. Generates a floating chat bubble in bottom-right of every page on their site. Bubble opens to chat interface with DCS branding. Auto-detects site URL on first load and queues content training (Phase 2). Same script works on DCS-built sites (auto-installed) and external sites (manual install). ~2 days for the script + CDN distribution + auto-config.' },
{ title: 'Auto-train on website content', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'When a site is connected to DCS Assistant, automatically crawl all pages, extract text content (excluding nav/footer boilerplate), chunk into 500-token segments, generate embeddings via OpenAI text-embedding-3-small, store in pgvector table. At chat time, embed the user question, semantic search for top 5 relevant chunks, include as context in Claude/GPT call. RAG pipeline standard but tuned for small business sites (10-50 pages typical). 1 week for crawler + chunking + embedding pipeline + retrieval.' },
{ title: 'Auto-train on business profile + hours', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Beyond content, train on structured business data from the user\'s DCS Business Info tab — name, address, phone, email, hours, services list, pricing. This data is more reliable than parsing website text. When user asks "what are your hours" or "what services do you offer", Assistant pulls from structured data, not RAG. Faster and more accurate than RAG for these common questions. 3 days for the structured data integration + prompt engineering.' },
{ title: 'Lead capture form integration', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'When Assistant detects high purchase intent ("I want to book", "how much for X service"), proactively show inline form to capture name/email/phone. Form submission routes to DCS Inbox + WhatsApp notification to business owner. Conversion path: visitor chats → asks question → gets answer → Assistant offers "want me to have someone call you?" → form → owner gets WhatsApp lead. ~2 days for intent detection + form UI + routing.' },
{ title: 'WhatsApp handoff for live questions', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'When Assistant cannot answer confidently (low confidence score, complex question, escalation request), offer handoff to live WhatsApp chat with business owner. Click "Talk to a human" button → opens WhatsApp deep link to owner\'s number with conversation context pre-filled (visitor question + Assistant\'s attempt). Owner gets context and continues conversation. Integrates with existing WhatsLink infrastructure. 3 days for handoff trigger logic + WhatsApp message formatting.' },
{ title: 'Multilingual (Hindi / Arabic / English)', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Critical for UAE + India target markets. Auto-detect visitor language from browser settings, respond in that language. Supported: English (default), Hindi (Devanagari script), Arabic (RTL layout), with Tagalog and Bahasa as Phase 2. Underlying LLM (Claude/GPT) handles language switching natively — just need UI infrastructure (RTL CSS, font fallbacks for Hindi/Arabic). Prompt engineering to instruct LLM "respond in {detected_language}". 1 week for UI + prompt work + testing across languages.' },
{ title: 'Per-site conversation analytics', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Show business owners how their Assistant is performing. Dashboard in /app22 Assistant tab: total conversations this week, most-asked questions (categorized), conversion rate (chats that led to lead capture), average response satisfaction (visitor can rate after chat), unanswered questions list (what owner should add to their site content). Drives upsell — owners see value, more willing to pay for premium tier. 3 days for analytics queries + dashboard UI.' },
{ title: 'Pricing tier ($19/mo per site)', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Standalone pricing model: $19/mo per site for DCS Assistant. Or bundled free in DCS Business OS / Vertical Suites at higher tier. Tier limits: 500 conversations/month included, $0.05 per additional conversation (matches token cost + margin). For non-DCS sites (paying $19 directly), same limits. Enterprise tier $99/mo unlocks: white-label (remove DCS branding from chat bubble), custom AI model training, API access to chat transcripts. 1 day for Stripe products + tier enforcement.' },
],
tech: {
stack: 'Anthropic Claude API or GPT-4 (via OpenAI compatible) · pgvector for RAG · WhatsApp Business API for handoff',
hosting: 'Embed script served from CDN · Chat backend on Railway · Embeddings on Supabase pgvector',
db_tables: 'assistant_sites · assistant_chunks · assistant_conversations · assistant_messages',
notes: 'Premium addon ($19/mo standalone) or bundled into Business OS / Vertical Suites at higher tier. Multilingual support is critical differentiator for UAE/India markets.',
},
},
{
id: 'p-agents-api', project: '02 · DCS CORE PRODUCTS', name: 'DCS Agents API (Industry Multi-Agent)',
kind: 'product', priority: 'low', launchReady: 'no',
desc: 'Sell the 641-agent industry-specific system as B2B API. Other AI companies use DCS vertical agents instead of building their own. Restaurant agent · Salon agent · Clinic agent etc. Per-invocation pricing. Different value capture from DCS Platform (which sells the full builder UX).',
items: [
{ title: 'Public API spec (auth + agent invocation)', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'OpenAPI 3.0 spec for the Agents API exposed at api.dcsai.ai/agents/v1. Endpoints: POST /agents/:agent_id/invoke (call any of the 641 agents with prompt + context, get structured output), GET /agents (list all agents with descriptions + pricing), GET /agents/:agent_id (single agent details). Auth via API key in header (X-DCS-API-Key). Rate limits per tier enforced via middleware. Documentation auto-generated from spec via Swagger UI. 1 week for spec + endpoint implementation + auth layer.' },
{ title: 'Agent catalog page (all 641 listed)', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Public marketing page at agents.dcsai.ai showing all 641 agents organized by industry vertical (Restaurant agents, Salon agents, Clinic agents, etc.). Each agent has: name, what it does (one sentence), example input → output, pricing per invocation, "Try it free" sandbox link, copy-paste API code snippet. Filterable by industry, complexity, pricing. Critical for SEO — programmatic page per agent could rank for niche queries like "restaurant menu AI agent". 3 days for the page + SEO setup + per-agent detail pages.' },
{ title: 'Per-agent invocation pricing model', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Each agent has a base price per invocation based on its underlying LLM cost + complexity. Simple agents (e.g., "format business hours") = $0.001 per call. Complex agents (e.g., "generate complete restaurant menu") = $0.05 per call. Premium "moat" agents (e.g., "Salon OS booking optimizer") = $0.20 per call. Price is base × multiplier per tier (Starter 1.5x, Growth 1.2x, Pro 1x — Pro gets wholesale, others have markup). 1 week for pricing table per agent + billing integration.' },
{ title: 'Rate limits + quotas per tier', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'API rate limits enforced by tier: Starter (1000 calls/day), Growth (10k calls/day), Enterprise (custom, typically 100k+ calls/day). Limits prevent abuse and ensure fair use. Customers hitting limits can upgrade or contact sales for custom tier. Rate limit headers (X-RateLimit-*) included in every response for client-side awareness.' },
{ title: 'SDK (Python / Node / cURL examples)', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Multi-language SDKs for developer adoption. Python: `pip install trd-agents`, idiomatic Python with async support. Node: `npm install @trd/agents`, TypeScript types. cURL: copy-paste examples for every endpoint. Each SDK includes: auth setup, agent invocation, streaming support (for long-running agents), error handling, retry logic. Examples in docs for common patterns ("call restaurant_menu_agent then format_for_pdf_agent"). 1 week for both SDKs + examples + npm/pip publishing.' },
{ title: 'Documentation site', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Developer docs at docs.dcsai.ai/agents. Sections: Quickstart (call your first agent in 5 minutes), Agent reference (auto-generated from agent metadata), SDK reference (Python + Node), Concepts (what is an agent, chaining, error handling), Examples (real-world recipes), Pricing, Rate limits, Status page. Built with Mintlify or similar docs framework. Search functionality across all agents. 3 days of writing + deployment.' },
{ title: 'First 5 B2B customers', status: 'upcoming', effort: 'M6+', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Validation milestone — sign 5 B2B customers using the Agents API. Target customer profile: AI app developers building vertical SaaS in industries DCS already supports (e.g., a developer building a restaurant POS app who needs menu-generation AI). They get to use our agents instead of building their own. Acquisition: outbound to relevant founders on Twitter, Indie Hackers, and AI hackathons. Pricing leverage: free tier (100/day) gets developers building, conversion to Starter at $49 when usage exceeds free. Month 6+ because gated on DCS Platform proving the underlying agents at scale first.' },
],
tech: {
stack: 'Existing 641-agent system (agent_orchestrator.ts) exposed via REST API · Auth + billing middleware · Redis for rate limits',
hosting: 'api.dcsai.ai on Railway · CloudFlare for DDoS protection',
db_tables: 'api_keys · api_usage_log · agent_pricing_tiers',
notes: 'Different value capture from DCS Platform — Platform sells the full UX to non-technical users, Agents API sells the underlying capability to other developers. Both monetize the same 641-agent moat.',
},
},
{
id: 'p-saas-builder', project: '02 · DCS CORE PRODUCTS', name: 'DCS SaaS Builder',
kind: 'product', priority: 'low', launchReady: 'no',
desc: 'Future evolution beyond websites. Build full SaaS apps with database, auth, billing, dashboards. Like Bubble.io + AI. "Describe your SaaS in a paragraph → working MVP in 10 minutes." YEAR 2 territory — needs DCS Platform stable + Memory layer + Inference at scale.',
items: [
{ title: 'Database schema generation from prompt', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'The hardest technical problem in SaaS builder. User describes app in natural language ("I want to build a CRM for plumbing companies"), agent analyzes intent, generates appropriate Postgres schema: tables (customers, jobs, technicians, invoices), columns with proper types, foreign keys, indexes, RLS policies. Validates schema for sensible relationships (no orphan tables, no missing FKs). Migration files generated and applied to fresh Supabase project per user app. 1 month because schema design is genuinely hard and getting it wrong cascades through everything else.' },
{ title: 'Auth flow (signup/login/forgot-password)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Every SaaS needs auth. Generate complete auth flow: signup page (email + password), login page (with magic link option), forgot password flow (Resend email + reset token), email verification, profile page, password change. Built on Supabase Auth as the backend (gives social login support, MFA option, password hashing for free). Frontend uses agents to generate components matching the rest of the user\'s app styling. 2 weeks for full flow + UI consistency + edge case handling (verification expiry, etc.).' },
{ title: 'Billing integration (Stripe Connect)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'For SaaS apps that need to charge their end users. Stripe Connect platform model — DCS has the master Stripe account, user\'s SaaS uses Stripe Connect to bill their own customers. User defines their pricing tiers in DCS\'s UI, system generates Stripe products, integrates checkout into the SaaS, webhook handlers update subscription status. DCS takes 1% platform fee on every transaction processed by user\'s SaaS. 2 weeks for Stripe Connect onboarding flow + checkout integration + webhook handlers.' },
{ title: 'Dashboard UI generation', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Generate the internal dashboard UI for the SaaS owner — view their data, run reports, manage users. Different from the customer-facing pages. Components: data tables (with sort/filter/search/pagination), CRUD modals, chart widgets, KPI tiles, navigation sidebar. Agents map from database schema to dashboard pages — e.g., a "customers" table gets a /customers page with searchable table + detail view + create/edit forms. 1 month because dashboard generation requires understanding intent (which fields matter, what visualizations help).' },
