab=1, Guild=first tool, bookstore=first dataset). Post-WithFriends-exit (late 2026) → 2–3 people. Mana/Jimmy Lee called out as potential aligned funder.

---

Five Clarifying Questions

1. Scope of “platform”: Are you treating Marty (SMS/Discord chatbot), the Manapool seller microservice, and the planned RPG/zine ingest tool as part of the bookstore-OS surface area for the strategy conversation, or only Guild + Emporium? The vault treats them as adjacent but separately deployed.
2. Tenant boundary intent: Multi-tenancy is shipped at the Org→Shop level, but tier definitions, tier copy, brand/visual direction, and most operator UX are still hardcoded as Dungeon Books. For the strategy conversation, should I be assuming the next pilot shop runs as another Org under your Stripe account (per [guild-multitenancy.md] onboarding steps) or as a fully separate deployment? The plan accommodates both but they imply very different roadmaps.
3. Authentik / selfhosted-app suite is designed but not deployed. Is that POC live yet, or should I treat the entire OIDC tier-gating layer (HedgeDoc, Kavita, etc.) as “designed only” for the strategy conversation?
4. Withfriends sunset and Guild PMF question ([journal/2026-04-23.md]) flagged Guild as “largely a novelty” pending data analysis to separate migration-failure from PMF-failure. Has that activity-data pull happened? It changes how to read the 12-member traction signal.
5. The cross-merchant identity layer in [references/cross-merchant-identity.md] is the most architecturally consequential piece in the vault that hasn’t been built (player schema as primary entity, not loyalty record). Is that the next platform-tier build, or is it parked behind the indie-supplier ingest tool ([journal/2026-04-23.md] Phase 1) and continued Emporium launch work?