{ title: 'Multi-page app routing', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'SaaS apps have many pages — dashboard, settings, billing, profile, plus app-specific (customers list, customer detail, jobs board, etc.). Need proper routing: client-side React Router or Next.js routing with code-splitting per route, dynamic routes (/customers/:id), nested routes (settings → profile, settings → billing), auth guards on protected routes. 2 weeks for routing architecture + agent prompts to generate correct route definitions.' },
{ title: 'Deploy to DCS Hosting', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Built SaaS apps deploy to DCS Hosting (Vercel under the hood currently). Each SaaS gets its own subdomain (.trdn.app) or custom domain. Includes backend API routes (serverless functions on Vercel), database (Supabase project), file storage (Supabase Storage). One-click deploy from /app22 SaaS builder. Environment variables auto-configured (Supabase keys, Stripe keys, etc.). 1 week for the deploy orchestration + env var management.' },
{ title: 'Pricing tier (premium-only)', status: 'upcoming', effort: 'YR2', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'DCS SaaS Builder is premium-only. Pricing: $199/mo Builder tier (1 SaaS, 1000 users), $499/mo Studio tier (5 SaaS, 10k users), $999/mo Agency tier (unlimited SaaS, unlimited users + white-label). Higher pricing reflects: (1) much more complex output than websites, (2) ongoing hosting costs (databases, backends), (3) target customer is technical product builder with bigger budget than salon owner. Year 2 launch because needs DCS Platform stable + Memory + Inference all proven at scale first.' },
],
tech: {
stack: 'Existing agent system + new database agents + Supabase Auth + Stripe Connect + React Router · Year 2 architecture',
hosting: 'DCS Hosting (Vercel-backed today, own infra Year 2) · per-app Supabase projects',
notes: 'YEAR 2 product — depends on DCS Platform reaching 1000+ stable customers first. Until then, SaaS Builder is on the backlog.',
},
},
{
id: 'p-mobile', project: '02 · DCS CORE PRODUCTS', name: 'DCS Mobile App',
kind: 'product', priority: 'low', launchReady: 'no',
desc: 'Mobile app builder. Generate iOS + Android apps from one prompt. Like FlutterFlow + AI. Use cases: restaurant ordering apps, salon booking apps, fitness coaching apps. Cross-platform via React Native or Flutter under the hood. YEAR 2.',
items: [
{ title: 'React Native (or Flutter) code generation', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Foundational decision: React Native vs Flutter for code generation target. React Native pros: shared JS/TS ecosystem with web (DCS already does), Expo simplifies tooling, larger community. Flutter pros: better default UI performance, single codebase truer (one Dart codebase), beautiful default widgets. Likely React Native because we already understand the JS ecosystem and can share agent training data with the web frontend agents. 1 month to retrain frontend agents on RN patterns + generate working sample apps.' },
{ title: 'Mobile-specific UI patterns library', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Mobile apps need fundamentally different UI than websites: bottom tab navigation, swipe gestures, pull-to-refresh, infinite scroll, modal sheets, system-native action sheets. Need to encode these patterns in our agents so they generate correct mobile UX, not just mobile-resized web UI. Pattern library: 30-50 reusable components (tab bars, list rows, cards, modals, forms, etc.) styled for iOS + Android conventions. 1 month for library + agent prompts.' },
{ title: 'iOS App Store submission flow', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Apple\'s submission process is notoriously painful — code signing, provisioning profiles, App Store Connect setup, screenshot requirements, privacy policy URLs, age ratings, review feedback cycles. DCS will hide this complexity: user finishes app, clicks "Submit to App Store", DCS handles signing (with user\'s Apple Developer account credentials), builds IPA, uploads to App Store Connect, fills metadata from app prompt, generates screenshots from app preview, submits for review. 2 weeks for the automation + Apple developer account flow.' },
{ title: 'Android Play Store submission flow', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Google Play submission is simpler than Apple but still complex: keystore management, app bundle (AAB) generation, Play Console listing, content ratings, store listing copy + graphics. DCS handles all of this similarly to iOS — user clicks "Submit to Play Store", DCS generates everything from app context, uploads, monitors review status. Bonus: Play Store reviews are typically 24-48 hours vs Apple\'s 1-7 days. 2 weeks for automation.' },
{ title: 'Push notification setup', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Push notifications are critical for engagement in mobile apps. Configure Firebase Cloud Messaging (FCM) for both iOS and Android, generate certificates, integrate React Native push notification library, provide API in DCS for sending notifications ("notify all users who haven\'t opened the app in 7 days"), template library for common notification patterns. 1 week for the cross-platform setup + DCS-side notification API.' },
{ title: 'In-app payments (Apple / Google billing)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'For apps selling digital goods (subscriptions, in-app purchases), MUST use Apple/Google billing (30% commission) — required by App Store rules. For physical goods/services (restaurant orders, salon bookings), can use Stripe (~3% commission). DCS agents distinguish based on what the app sells and integrate the correct billing system. For digital: react-native-iap library for both stores. For physical: existing Stripe integration. 2 weeks for both paths + agent prompt updates to choose correctly.' },
{ title: 'Pricing tier (premium-only)', status: 'upcoming', effort: 'YR2', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'DCS Mobile is premium-only, even higher than SaaS Builder. Pricing: $499/mo Builder (1 app, basic features), $999/mo Studio (3 apps, push notifications, in-app payments), $1999/mo Agency (unlimited apps, white-label, custom domains, agency support). Higher because: (1) ongoing Apple/Google developer account costs ($99/year + $25 one-time), (2) more complex submission/review cycles needing DCS support, (3) target customer is established business willing to pay for native mobile presence. Year 2 launch.' },
],
tech: {
stack: 'React Native + Expo (likely) · Firebase Cloud Messaging · Apple App Store Connect API · Google Play Console API',
hosting: 'Codepush for OTA updates (Expo Updates) · Native binaries hosted on App/Play Stores',
notes: 'YEAR 2 product — biggest engineering lift in the roadmap. Could spawn off as separate team eventually.',
},
},
{
id: 'p-sentinel-pro', project: '02 · DCS CORE PRODUCTS', name: 'DCS Sentinel (Code Quality)',
kind: 'product', priority: 'low', launchReady: 'no',
desc: 'Productize the Sentinel protection system as "AI Safety for Solo Founders". 10 protected zones in agent_orchestrator.ts (16,156 lines), 60-second Railway red on tampering, [DCS-EDIT] handshake protocol. Internal version in production since May 9. Sold to indie devs paranoid about AI agents breaking prod.',
items: [
// Internal Sentinel - shipped
{ title: 'Internal Sentinel — 10 protected zone markers in agent_orchestrator.ts', status: 'completed', effort: 'in production', owner: 'Cowork', addedDate: '2026-05-09', phase: 'Foundation', testResult: 'pass', note: 'Internal Sentinel system shipped May 9, 2026. Ten protected zone markers placed at critical points in agent_orchestrator.ts (16,156 lines) — each marker is a comment block following a specific format. Railway CI/CD pre-deploy hook scans for required marker count; if any missing, deploy fails within 60 seconds and Railway turns RED. Prevents accidental deletion of critical orchestration logic via the [DCS-EDIT] handshake protocol requirement.' },
{ title: 'WL-3 import marker @ line 7', status: 'completed', effort: 'shipped', owner: 'Cowork', addedDate: '2026-05-09', phase: 'Foundation', note: 'Sentinel marker at line 7 protecting the WL-3 (WhatsLink-3) import block. This import is critical because WhatsLink form injection runs on every DCS-built site. Without this import, all built sites lose the Reserve a Table / WhatsApp widget functionality. Marker comment includes phase reference + must-not-modify warning.' },
{ title: 'Phase 43.3 trackVendorUsage marker @ line 40', status: 'completed', effort: 'shipped', owner: 'Cowork', addedDate: '2026-05-12', phase: 'Foundation', note: 'Sentinel marker at line 40 protecting Phase 43.3 trackVendorUsage function (commit 568c8ea, May 12). Function logs every Cerebras API call to vendor_usage_events table for cost tracking. Removing this breaks all vendor spend visibility going forward. Marker prevents accidental deletion during refactors.' },
{ title: 'Phases 42.14-42.27 markers', status: 'completed', effort: 'shipped', owner: 'Cowork', addedDate: '2026-05-10', phase: 'Foundation', note: 'Sentinel markers protecting Phase 42.14 through 42.27 fixes — industry guards, nav filters, contrast guards, gallery deduplication. All shipped May 9-10 as part of the bug cleanup track. Markers prevent regression: if a future change deletes the fix, Sentinel catches it within 60 seconds via Railway deploy check.' },
{ title: '[DCS-EDIT] handshake protocol enforced', status: 'completed', effort: 'in production', owner: 'Cowork', addedDate: '2026-05-09', phase: 'Foundation', note: 'Critical protocol: Cowork (WhatsApp agent) must send [DCS-EDIT] message to founder BEFORE AND AFTER every push to main that touches sentinel-protected files. Message format: file path, line range affected, commit SHA, summary of change. 138 handshakes recorded in Cowork session history. Protocol prevents Cowork from silently modifying critical code paths. Founder reviews handshake messages and can intervene if change looks risky.' },
{ title: 'Railway redirect alarm (60s RED on missing marker)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-09', phase: 'Foundation', note: 'Railway pre-deploy hook script scans agent_orchestrator.ts for required marker count (10). If count != 10, exit code 1 triggers deploy failure within 60 seconds. Railway dashboard shows red status, email alert sent to founder. Recovery procedure: identify what was deleted via git diff, restore via git, push fix. This is the safety net that prevents silent disasters from AI agent edits.' },
// Productization
{ title: 'Open-source Sentinel core', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Phase 1', note: 'Productization Phase 1: open-source the core Sentinel logic (marker checking, pre-deploy hook scripts) as an MIT-licensed npm package. Goal: get indie devs using it on their own projects, validate the value, build community trust. The hosted/managed version (Phase 2) becomes the monetization layer on top of free open-source core. Strategy: same playbook as Sentry/PostHog/Plausible — open-core, paid hosted.' },
{ title: 'Hosted Sentinel dashboard for paying customers', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Phase 2', note: 'Productization Phase 2: managed Sentinel dashboard for paying customers. Shows: protected zone status across all monitored repos, alert history, marker change audit log, integration with GitHub/GitLab for auto-PR-blocking. Pricing: $19/mo individual, $99/mo team, $499/mo enterprise. Target: indie devs and small teams using AI coding agents (Cursor, Copilot, Aider, Devin) who want safety net against agent screw-ups.' },
{ title: 'GitHub Actions integration', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Phase 2', note: 'GitHub Actions workflow that runs Sentinel marker check on every PR. Blocks merge if markers missing or modified without handshake. Lower friction than running a separate service — most teams already use GitHub Actions. Comes pre-configured: install via marketplace, point at your repo, done in 5 minutes. This is the standalone product version of DCS\'s internal Railway pre-deploy hook.' },
{ title: 'Custom alert channels (WhatsApp / email / Slack)', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Phase 2', note: 'Multi-channel alerts when Sentinel detects an issue: WhatsApp (primary for solo founders, mobile-friendly), email (everyone), Slack (teams), Discord (open-source communities), Telegram (international users). Each channel configurable per repo + per severity. Premium feature — included in $99/mo team tier and above.' },
{ title: 'Per-protected-zone pricing model', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Phase 3', note: 'Pricing strategy: charge per protected zone monitored, not per repo or per user. Logic: more zones = more value delivered = higher willingness to pay. Tier examples: $9/mo for 5 zones (individual), $29/mo for 25 zones (small team), $99/mo for 100 zones (team), $499/mo for unlimited zones + SLA (enterprise). Aligns price with value delivered.' },
{ title: 'First 10 paying customers', status: 'upcoming', effort: 'M6+', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Growth', note: 'Month-6+ target: 10 paying Sentinel customers = ~$1k MRR. Acquisition: Show HN launch for open-source version, content marketing about AI-agent safety, integration with popular AI coding tools (Cursor extension, Aider plugin), word-of-mouth from DCS Platform users who experienced the internal benefits. Slow ramp expected — niche product.' },
],
tech: {
stack: 'Marker comments in agent_orchestrator.ts (16,156 lines) · Railway deploy check + alarm webhook',
hosting: 'Internal in production · Future: SaaS dashboard at sentinel.dcsai.ai',
notes: 'Cowork file shows 169 sentinel references and 138 [DCS-EDIT] handshakes — the protocol works at scale',
},
},
{
id: 'p-teams', project: '02 · DCS CORE PRODUCTS', name: 'DCS Teams',
kind: 'product', priority: 'medium', launchReady: 'no',
desc: 'Multi-user collaboration tier on DCS Platform. Multiple seats per account, role-based access (owner / editor / viewer), shared workspace, comment threads. Targets agencies + small teams building sites for clients. Direct lead-in for DCS Agency White-Label product.',
items: [
{ title: 'Multi-user authentication', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Foundation for team support: allow multiple users to access the same DCS account. Implementation: extend Supabase auth.users with organizations + memberships tables. Each user has email + password; each organization has multiple users. Single sign-on between DCS Platform, Compute, Storage. Required before any team features can ship. Estimated 3-5 days work.' },
{ title: 'Role-based access control (RBAC)', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Different team members get different permissions: Owner (everything), Admin (everything except billing + delete account), Editor (build sites, edit content, can\'t delete), Viewer (read-only). Critical for agencies and small teams to share access without granting full control. Implementation: roles table joined via memberships, permission check middleware on every protected route.' },
{ title: 'Workspace / team concept in data model', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Refactor the data model: every site, asset, billing record gets owned by an organization (team workspace), not directly by a user. User belongs to organization via membership. Allows team transfer (transfer ownership of all sites from one user to another), team billing (pay once for whole team), team analytics (aggregate metrics across team\'s sites). Large refactor — high effort but foundational.' },
{ title: 'Invite + onboarding flow for team members', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Owner invites team member via email — email contains signed token, recipient clicks → creates account or logs in → joins the organization with specified role. Onboarding: quick tour highlighting their role\'s permissions, where to find what, how to ask for elevated permission. Inspired by Linear/Notion invitation flows.' },
{ title: 'Comment threads on builds', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Team collaboration feature: any team member can comment on a specific build, page, or section. Threaded replies. @-mentions notify via email/in-app. Resolved/unresolved state. Critical for agencies reviewing client work before publishing. Inspired by Figma/Notion comment patterns.' },
{ title: 'Activity log per team', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Audit log of all team actions: who built which site, who edited what, who deleted what, who invited whom, who changed billing. Searchable + filterable by user, action type, date range. Critical for: post-mortem when something goes wrong, security audits, accountability in larger teams. Stored in audit_log table with 1-year retention.' },
{ title: 'Per-seat pricing model', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Pricing: base subscription (Pro $99/mo) + $19/seat for additional team members. First seat included free. Discourages share-one-account behavior while remaining affordable for small teams. Annual billing offers ~15% discount. Enterprise tier offers volume discounts for 20+ seats.' },
],
},
{
id: 'p-pilot', project: '02 · DCS CORE PRODUCTS', name: 'DCS Pilot (Autonomous Layer)',
kind: 'product', url: 'https://pilot.dcsai.ai', priority: 'high', launchReady: 'partial',
desc: 'Autonomous software for small-business work — vertical, specialized, one-command. Built on the DCS stack (Platform + Compute + Storage + OS), packaged as a single autonomous experience. The destination this whole architecture was pointed at. Marketing site shipped May 14 2026 as full product-detail waitlist page. Substrate audit (May 14 2026) found most of Layer 1 already running in production: memory + RAG, 5-tier inference, permanence + audit logs, multi-channel inbox, Predictive Outcome Engine in observability mode. Phase 1 of Pilot = productization step (wrap running substrate into one branded experience), not invention. Five layers total: Layer 1 persistence (substrate live), Layer 2 command center (Q3-Q4 2026 compressed), Layer 3 vertical computer use (Q1 2027, only genuinely new build), Layer 4 predictive (Q4 2026, activation not invention), Layer 5 vertical OS (Salon/Clinic Q3 2026 per platform roadmap, rest H2 2027).',
items: [
{ title: 'Layer 1 · Memory layer with RAG retrieval', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-04-18', phase: 'Layer 1', note: 'agent_memory.ts shipped April 18 2026 as long-term memory layer for the orchestrator. Wave M Shot 31 (Zone 69) added memory RAG retrieval into the build-prompt hook — every build now retrieves relevant past context before agents fire. Wave M-31v2 (Zone 77) extended this into WhatsApp replies — Cowork agents pull memory context before responding. Cross-session persistent memory works in production today. This is Pilot Layer 1\'s core substrate, already running.' },
{ title: 'Layer 1 · 5-tier inference chain', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', phase: 'Layer 1', note: 'Inference routing across 5 tiers chosen by cost + capability + availability. Tier 0: DCS Inference (own GPU). Tier 1: Cerebras. Tier 2: SambaNova / Together. Tier 3: Gemini Flash-Lite. Tier 4: DeepSeek. Plus Pod Mode for $0.00 multi-agent builds. Pilot inherits the entire chain without rebuilding anything. Live in production, agent_orchestrator.ts handles routing.' },
{ title: 'Layer 1 · Permanence dashboard + audit log', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Layer 1', note: 'Shot 32 — Permanence dashboard at /admin/permanence.html. Shot 36 — Audit log at /audit-log.html. Shot 22 — Permanence footer-link injector (Zone 49) deployed to every site. Every site published mirrors to Filecoin with content ID. Every autonomous action emits a signed receipt pinned to DCS Storage. Pilot uses these as its signed-action-receipt layer with zero new infrastructure work.' },
{ title: 'Layer 1 · Conversational autonomy (WhatsApp agents live)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Layer 1', note: 'Cowork WhatsApp agents already perform live edits in production. Phase 1 WhatsApp agent operational. Phase 2 WL-3 block inserted after Phase A13 in agent_orchestrator.ts. Sentinel 10 protected zones enforced at runtime. Multi-channel inbox aggregator (Day 6) — unified WhatsApp + Telegram + IG + SMS. Cowork Shots 1-25 shipped. Pilot Layer 1 inherits all of this directly.' },
{ title: 'Layer 1 · Sentinel bounded autonomy (10 protected zones)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Layer 1', note: 'Sentinel hardened to 10 protected zone markers in agent_orchestrator.ts. Build-time sentinel guard on Railway prebuild. Industry guard (Phase 42.14) live. Phase 42.15 nav filter. Sentinel zone 8 (phone sanitizer protection). All protected zones enforced at runtime — not as polite guidance to the model. Pilot uses Sentinel as its bounded-autonomy guardrail; no new safety layer needed.' },
{ title: 'Layer 1 · Long-running agent mode (LRA)', status: 'pending', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 1', note: 'Extend orchestrator to support agents that run for minutes, hours, or until task complete — not just one prompt-response. Owner says \'rewrite my menu page for autumn\' once; agent works through it, asks one clarifying question if needed, returns when done. Pattern lifted from ServiceNow Project Arc and Anthropic Computer Use, adapted for DCS\'s vertical context. Substrate exists (memory + orchestrator + audit); what\'s missing is the loop that keeps an agent active across hours. Estimated 2 weeks Cowork work.' },
{ title: 'Layer 1 · DCS Agents API (public endpoints)', status: 'pending', effort: '3 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 1', note: 'Expose 641 agents as paid endpoints — POST /agents/restaurant/menu-rewrite, /agents/salon/booking-script, etc. External workflow tools (n8n, Make, Zapier) call into DCS instead of bolting in a generic GPT node. DCS becomes infrastructure for everyone else\'s automations. Internal Tier 0 Inference API already shipped; this is the public, vertical-named layer above it. ~3 weeks productization work.' },
{ title: 'Layer 2 · Journey presets (one-click workflow templates)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Layer 2', note: 'Cowork Shot 42 shipped one-click journey preset templates. Sequences customers might run repeatedly — monthly review, seasonal refresh, new-product launch — saved as reusable templates. Pilot Layer 2 extends these from messaging journeys to whole-business workflows touching site + WhatsApp + booking + social in lockstep.' },
{ title: 'Layer 2 · Multi-channel inbox aggregator', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Layer 2', note: 'Cowork Day 6 — Unified multi-channel inbox aggregator across WhatsApp, Telegram, Instagram, SMS. Customer messages from any channel land in one inbox. Pilot uses this as the conversational interface layer — owner messages Pilot on whatever channel they already use; Pilot routes through the inbox infrastructure.' },
{ title: 'Layer 2 · A/B testing + bulk broadcast', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Layer 2', note: 'A/B template testing (split-test variants, conversion tracking) shipped in Cowork. Bulk broadcast — send template to 100+ customers with rate limiting. Multi-language template variants pre-approved per language. Retention emails (Shot 28b) via Resend templates + cron — proves the scheduled-watcher pattern. Pilot Layer 2 watcher agents extend this cron pattern across all metrics, not just messaging.' },
{ title: 'Layer 2 · Natural-language command bar UI', status: 'upcoming', effort: '4 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 2', note: 'Single input at the top of the DCS dashboard. \'Add a winter promotion banner and tell my regulars about it.\' The orchestrator parses intent, picks agents from the 641 specialists, runs the workflow, surfaces a clean diff for one-tap approval. This is the genuinely new UI layer that wraps everything underneath. Estimated 4 weeks Cowork work — biggest single piece of Layer 2.' },
{ title: 'Layer 2 · Always-on watcher agents (cross-metric)', status: 'upcoming', effort: '3 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 2', note: 'Background agents that watch site, bookings, inbox, competitors. They report what\'s changing and propose actions; owner approves, they execute. Retention-email cron + A/B test observability already prove the pattern in messaging; Layer 2 extends it to every metric a business owner cares about. ~3 weeks Cowork build.' },
{ title: 'Layer 2 · Self-healing quality loop', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 2', note: 'DCS Quality Scoring system runs on every build today (Shot 20, Zone 41 — observability live). Pilot Layer 2 closes the loop: when a score drops below threshold, the orchestrator routes the failing site to a fix agent automatically. Failed contrast → contrast agent rewrites the section. Failed Lighthouse → performance agent compresses and re-tests. Self-correcting builds. ~2 weeks work building on existing observability hook.' },
{ title: 'Layer 3 · DCS Vertical Computer Use (orchestration runtime)', status: 'upcoming', effort: '6 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 3', note: 'The biggest net-new engineering layer. Computer-use runtime that drives external platforms in industry-aware ways — knows the difference between updating a Toast menu and a Zomato menu because each connector is industry-coded. Bounded autonomy with audit trail by default. Multi-window synchronizer pattern for chains (one agent fans out to many connected accounts). Failure-to-human escalation. ~6 weeks Cowork build for the runtime; connector library is the long tail on top.' },
{ title: 'Layer 3 · Per-industry connector library (restaurants first)', status: 'upcoming', effort: '8 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 3', note: 'Connector library by industry. Restaurants: Toast, Square, OpenTable, Zomato, delivery apps. Salons: appointment systems, retail POS. Retail: Shopify, marketplaces, inventory tools. Each connector is a thin wrapper — official API where possible, browser automation where not. Released as the customer base for that industry hits 100+. Restaurant connectors first (largest industry), then salon/clinic, then expansion across 87 industries. Community partner program contributes the long tail. ~8 weeks Cowork work for the first 4 connectors.' },
{ title: 'Layer 3 · Multi-account synchronizer + failure escalation', status: 'upcoming', effort: '3 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 3', note: 'Some tasks need the same action across many accounts at once — a chain pushing the same promotion to 5 location Instagram accounts. AdsPower built this for browsers; DCS builds it for vertical agents. Plus never-silent failure escalation — when an agent can\'t proceed (captcha, UI change, ambiguous decision), it pauses, screenshots, pings owner on preferred channel. Owner replies; agent continues. ~3 weeks build.' },
{ title: 'Layer 4 · Predictive Outcome Engine (Wave Q-1 substrate)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Layer 4', note: 'Wave Q-1 of the Predictive Outcome Engine (Zone 75) is already live in observability mode — watching metrics today, accumulating training signal. Pilot Layer 4 is the ACTIVATION step (Q4 2026): moving from observation to action, exposing suggestions to the owner, integrating with the command bar. Engine substrate exists; activation is a much smaller job than building from scratch.' },
{ title: 'Layer 4 · Predictive engine activation (observability → active)', status: 'pending', effort: '4 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 4', note: 'Move the Predictive Outcome Engine from observability mode to active mode. Wire suggestions into the command bar — \'Your Thursday bookings are tracking 22% below last month. Want to run any of these three things?\' Implement approval gating, suggestion ranking, action templates. ~4 weeks build now that observability data is accumulating.' },
{ title: 'Layer 4 · Agents that write their own tools', status: 'upcoming', effort: '6 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 4', note: 'When customer signs up for a niche platform DCS doesn\'t know yet, an agent inspects it, drafts a connector skeleton, escalates to DCS review and ratification. Long-tail platform coverage at near-zero marginal cost. Significant ML + safety work — high effort. Estimated 6 weeks Cowork build.' },
{ title: 'Layer 4 · DCS Verified Badge (F5)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 4', note: 'Already named in inventory as Strategic Product F5. Businesses that run DCS Pilot for 90 days with a clean autonomous-action audit log earn a public DCS Verified mark — trust signal for their customers. Builds on the audit-trail substrate (Layer 1) and the watcher observability (Layer 2). ~2 weeks build.' },
{ title: 'Layer 4 · Multimodal content engine', status: 'upcoming', effort: '4 wk', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 4', note: 'Text, image, audio, video from one prompt. Image already covered by Image Bank (3,520 SDXL images shipped). Extension to short-form video for social, voice replies for IVR/WhatsApp, multilingual variants in one pass. ~4 weeks build on top of existing image generation pipeline.' },
{ title: 'Layer 5 · Salon OS', status: 'pending', effort: 'M3', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 5', note: 'Salon OS is on DCS Platform\'s Q3 2026 roadmap as a full vertical template set — ships alongside Pilot Phase 1, NOT in H2 2027. Appointment management, stylist scheduling, no-show recovery, retention campaigns, product upsell, review collection. The \'I run my whole salon by texting Pilot\' product. Per-industry agents + connectors + templates bundled as one product.' },
{ title: 'Layer 5 · Clinic OS', status: 'pending', effort: 'M3', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 5', note: 'Clinic OS is on DCS Platform\'s Q3 2026 roadmap as a full vertical template set. Patient intake, appointment management, prescription reminders, compliance-aware comms, common-EHR connectors. Heavier regulatory surface, higher per-customer revenue, narrower initial market. BAA available per existing Sovereign positioning.' },
{ title: 'Layer 5 · Restaurant OS', status: 'upcoming', effort: 'M5-6', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 5', note: 'All DCS products configured for restaurants, sold as one product. Site, ordering, menu sync across delivery apps, bookings, loyalty, daily-specials automation, predictive promotions, weekly reporting — one command runs the whole restaurant\'s online presence. H1 2027 ship target.' },
{ title: 'Layer 5 · Retail OS', status: 'upcoming', effort: 'M6-7', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 5', note: 'For shops and small retail chains. Inventory across channels, marketplace listings, promotion sync, customer loyalty, supplier coordination, sales reporting. One command, all sales channels in step. H1 2027 ship target.' },
{ title: 'Layer 5 · DCS Business OS (white-label core)', status: 'upcoming', effort: 'M7-8', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 5', note: 'The underlying platform that powers all vertical OS products — sold to partners launching their own vertical product for industries DCS doesn\'t directly target (accounting OS, law-firm OS, real-estate-broker OS). Partner program + partner marketplace. B2B2B revenue model. H2 2027 ship.' },
{ title: 'Layer 5 · DCS Framework Export', status: 'upcoming', effort: 'M8-9', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Layer 5', note: 'Customers who outgrow the hosted product can export their DCS setup as a self-hosted framework — same orchestrator, same agents, their own infrastructure. Removes the \'what if DCS shuts down\' objection forever. Adjacent to DCS Sovereign positioning. H2 2027 ship.' },
{ title: 'Marketing site (pilot.dcsai.ai)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Marketing', note: 'Full single-HTML product-detail page (136KB, 2832 lines, 17 sections) shipped May 14 2026. Same dark theme as compute/storage. Hero with "First waves shipping now · Full vision Q3 2026" eyebrow + command demo. Built-on trust bar. Detailed 9-row comparison table vs current options. 6 industry use case walkthroughs. 87-industry coverage grid. 6 key capabilities. 5-layer capability deep-dive with green "Already running" callout strip. Architecture (5 layers). Security & Trust pillars. Pricing tiers (Starter/Growth featured/Pro indicative). 6 "What Pilot doesn\'t do" cards. Roadmap timeline. 5-node Ecosystem loop. 5-region markets grid. 18-question FAQ. Waitlist CTA. Footer. Ready for Cowork to deploy to pilot.dcsai.ai.' },
{ title: 'pilot.dcsai.ai DNS + Cloudflare Pages setup', status: 'pending', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Deploy', note: 'Cowork deploy step. Create pilot.dcsai.ai subdomain in Cloudflare DNS. Create Cloudflare Pages project pointing at a new GitHub repo (trd-pilot-frontend) or sub-folder of trd-frontend. Deploy pilot-index.html. Wire SSL. Estimated 1 day Cowork work.' },
{ title: 'Waitlist endpoint wiring', status: 'pending', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-14', phase: 'Deploy', note: 'Marketing page has placeholder onsubmit alert. Wire to POST /api/pilot-waitlist (Railway) or reuse existing waitlist infra. Send confirmation email via Resend. Add Supabase row to waitlists table with source=pilot. Optional: SSO check against trd-auth localStorage key to greet returning visitors. 1 day work.' },
],
tech: {
site: 'https://pilot.dcsai.ai (marketing/waitlist, deploy pending)',
stack: 'Inherits the full DCS stack — Platform · Compute · Storage · OS · Sentinel · agent_memory.ts',
substrate_live: 'agent_memory.ts + RAG (Zone 69, Zone 77) · 5-tier inference · /admin/permanence.html · /audit-log.html · multi-channel inbox · WhatsApp agents · Sentinel 10 zones · Predictive Engine Wave Q-1 (Zone 75 observability)',
genuinely_new: 'Long-running mode · DCS Agents API public endpoints · NL command bar UI · Watcher agents · Self-healing quality loop · Vertical Computer Use runtime · Per-industry connectors · Predictive engine activation · Vertical OS products',
timeline: 'Layer 1 substrate already running · Layer 2 Q3-Q4 2026 · Layer 3 Q1 2027 · Layer 4 Q4 2026 (activation) · Layer 5 Salon/Clinic Q3 2026 + rest H2 2027',
notes: 'Pilot is the productization layer wrapping running DCS infrastructure into one branded autonomous experience. Not a from-scratch product — most of Layer 1 is already live in production as of May 14 2026 audit.',
},
},
// ============== 03 · DCS PLATFORM INTERNALS ==============
{
id: 'i-build', project: '03 · DCS PLATFORM INTERNALS', name: 'Build Pipeline',
kind: 'internal', priority: 'critical', launchReady: 'partial',
desc: '17-agent orchestrator. Core engine that produces every DCS site. 127s p50 → 90s target.',
items: [
{ title: '5-tier inference chain (Tier 0–4)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Inference routing across 5 tiers chosen by cost + capability + availability. Tier 0: DCS Inference (own GPU on RunPod H200, $0.00 cost when available). Tier 1: Cerebras (5 keys, 15M TPM combined, primary for most agents). Tier 2: Together GPT-OSS 120B (for heavy lifters). Tier 3: Gemini Flash-Lite (fast fallback). Tier 4: DeepSeek (slow but cheapest final fallback). Routing logic in agent_orchestrator.ts picks tier based on agent type + current capacity. Cost per build: $0.04-0.16 depending on tier mix.' },
{ title: 'Sentinel hardened (10 protected zones)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-09', note: 'Sentinel system hardened on May 9 to 10 protected zone markers across agent_orchestrator.ts, index.ts, whatslink files, sovereign_leads.ts, and inbox.ts. Each marker is a critical code block where changes require [DCS-EDIT] handshake. Railway deploy check enforces marker count. Critical safety net as code base grew past 16k lines.' },
{ title: 'Reduce p50 build 127s → ≤90s', status: 'upcoming', effort: '1 week', owner: 'Deepak Dudi', addedDate: '2026-05-05', note: 'Current p50 build time = 127 seconds for a complex DCS Platform site (17 agents, full pipeline). Target ≤90 seconds for launch. Two paths: Path A — reduce agent count from 17 to ~14 by merging or skipping low-value agents (faster but reduces capability). Path B — parallelize sequential agents that don\'t actually have data dependencies (preserves capability, more complex). Path B requires [DCS-EDIT] handshake to Cowork agent first because orchestrator changes touch Sentinel zones.' },
{ title: 'Cerebras 5-key TPM ceiling fix', status: 'upcoming', effort: '30 min', owner: 'Deepak Dudi', addedDate: '2026-05-12', testResult: 'fail', note: 'Current bug: when Tier 3 chunk 2 of a complex build exhausts all 5 Cerebras API keys\' TPM (tokens-per-minute) ceiling, the orchestrator falls back to DeepSeek which takes 366 seconds — kills build time. Fix options: (a) add 5 more Cerebras keys (~$50/mo more), (b) better load balancing across the 5 existing keys, (c) graceful fallback to Together instead of DeepSeek when Cerebras maxed. Diagnostic + fix needed.' },
{ title: 'Prompt caching across all 8 callXXX', status: 'upcoming', effort: '4 hrs', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Anthropic prompt caching can reduce token costs by 90% for repeated prompt prefixes. Currently used in some callXXX functions but not all 8. Apply consistently across callCerebras, callDeepSeek, callGemini, callGroq, callSambaNova, callTogether, callTrdInference, callAnthropic. Estimated savings: 30-50% reduction in total inference cost at scale. Implementation: ~2 hours per function, ~16 hours total.' },
],
},
{
id: 'i-templates', project: '03 · DCS PLATFORM INTERNALS', name: 'Template System',
kind: 'internal', priority: 'critical', launchReady: 'partial',
desc: '6 → 500 templates across 10+ industries. Inherits header/footer caps, widgets, palette, 6-template rotation.',
items: [
{ title: '6 restaurant templates locked (girl-goat 9/10)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-08', testResult: 'pass', note: 'Six restaurant templates locked May 8: noma-editorial, gucci-osteria, girl-goat (highest quality at 9/10), gramercy, sunday-brooklyn, nobu-signature. Each template is a complete restaurant website with menu structure, booking widget, gallery, location info, contact section. Locked status means DCS Foundation auto-inherits these for every restaurant build via Phase G1 template rotation. girl-goat is the gold standard reference.' },
{ title: 'Beauty 6 · Healthcare 6 · Travel 6 · Visa 6', status: 'upcoming', effort: '4 weeks', owner: 'Cowork', deps: 'CSS Variable System', addedDate: '2026-05-08', note: 'Next 4 industries to get 6-template sets: Beauty (salons, spas, cosmetic clinics), Healthcare (dentists, physiotherapy, general practice), Travel (hotels, tour operators, travel agencies), Visa (immigration consultants — large UAE/India market). Each industry takes ~1 week to research, design, and integrate. Total: 24 new templates across 4 weeks. Queued behind CSS Variable System work.' },
{ title: 'Fitness 6 · Fashion 6 · NGO 6 · Salon 6', status: 'upcoming', effort: '4 weeks', owner: 'Cowork', addedDate: '2026-05-08', note: 'Following 4 industries: Fitness (gyms, personal trainers, yoga studios), Fashion (boutiques, designers, e-commerce fashion), NGO (non-profits, charities, foundations), Salon (already covered under Beauty? — TBD whether merge or separate). Each ~1 week. Brings total to ~50 industry-specific templates across 8 industries.' },
{ title: 'Scale to 500 templates total', status: 'upcoming', effort: '10 industries × 50', owner: 'Cowork', addedDate: '2026-05-01', note: 'M1-M3 stretch goal: 500 templates total. Paths to scale: (a) more industries 6 templates each = ~80 industries × 6 = 480 templates (slow), (b) sub-categories within industries (e.g. fine-dining vs casual vs fast-food vs ghost kitchen) × 6 each, (c) AI-generated template variations from existing templates as premium feature, (d) community-submitted templates via Marketplace (Year 2). Realistic blend of all 4 approaches.' },
],
},
{
id: 'i-css', project: '03 · DCS PLATFORM INTERNALS', name: 'CSS Variable System',
kind: 'internal', priority: 'critical', launchReady: 'no',
desc: 'All 658 sections use var(--tpl-*) instead of hardcoded colors. PREREQUISITE for all template expansion.',
items: [
{ title: 'Add css_vars block to template JSONs', status: 'upcoming', effort: '2 hrs', owner: 'Cowork', addedDate: '2026-05-08', note: 'Refactor: each template JSON gets a css_vars block listing all CSS custom properties it uses (--color-primary, --color-bg, --spacing-md, --font-heading, etc.). Allows: per-customer theme generation (override vars based on brand), dark mode support, automatic palette extraction from customer logo, theme presets. Foundation step before any visual customization features.' },
{ title: 'Python script: replace hardcoded colors in 658 sections', status: 'upcoming', effort: '3 hrs', owner: 'Cowork', addedDate: '2026-05-08', note: 'Scripted batch operation by Cowork agent: scan all section HTML files (658 across all templates), replace hardcoded color values (#FF6B35, rgb(50,50,50), etc.) with CSS variable references (var(--color-accent), var(--color-text-secondary)). ~5 hours of automated work. Critical step toward theme customization — until done, theme changes require regenerating sections.' },
{ title: 'Smoke: 6 restaurant templates render differently', status: 'upcoming', effort: '30 min', owner: 'Deepak Dudi', addedDate: '2026-05-08', note: 'After CSS variable migration, smoke test: build 6 different restaurant sites using the 6 locked templates with same content. Each should render visually distinct (different colors, different typography, different spacing) — confirms templates aren\'t accidentally sharing CSS through hardcoding. Critical regression test before declaring CSS Variable System done.' },
],
},
{
id: 'i-phase42', project: '03 · DCS PLATFORM INTERNALS', name: 'Phase 42 Bug Cleanup',
kind: 'internal', priority: 'critical', launchReady: 'partial',
desc: 'Contrast guards, cache headers, gallery dedup, industry guards, header rewriter. Smoke majors 61 → 49 → <20 target. The bug-cleanup track that locked DCS Foundation at commit afb7c88.',
items: [
// Early Phase 42
{ title: 'Phase 42.11 + 42.12 — Speakeasy template + Velvet & Smoke build', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-09', phase: 'Phase 42', commits: '2f828a1', note: 'Phase 42.11 and 42.12 shipped May 9 (commit 2f828a1). Speakeasy template subset (cocktail bar variant of the restaurant industry templates) had multiple rendering bugs: gallery overlap, header spacing issues, mobile menu broken on iOS Safari. Velvet & Smoke was the live test build that surfaced these bugs — built with the speakeasy template, customer-facing site at velvet-smoke.dcsai.ai. Fixed in 2f828a1 with surgical CSS adjustments + JS event handler fixes.' },
{ title: 'Phase 42.13 — More speakeasy template fixes', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-09', phase: 'Phase 42', note: 'Continued speakeasy template fixes after Phase 42.11/12. Speakeasy is the cocktail-bar restaurant template subset that had several rendering bugs (gallery overlap, header spacing, mobile menu issues). Phase 42.13 fixes target subset cleanup before moving to other industries.' },
{ title: 'Phase 42.14 — Industry guard', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-09', phase: 'Phase 42', note: 'Industry guard prevents agent confusion when building sites for industries with overlapping templates. Example: a restaurant prompt could accidentally trigger e-commerce templates if not properly gated. Phase 42.14 adds explicit industry verification step before template selection — prevents mismatched template/content pairing.' },
{ title: 'Phase 42.15 — Nav filter', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-09', phase: 'Phase 42', note: 'Navigation filter ensures only relevant nav items appear for each industry. Example: a restaurant site shouldn\'t have \'Shop\' in nav (e-commerce-only item), a portfolio site shouldn\'t have \'Menu\' (restaurant-only). Filter logic runs on every build during the page-generation phase to clean up inappropriate nav items.' },
// Mid Phase 42
{ title: 'Phases 42.19-42.30 — 27 contrast guards', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', phase: 'Phase 42', testResult: 'pass', note: '27 contrast guards shipped across Phases 42.19 through 42.30 (May 9-10). Contrast guard = CSS post-processor that ensures every text/background combination on built sites meets WCAG AA accessibility standards (4.5:1 ratio for body text, 3:1 for large text). Previously some agent-generated CSS produced unreadable color combinations (light grey text on white background). Guard auto-corrects to nearest accessible color. 27 specific guard rules covering: hero sections, navigation, buttons, footers, body text, headings, captions, links. Result: smoke test major failures dropped from 61 → 49.' },
{ title: 'Phase 42 — Gallery deduplication', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', phase: 'Phase 42', note: 'Bug fix: gallery sections were sometimes duplicating images when the agent generated content for a multi-section page. Root cause: AI agents sometimes call gallery population twice. Phase 42 dedup logic checks for duplicate image URLs and removes them before render. Reduces visual confusion and slight performance improvement.' },
{ title: 'Phase 42 — Cache headers fixed', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', phase: 'Phase 42', note: 'Bug fix: Cache headers on published DCS sites were too aggressive — customers couldn\'t see their updates after editing because browsers cached old version for hours. Phase 42 fix: shorter cache TTL on HTML (5 minutes), longer cache TTL on static assets (1 year with versioning). Customers now see edits within 5 minutes of publishing.' },
{ title: 'Phase 42 — Header rewriter', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', phase: 'Phase 42', note: 'Header rewriter ensures every page on a built site has consistent header structure: logo, navigation, CTA button, contact info. Different agents previously generated slightly different headers across pages — broke user expectation of consistency. Phase 42 fix: single header template applied across all pages of a site via post-build rewrite step.' },
{ title: 'Phase 42.30 — DCS Foundation locked at commit afb7c88', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', phase: 'Phase 42', commits: 'afb7c88', testResult: 'pass', note: 'Phase 42.30 shipped May 10 (commit afb7c88). Critical milestone: DCS Foundation locked, meaning every new industry template added going forward automatically inherits all the chrome and engineering — header/footer caps, widgets, sitemap generator, palette extractor, 6-template rotation logic. New templates only need to define industry-specific content (page structure, copy patterns) — the underlying framework is shared. This removes huge engineering overhead per new industry and is the reason adding Beauty/Healthcare/Travel/etc. templates only takes ~1 week each.' },
// Open bugs
{ title: 'Phase 42.27 missed 6 pages (regex bug)', status: 'pending', effort: '1 hr', owner: 'Cowork', addedDate: '2026-05-12', phase: 'Phase 42', testResult: 'fail', rootCause: 'Template-prefix injection only worked on 21/27 pages', note: 'May 12 bug. Phase 42.27 was supposed to inject the template-color-prefix CSS class into all 27 generated pages of a build. Worked correctly on 21 pages, silently skipped 6: gift-cards, reservation, services, testimonials, menu, wine-list, order. Root cause: the regex pattern that detected "this is a page that needs prefix injection" didn\'t match the file naming convention for these 6 page types (they have hyphens or are nested). Result: those 6 pages render with default colors instead of the customer\'s brand palette — visible color inconsistency. Fix in flight by Cowork agent. Effort ~1 hour to update regex + verify across all 27 page types.' },
{ title: 'Footer black-on-black 42.29 patch (!important)', status: 'pending', effort: '45 min', owner: 'Cowork', addedDate: '2026-05-12', phase: 'Phase 42', testResult: 'fail', rootCause: 'Phase 42.29 contrast guard ran but did not fix Girl & Goat footer · element-level inline style override needed', proposedFix: 'Add !important to footer color rules · target via [data-trd-light-page="1"] body attribute', note: 'May 12 bug specific to Girl & Goat template footer. Phase 42.29 contrast guard ran, detected the dark footer text on dark background, attempted to fix — but the fix didn\'t apply because the original element has inline style attribute that takes precedence over the contrast guard\'s generated CSS class. Root cause: contrast guard generates external CSS, but inline styles always win in CSS cascade. Proposed fix: add !important to footer color rules in the contrast guard output, OR target via [data-trd-light-page="1"] body attribute that the guard already sets. Owned by Cowork. ~45 minutes effort.' },
// Upcoming
{ title: 'Smoke test majors 49 → <20', status: 'upcoming', effort: '1 week', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Phase 42', note: 'Smoke test is the automated nightly run that prompts ~30 representative builds across industries and counts major issues per build. Current: 49 major issues across the suite (down from 61 in early May after Phase 42 cleanup). Target before June 4: <20 majors. "Major" = build fails, page fails to render, critical UX broken. "Minor" = cosmetic issue or non-critical content gap. Path forward: identify top 5 most-frequent failure modes from the 49 current issues, batch-fix them. Most issues cluster around contrast/color (now mostly handled by Phase 42 guards) and content quality (which requires prompt engineering, not code fixes).' },
],
tech: {
stack: 'agent_orchestrator.ts (16,156 lines) · contrast guards · gallery dedup · cache headers',
recent_commits: 'afb7c88 (Phase 42.30 lock) · 2f828a1 (42.11-13)',
notes: 'Cleanup track ran parallel with feature work May 9-12. Foundation locked at afb7c88 prevents regression.',
},
},
{
id: 'i-whatsapp-widget', project: '03 · DCS PLATFORM INTERNALS', name: 'WhatsApp Widget on Built Sites',
kind: 'internal', priority: 'high', launchReady: 'no',
desc: 'Widget embedded on every DCS-built site. Table stakes for UAE/India target markets. (Separate from DCS Cowork agent.)',
items: [
{ title: 'WhatsApp widget embed on every site', status: 'upcoming', effort: '~4 hrs', owner: 'Deepak Dudi', addedDate: '2026-05-08', note: 'Click-to-WhatsApp floating button embedded on every DCS-built site. Customer clicks → opens WhatsApp with pre-filled message addressed to business owner. Critical for UAE/India target markets where WhatsApp is primary communication channel. Table stakes — sites without it feel incomplete to local users. Estimated 4 hours work to roll out across all 6 restaurant templates initially.' },
{ title: 'WhatsApp Business API approval', status: 'upcoming', effort: '2–4 weeks', owner: 'ops', addedDate: '2026-05-08', note: 'WhatsApp Business API approval required for high-volume usage (>1000 messages/day per number). Application via Meta Business Manager. Required documentation: business registration (DCS AI Tech LLC license), use case description, expected message volume, opt-in flow demonstration. Approval timeline: 2-4 weeks. Until approved: use click-to-chat which works without approval but has volume limits.' },
],
},
{
id: 'i-content-editor', project: '03 · DCS PLATFORM INTERNALS', name: 'Content Editor M1–M6',
kind: 'internal', priority: 'medium', launchReady: 'no',
desc: 'Per-tab inline editing: Services / Testimonials / FAQ / Gallery / Menu / Portfolio / Blog / Team / Booking / Product variants / Inline.',
items: [
{ title: 'M1: Services / Testimonials / FAQ / Gallery', status: 'upcoming', effort: '1 week', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Month 1 deliverables for Content Editor: inline editors for the four highest-volume content types. Services (most industries use this), Testimonials (universal — every site has these), FAQ (SEO-critical, generates schema markup), Gallery (visual sites need this). One week each = 4 weeks total. Implementation: per-tab UI in Edit Website page, save to Supabase content tables, deploy via reinject route.' },
{ title: 'M2: Menu / Portfolio', status: 'upcoming', effort: '1 week', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Month 2: Menu editor (restaurant-specific, complex due to categories/items/prices/dietary tags) and Portfolio editor (creative industries — designers, photographers, agencies). Each ~1 week. Menu editor includes structured data generation (schema.org/Menu) for Google food search visibility. Portfolio supports multi-image projects with descriptions.' },
{ title: 'M3: Blog / Team', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Month 3: Blog editor (rich-text WYSIWYG, image embedding, post scheduling) and Team editor (member profiles, photos, bios, social links). 1 month total — Blog is more complex than other content types due to rich text + media handling. Schema.org/BlogPosting generated for SEO. Team page supports filterable view by department/role.' },
{ title: 'M4: Booking', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Month 4: Booking editor (services + staff calendars + availability rules + customer notification). Major feature for service businesses (salons, clinics, consultants). 1 month effort. Includes: service definition, staff calendars, time-slot availability, holiday/blocked time, customer reminders via WhatsApp + email. Replaces Calendly/Acuity for DCS customers.' },
{ title: 'M5: Product variants', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Month 5: Product variants editor (e-commerce — sizes, colors, quantities, SKU management). Critical for e-commerce sites. 1 month effort. Includes: variant definition (size + color matrix), per-variant inventory tracking, per-variant pricing, image gallery per variant. Foundation for proper e-commerce checkout in M6+.' },
{ title: 'M6: Inline editor', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Month 6: Fully inline editing — customer clicks any text directly on rendered site and types replacement (no separate Edit page). End goal of Content Editor roadmap. Most complex implementation: requires save-back-to-source mechanism across all content types. Last in roadmap because previous M1-M5 features cover 80% of edit use cases without requiring this level of UX polish.' },
],
},
{
id: 'i-stripe', project: '03 · DCS PLATFORM INTERNALS', name: 'Stripe Pricing Tiers',
kind: 'internal', priority: 'high', launchReady: 'no',
desc: 'Starter $19 · Growth $49 · Pro $99 (planned). Pricing the Platform separately from Compute (which already has Stripe live).',
items: [
{ title: 'Stripe Connect setup for Platform', status: 'upcoming', effort: '~2 wk approval', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Stripe Connect application + approval needed for marketplace flows: Agency tier collecting payments from clients, Partner Program paying commissions, Enterprise tier with custom payment terms. Application via Stripe Dashboard, approval typically 2-3 weeks. Required docs: business registration (DCS AI Tech LLC), expected volume, use case. Separate from basic Stripe integration that\'s already live for DCS Compute.' },
{ title: 'Pricing page UI ($19 / $49 / $99)', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Public 3-tier pricing page: Starter $19/mo (1 site, DCS branded, basic features), Growth $49/mo (3 sites, white-label option, advanced agents), Pro $99/mo (unlimited sites, framework export, priority support). Monthly + annual billing toggle (annual = 2 months free). Built before launch — required for accepting payments. Includes feature comparison table, FAQ, testimonial section (when available).' },
{ title: 'Tier enforcement in /app22', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-01', note: 'Enforce feature gates in /app22 based on user\'s active Stripe subscription. Starter users see \'Upgrade to Pro\' CTAs on locked features. Soft enforcement (graceful UX messaging) rather than hard blocks — better conversion than aggressive paywalls. Subscription status synced from Stripe webhook on every change (subscription.created, updated, canceled, payment_failed). Feature check middleware on every protected route.' },
],
},
{
id: 'i-sentinel', project: '03 · DCS PLATFORM INTERNALS', name: 'Sentinel Protected Zones',
kind: 'internal', priority: 'high', launchReady: 'yes',
desc: '10 protected markers across agent_orchestrator.ts (16,156 lines). Railway turns RED in 60s if any marker missing. [DCS-EDIT] handshake required.',
items: [
{ title: 'Sentinel hardened to 10 markers', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-09', testResult: 'pass', note: 'Sentinel hardened May 9 to 10 protected zone markers across agent_orchestrator.ts (16,156 lines), src/index.ts, whatslink files, sovereign_leads.ts, and inbox.ts. Each marker is a specifically-formatted comment block. Railway pre-deploy hook scans for the count; if any missing, deploy fails in 60 seconds and Railway turns RED. Critical safety net as code base grew past 16k lines.' },
{ title: 'WL-3 import @ line 7', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-09', note: 'Sentinel marker at line 7 of agent_orchestrator.ts protecting the WL-3 (WhatsLink-3) import block. Critical because WhatsLink form injection runs on every DCS-built site. Removing this import breaks the Reserve a Table button on all built sites. Marker prevents accidental deletion during refactors.' },
{ title: 'Phase 43.3 trackVendorUsage @ line 40', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Sentinel marker at line 40 protecting Phase 43.3 trackVendorUsage function (shipped May 12, commit 568c8ea). This function logs every Cerebras API call to vendor_usage_events table for cost tracking. Removing breaks all vendor spend visibility going forward. Marker prevents deletion during routine refactors.' },
{ title: 'Phases 42.14–42.27 markers', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Sentinel markers protecting Phase 42.14 through 42.27 bug fixes: industry guards, nav filters, contrast guards, gallery deduplication. All shipped May 9-10 as part of bug cleanup track. Markers prevent regression — if a future change deletes the fix, Sentinel catches it within 60 seconds via Railway deploy check.' },
],
},
{
id: 'i-infrastructure', project: '03 · DCS PLATFORM INTERNALS', name: 'Infrastructure & Deploy',
kind: 'internal', priority: 'critical', launchReady: 'yes',
desc: 'Complete map of every repo, every hosting platform, every deploy flow, every DNS record. The single source of truth for "where does X live and how does it deploy." Captured May 13, 2026 from live system after confirming with founder.',
items: [
// ============ REPOS ============
{ title: 'Backend repo: github.com/TRDnetwork/trd-backend-clean', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Backend', note: 'Main DCS Platform backend. Local path: /Users/NEWUSER/Desktop/Project TRDN/trd-backend-clean. Path has spaces — always quote when used in terminal commands. Contains: agent_orchestrator.ts (16,156 lines), src/index.ts (router mounts), src/services/* (vercel_deploy, github_deploy, callXXX inference functions), src/routes/* (REST endpoints), migrations/. TypeScript with NodeNext module resolution. Deploys to Railway on push to main.' },
{ title: 'Frontend repo: github.com/TRDnetwork/trd-frontend', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-03-15', phase: 'Frontend', note: 'Main DCS Platform frontend. Local path: /Users/NEWUSER/Desktop/Project TRDN/trd-frontend. Contains app22.html (the dashboard) at REPO ROOT (not in public/ or dist/ — directly at trd-frontend/app22.html). Also: index.html, about-us.html, career.html, contact.html, security-audits.html, and 13 other static HTML files. Deploys to Cloudflare Pages on push to main.' },
{ title: 'Storage backend repo: github.com/TRDnetwork/trd-storage', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-07', phase: 'Backend', note: 'Standalone DCS Storage Express+TypeScript backend. Separate Railway project (not in main backend). NodeNext module resolution. Lighthouse SDK 0.4.5. Routes: /api/storage/* (assets CRUD, certificates, mirrors). Deploys to Railway on push to main.' },
{ title: 'Storage frontend repo: github.com/TRDnetwork/trd-storage-frontend', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-07', phase: 'Frontend', note: 'Standalone DCS Storage marketing site + asset vault UI. Single-file index.html ~5000 lines. Deploys to Cloudflare Pages on push to main. DNS: storage.dcsai.ai CNAME to pages.dev project.' },
{ title: 'Compute backend repo: github.com/TRDnetwork/trd-cn-backend', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-05', phase: 'Backend', note: 'DCS Compute Network (DCS-CN) backend. Separate Railway project. Shared Supabase with main DCS via compute_* prefixed tables. Routes: /api/compute/* (workers, jobs, payouts), /api/sso/* (HMAC token exchange with /app22). Deploys to Railway on push to main.' },
{ title: 'Compute frontend repo: github.com/TRDnetwork/trd-cn-frontend', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-05', phase: 'Frontend', note: 'DCS-CN marketing site + worker dashboard. Deploys to Cloudflare Pages on push to main. DNS: compute.dcsai.ai CNAME to pages.dev project. Worker SDK install instructions, GPU tier earnings calculator, FAQ.' },
// ============ HOSTING PLATFORMS ============
{ title: 'Railway hosts all 3 backends', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Hosting', note: 'Railway hosts 3 separate projects: (1) main DCS backend serving api.dcsai.ai — the agent orchestrator + REST routes for /app22. (2) DCS-CN backend at api.compute.dcsai.ai (or similar). (3) DCS Storage backend at api.storage.dcsai.ai. Each project has independent env vars, independent deploys, independent Sentinel checks. Auto-deploy on push to main branch via Railway GitHub integration.' },
{ title: 'Cloudflare Pages hosts all 3 frontends', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Hosting', note: 'Cloudflare Pages hosts the static frontends: (1) trd-frontend → dcsai.ai + all subpaths including /app22, /about-us, etc. (2) trd-storage-frontend → storage.dcsai.ai. (3) trd-cn-frontend → compute.dcsai.ai. Auto-deploy on push to main branch. Build command typically not needed (static HTML). Global CDN means <1s page load worldwide. SSL auto-issued.' },
{ title: 'Cloudflare DNS for dcsai.ai and all subdomains', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'DNS', note: 'Cloudflare manages DNS for dcsai.ai. Records: dcsai.ai → Cloudflare Pages (trd-frontend project). storage.dcsai.ai → Cloudflare Pages (trd-storage-frontend). compute.dcsai.ai → Cloudflare Pages (trd-cn-frontend). sovereign.dcsai.ai → Cloudflare Pages. api.dcsai.ai → Railway backend (main DCS). api.storage.dcsai.ai → Railway (storage backend). api.compute.dcsai.ai → Railway (DCS-CN backend). All records have orange-cloud proxy enabled for DDoS protection.' },
{ title: 'Hostinger hosts visual-editor.html', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-01', phase: 'Hosting', note: 'Visual Editor frontend (visual-editor.html, split from app22.html on May 1) hosted on Hostinger separately. Reason: CORS configuration easier on Hostinger for the iframe-based preview. May be migrated to Cloudflare Pages later for consolidation, but currently works.' },
{ title: 'Supabase hosts the database (shared across all products)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-03-20', phase: 'Database', note: 'Single Supabase project shared across DCS Platform, DCS Storage, DCS Compute, DCS Sovereign. Same auth.users table for unified sign-on. Per-product table prefixes for separation: agents (Platform), compute_* (Compute), storage_* (Storage), session_messages (chat), trd_inference_logs (Inference), vendor_usage_events (cost tracking). RLS policies tied to auth.uid() on every product table.' },
{ title: 'RunPod hosts GPU compute', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-08', phase: 'GPU', note: 'RunPod provides the underlying GPU infrastructure that DCS GPU resells. Also hosts the DCS Inference vLLM pod (H200 with Qwen3-32B-FP8). Pod management API used by agent_orchestrator.ts in Pod Mode. Each customer pod gets its own RunPod instance with billing tracked per-second.' },
// ============ DEPLOY FLOWS ============
{ title: 'Backend deploy: git push origin main → Railway auto-deploys', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Deploy flow', note: 'For trd-backend-clean: cd to "/Users/NEWUSER/Desktop/Project TRDN/trd-backend-clean" → git add -A → git commit -m "..." → git push origin main. Railway auto-deploys within 60-90 seconds. Pre-deploy hook runs Sentinel marker check; if marker count != 10, deploy fails and Railway turns RED. Same flow for trd-storage and trd-cn-backend repos (separate Railway projects).' },
{ title: 'Frontend deploy: git push origin main → Cloudflare Pages auto-deploys', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-03-20', phase: 'Deploy flow', note: 'For trd-frontend: cd to "/Users/NEWUSER/Desktop/Project TRDN/trd-frontend" → git add app22.html (or whichever files changed) → git commit -m "..." → git push origin main. Cloudflare Pages auto-deploys within 30-60 seconds. Deploys are static (no build step for plain HTML). Same flow for trd-storage-frontend and trd-cn-frontend repos.' },
{ title: 'Cowork agent push protocol: [DCS-EDIT] handshake required', status: 'completed', effort: 'in production', owner: 'Cowork', addedDate: '2026-05-09', phase: 'Deploy flow', note: 'WhatsApp Cowork agent must send [DCS-EDIT] handshake message to founder BEFORE pushing changes touching Sentinel-protected files. Message format: file path + line range + commit SHA + change summary. Founder reviews and can intervene. 138 handshakes recorded to date. Protocol prevents silent breakage of critical orchestrator logic.' },
// ============ ENV VARS ============
{ title: 'Main DCS backend env vars (Railway)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-15', phase: 'Config', note: 'Set in Railway dashboard for trd-backend-clean project: ANTHROPIC_API_KEY, CEREBRAS_API_KEY_1 through CEREBRAS_API_KEY_5 (5 keys for TPM aggregation), TOGETHER_API_KEY, GEMINI_API_KEY, DEEPSEEK_API_KEY, GROQ_API_KEY (pending dev tier access), SAMBANOVA_API_KEY (Phase 41.2), SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY, RESEND_API_KEY, STRIPE_SECRET_KEY, STRIPE_WEBHOOK_SECRET, RUNPOD_API_KEY, VERCEL_TOKEN (pending), GITHUB_TOKEN, NOWPAYMENTS_API_KEY, HMAC_SSO_SECRET, TRD_INFERENCE_URL, TRD_INFERENCE_TRAFFIC_WEIGHT (0.5-0.7), TRD_INFERENCE_TIMEOUT_MS=60000.' },
{ title: 'Storage backend env vars (Railway)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-07', phase: 'Config', note: 'Set in Railway dashboard for trd-storage project: LIGHTHOUSE_API_KEY, SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY, RESEND_API_KEY. Minimal env footprint — service is focused and has few external dependencies.' },
{ title: 'Compute backend env vars (Railway)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-05', phase: 'Config', note: 'Set in Railway dashboard for trd-cn-backend project: SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY, HMAC_SSO_SECRET (must match main DCS backend for SSO to work), STRIPE_SECRET_KEY (separate Stripe account for compute), NOWPAYMENTS_API_KEY (crypto on-ramp for renters).' },
// ============ COMMITS / VERSIONS ============
{ title: 'Recent backend commits (main)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-12', phase: 'Versions', note: 'Most recent commits on trd-backend-clean main: 568c8ea (Phase 43.3 Cerebras vendor usage tracker), 6954cd6 (vendor usage tracker base), 6121845 (v4 auto-fetch worker), cbf0d97 (v3 timeseries), b5462ea (vendor-spend phase 2), 466ee15 (DCS Inference Tier 0 shipped), afb7c88 (Phase 42.30 DCS Foundation locked), 2f828a1 (Phases 42.11-13 speakeasy fixes). Run "git log --oneline | head -15" in repo to see latest.' },
{ title: 'Smoke test infrastructure (Phase 42 cleanup)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-09', phase: 'Testing', note: 'Automated smoke test runs 5 representative builds end-to-end (restaurant, beauty, e-commerce, portfolio, services) and checks: build completes, all pages render, no broken links, lighthouse >85, mobile responsive, contact form works. Currently 49 major failures across 5 builds (~10 issues per build). Target: <20 majors before launch. Run via internal command (see Phase 42 internal product for details).' },
// ============ KEY OPERATIONAL RULES ============
{ title: 'Repo paths must be quoted (spaces in path)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Operational', note: 'Local Mac paths contain spaces: "/Users/NEWUSER/Desktop/Project TRDN/trd-backend-clean" and "/Users/NEWUSER/Desktop/Project TRDN/trd-frontend". ALWAYS quote these in terminal commands. Unquoted paths break cd, cp, ls, git commands silently or with confusing errors. This is one of the most common Cowork mistakes — strict rule.' },
{ title: 'ASCII-only commit messages, no leading # in pasted code', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Operational', note: 'Operational rule from extensive trial and error: commit messages must be ASCII only (no emojis, no special characters) — otherwise git on some systems breaks. Code blocks pasted into terminal must not start with # comment lines — zsh interprets some # patterns as history expansion and corrupts the paste. Patches should be written as .py files and executed from disk, not heredoc-piped into commands.' },
{ title: 'One change per commit (Rule 25)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-15', phase: 'Operational', note: 'Each commit touches one logical change. Makes git log readable, makes reverts surgical, makes Cowork [DCS-EDIT] handshakes accurate. Bad pattern: "fix bugs and add feature" commits. Good pattern: separate commits for each fix and the feature, each pushed sequentially. Slower in the moment, faster long-term when debugging.' },
{ title: 'SQL + logs first, no theorizing without data (Rule 26)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-15', phase: 'Operational', note: 'Debugging rule: before proposing fixes, gather evidence. Read Supabase rows via SQL, read Railway logs, read git log. Don\'t guess from symptoms. Faster overall because most "bugs" turn out to be different from what symptoms suggested. Rule 26 paired with Rule 25: gather all evidence in turn 1, root-cause in turn 2, fix in turn 3. Never patch-test-patch cycle.' },
// ============ KNOWN PENDING / ISSUES ============
{ title: 'GROQ_API_KEY pending dev tier access', status: 'pending', effort: 'awaiting Groq', owner: 'Deepak Dudi', addedDate: '2026-04-30', phase: 'Pending', note: 'Groq API access pending — applied for developer tier but not yet approved. Once available, wire as Tier 2 of inference chain (between Cerebras and SambaNova). Will make 5-tier chain → 6-tier. Groq offers extremely fast inference (~750 tok/s on Llama models) at competitive cost. Currently inference chain works fine without Groq but adding it improves tier diversity and capacity.' },
{ title: 'VERCEL_TOKEN pending in Railway env', status: 'upcoming', effort: '10 min', owner: 'Deepak Dudi', addedDate: '2026-05-05', phase: 'Pending', note: 'Generate Vercel API token in Vercel dashboard → add to Railway env vars on main DCS backend → enables Visual Editor single-page deploy feature (deploy a modified page back to Vercel without rebuilding the whole site). Currently blocking Visual Editor Phase 2 launch.' },
{ title: 'vercel_project_id column on agent_sessions table', status: 'upcoming', effort: '15 min', owner: 'Deepak Dudi', addedDate: '2026-05-05', phase: 'Pending', note: 'ALTER TABLE agent_sessions ADD COLUMN vercel_project_id text. Required for Visual Editor to track which Vercel project each build deploys to, so single-page redeploys go to the right project. Combined with VERCEL_TOKEN env var (above), unlocks Visual Editor Phase 2.' },
],
tech: {
repo: 'See individual repo items above for all 6 repos',
stack: 'Backends: Railway + Express/TS · Frontends: Cloudflare Pages + static HTML · DB: Supabase (shared) · GPU: RunPod · Filecoin: Lighthouse',
hosting: 'Railway (3 backends) + Cloudflare Pages (3 frontends) + Hostinger (visual-editor) + Supabase (DB) + RunPod (compute)',
db_tables: 'Shared Supabase project with per-product prefixes: agents · compute_* · storage_* · session_messages · trd_inference_logs · vendor_usage_events',
env_vars: 'See individual env var items above for each backend',
recent_commits: '568c8ea · 6954cd6 · 6121845 · cbf0d97 · b5462ea · 466ee15 · afb7c88 · 2f828a1',
notes: 'Captured May 13, 2026 from live system confirmation with founder. Update this feature whenever infrastructure changes — repo move, hosting platform change, new env var added, etc.',
},
},
{
id: 'i-vendor-map', project: '03 · DCS PLATFORM INTERNALS', name: 'Vendor & Hosting Map',
kind: 'internal', priority: 'high', launchReady: 'yes',
desc: 'Every vendor, every service, every URL, every billing. The single source of truth for "what is hosted where + why + how much does it cost." Use this to evaluate consolidation opportunities and prevent paying duplicate vendors for the same thing.',
items: [
// ============ VERCEL ============
{ title: 'Vercel — hosts customer-built sites (PAID plan)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Vercel', note: 'Vercel Pro plan ($20/month). Hosts every site that DCS Platform users generate via /app22. Each customer build creates a new Vercel project via the Vercel API (using VERCEL_TOKEN env var on Railway backend). Subdomain pattern: app-{slug}.dcsai.ai for each customer site, OR custom domains when customer brings their own. Why Vercel for customer sites: (1) handles React/Next.js builds natively (DCS agents output React for some templates), (2) per-deployment URLs for preview before publish, (3) instant rollback per project, (4) edge function support for any dynamic logic in customer sites.' },
{ title: 'Vercel — customer site subdomains', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Vercel', note: 'Pattern for customer site URLs: app-{customer-slug}.dcsai.ai (e.g. app-bean-and-brew.dcsai.ai). Configured via Vercel custom domains feature with CNAME wildcard in Cloudflare DNS pointing back to Vercel. When customer wants their own custom domain (e.g. beanandbrew.com), they add a CNAME in their DNS pointing to Vercel and DCS adds it as a custom domain in the Vercel project.' },
{ title: 'Vercel — separate from DCS\'s own frontends', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Vercel', note: 'IMPORTANT distinction: Vercel ONLY hosts customer sites (the sites users build via DCS Platform). DCS\'s own marketing sites (dcsai.ai main site, compute.dcsai.ai, storage.dcsai.ai, sovereign.dcsai.ai) live on Cloudflare Pages. This separation is intentional — customer sites benefit from Vercel\'s build features; DCS\'s own sites are static and benefit from Cloudflare\'s free tier + speed. Do not move DCS\'s own sites to Vercel — wastes money.' },
// ============ CLOUDFLARE PAGES ============
{ title: 'Cloudflare Pages — hosts all DCS frontends (PAID plan)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-03-20', phase: 'Cloudflare', note: 'Cloudflare paid plan. Hosts ALL of DCS\'s own static frontends. Per-site setup: each Cloudflare Pages project connected to a GitHub repo, auto-deploys on push to main branch. Currently hosting: (1) trd-frontend → dcsai.ai + all subpaths like /app22, (2) trd-storage-frontend → storage.dcsai.ai, (3) trd-cn-frontend → compute.dcsai.ai, (4) sovereign-trdn-io → sovereign.dcsai.ai (B1+C3+Sentinel). Global CDN means <1s page load worldwide. SSL auto-issued. DDoS protection included.' },
{ title: 'Cloudflare DNS — manages dcsai.ai (transferred from elsewhere)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-03-20', phase: 'Cloudflare', note: 'dcsai.ai domain transferred TO Cloudflare from previous registrar for DNS + management consolidation. Records: dcsai.ai → Cloudflare Pages (trd-frontend), storage.dcsai.ai → CF Pages (storage-frontend), compute.dcsai.ai → CF Pages (cn-frontend), sovereign.dcsai.ai → CF Pages (sovereign-frontend), api.dcsai.ai → Railway backend, api.storage.dcsai.ai → Railway (storage backend), api.compute.dcsai.ai → Railway (DCS-CN backend). All records orange-cloud proxied for DDoS protection + analytics.' },
{ title: 'Cloudflare — DDoS + WAF + bot protection included', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-03-20', phase: 'Cloudflare', note: 'Cloudflare paid plan includes: DDoS protection (unlimited bandwidth, even at attack volumes), WAF (Web Application Firewall with managed rule sets), Bot Fight Mode (blocks malicious bots), Cache rules (custom TTL per file type), Page Rules (URL redirects + cache overrides), Email Routing (for dcsai.ai email forwarding — currently unused, opportunity below). These features alone justify the paid plan even before considering Pages hosting.' },
// ============ HOSTINGER ============
{ title: 'Hostinger — hosts trdnetwork.com domain + email', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-03-15', phase: 'Hostinger', note: 'Hostinger hosts the trdnetwork.com domain (defensive registration to prevent brand squatting on alternate spelling of dcsai.ai) PLUS the email service for that domain. Email mailboxes accessible via Hostinger webmail or IMAP. Currently this is the only thing Hostinger is being used for. Cost: ~$3-5/month for the email + domain bundle. CONSOLIDATION OPPORTUNITY: trdnetwork.com domain could move to Cloudflare (already managing dcsai.ai there) and email could move to Cloudflare Email Routing (free) — eliminate Hostinger entirely. See Vendor Consolidation analysis below.' },
{ title: 'Hostinger — hosts visual-editor.html (legacy)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-01', phase: 'Hostinger', note: 'Visual-editor.html (DCS\'s WYSIWYG editor surface, split from app22.html on May 1) currently hosted on Hostinger separately from main frontend. Reason originally: CORS configuration easier on Hostinger at the time for iframe-based preview. Should be migrated to Cloudflare Pages (where the rest of trd-frontend lives) for consolidation — currently the only DCS frontend not on Cloudflare. Migration effort: ~1 hour. Saves: nothing on hosting cost (Hostinger already paid for email), but consolidates ops + makes deploys consistent.' },
// ============ NAMECHEAP ============
{ title: 'Namecheap — registrar for trdn.ai only', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Namecheap', note: 'Namecheap is the registrar (NOT host) for trdn.ai domain. Just registration, no hosting. Calendar reminder set for April 2028 renewal. Cost: ~$80-100/year for .ai TLD (premium TLD pricing). CONSOLIDATION OPPORTUNITY: trdn.ai could transfer to Cloudflare Registrar (which is at-cost — Cloudflare doesn\'t mark up domain registration). Migration effort: ~15 minutes (initiate transfer at Cloudflare, approve at Namecheap). Saves: small amount on renewal fees (Cloudflare at-cost is typically 5-15% cheaper) + management consolidation. Risk: 5-7 day transfer window during which domain is in transit.' },
// ============ RAILWAY ============
{ title: 'Railway — hosts all 3 backends (PAID plans)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Railway', note: 'Railway hosts all 3 DCS backends as separate projects: (1) trd-backend-clean (main DCS Platform backend with agent_orchestrator.ts, REST routes for /app22) → api.dcsai.ai. (2) trd-storage (Storage backend, Lighthouse SDK integration) → api.storage.dcsai.ai. (3) trd-cn-backend (DCS-CN Compute Network backend with worker job dispatch) → api.compute.dcsai.ai. Each Railway project has independent: env vars, deploy logs, Sentinel checks, scaling configuration, billing line item. Auto-deploys on git push to main. Pre-deploy hook runs Sentinel marker check on main backend. NO real alternative to Railway at this scale — keep.' },
{ title: 'Railway — pricing scales per project usage', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Railway', note: 'Railway pricing model: each project billed based on actual usage (RAM hours + CPU hours + egress bandwidth). Typical DCS backend: $5-20/month per project depending on traffic. Total Railway bill: $30-60/month across 3 backends. Reasonable for the value delivered (zero ops, fast deploys, good logs). No consolidation possible — backends need to stay separate because they serve different domains and have different dependency requirements.' },
// ============ SUPABASE ============
{ title: 'Supabase — shared database for all products (FREE tier currently)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-03-20', phase: 'Supabase', note: 'Single Supabase project shared across DCS Platform, DCS Storage, DCS Compute, DCS Sovereign. Currently on FREE tier — sufficient for pre-launch volume. Database includes: auth.users (unified across all products via single sign-on), agents table (641 rows for orchestrator), compute_* prefixed tables (DCS-CN data), storage_* prefixed tables (DCS Storage), session_messages (chat history), trd_inference_logs, vendor_usage_events (Phase 43.3 cost tracking). RLS policies enforce per-user data isolation. Will need to upgrade to Pro tier ($25/mo) at ~5000 active users or 8GB DB size — whichever comes first. Critical service, no alternative.' },
// ============ RUNPOD ============
{ title: 'RunPod — GPU compute for Inference + Pod Mode', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-08', phase: 'RunPod', note: 'RunPod provides underlying GPU infrastructure. Two distinct uses: (1) DCS GPU (reseller) — customer pods spun up on-demand for their workloads (per-second billing passed through). (2) DCS Inference — dedicated H200 pod running vLLM + Qwen3-32B-FP8 for free Tier 0 inference. Currently $200-400/mo on the DCS-owned H200 pod plus pass-through cost for customer GPU rentals. NOWPayments handles crypto-on-ramp for renters; Stripe for fiat. No alternative at this scale — keep.' },
// ============ LIGHTHOUSE / FILECOIN ============
{ title: 'Lighthouse — Filecoin storage layer (PAID for DCS Storage)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-05-07', phase: 'Lighthouse', note: 'Lighthouse SDK 0.4.5 provides the Filecoin permanence layer that powers DCS Storage. Per-file pin cost + small monthly account fee. Currently spending ~$30-60/month depending on volume. Lighthouse 402 gateway issue (Bug in DCS Storage section) — public gateway returns 402 Payment Required at scale, needs self-hosted IPFS gateway ($6/mo DigitalOcean) OR Cloudflare Workers proxy as fix. Critical for Storage product, no alternative.' },
// ============ ANTHROPIC + CEREBRAS + OTHER AI VENDORS ============
{ title: 'Anthropic API — premium agents (Claude)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'AI vendors', note: 'Anthropic provides Claude API for the highest-quality agents in the DCS pipeline (Architect Agent, QA Agent, sometimes DevOps Agent). Most expensive per-token but highest quality. Used selectively — Phase 27a moved DevOps off Claude to Cerebras to save ~$0.049 per build. Now primarily used for Architect Agent (planning the build) which justifies the cost. Daily spend: $50-100 depending on build volume.' },
{ title: 'Cerebras API — primary inference (5 API keys)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-18', phase: 'AI vendors', note: 'Cerebras provides extremely fast inference at low cost. Uses 5 API keys for combined TPM (tokens per minute) ceiling of 15M tokens/min. Phase 27a-27c moved most agents to Cerebras as Tier 1 in the inference chain, dropping cost per build from $0.76 to $0.04 (95% reduction). Currently the workhorse of DCS\'s inference. Monthly spend: $50-150 depending on volume. KNOWN ISSUE: when Tier 3 chunk 2 exhausts all 5 keys\' TPM, falls back to slow DeepSeek (366s). Fix planned: better load balancing OR more keys.' },
{ title: 'Other AI vendors — Together, Gemini, DeepSeek, Groq pending, SambaNova', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-30', phase: 'AI vendors', note: 'Tier 2 onward in inference chain: Together (GPT-OSS 120B for heavy agents), Gemini Flash-Lite (fast fallback), DeepSeek (cheapest final fallback), SambaNova (Tier 2 added Phase 41.2), Groq (pending dev tier approval, will add as Tier 1.5). Each used selectively for different agent types. Combined monthly spend: $20-50. Vendor tracking via Phase 43.3 vendor_usage_events table.' },
// ============ RESEND ============
{ title: 'Resend — transactional email API', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-02', phase: 'Resend', note: 'Resend handles ALL outbound transactional email across DCS products: contact form submissions on dcsai.ai, signup confirmations, password resets, DCS Storage waitlist double-opt-in, vendor notification emails. Sends from no-reply@dcsai.ai. Free tier: 3000 emails/month — currently sufficient. Will need paid plan ($20/mo for 50K emails) once user volume scales post-launch. Note: this is OUTBOUND only — INBOUND email for the trdnetwork.com domain still goes through Hostinger (consolidation opportunity).' },
// ============ STRIPE ============
{ title: 'Stripe — payment processing (multiple accounts)', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-01', phase: 'Stripe', note: 'Stripe handles fiat payments. Multiple Stripe accounts planned for separation: (1) DCS Compute account (renter payments for GPU rentals — already live). (2) DCS Platform account (subscriptions $19/$49/$99 — pending Stripe Connect approval). Separation keeps per-entity revenue clean for accounting. Standard Stripe fees: 2.9% + $0.30 per transaction (US), varies by country. No alternative for fiat payments at our scale.' },
// ============ NOWPAYMENTS ============
{ title: 'NOWPayments — crypto payment processing', status: 'completed', effort: 'in production', owner: 'Deepak Dudi', addedDate: '2026-04-09', phase: 'NOWPayments', note: 'NOWPayments handles crypto-on-ramp (USDT, USDC) for DCS Compute renters and potentially DCS Storage. Stablecoins settle to USD in bank account daily, avoiding crypto volatility exposure. Fee: 0.5-1% per transaction. Critical for global users and crypto-friendly DePIN audience. Cheaper than international wire fees. No real alternative — Coinbase Commerce has higher fees, smaller crypto coverage.' },
// ============ POSTHOG (FUTURE) ============
{ title: 'PostHog — analytics (decision made, not yet implemented)', status: 'upcoming', effort: '1 hr setup', owner: 'Deepak Dudi', addedDate: '2026-04-18', phase: 'PostHog', note: 'PostHog selected as analytics backend (decision Phase 0 of DCS Analytics product). EU region recommended for GDPR. Will be embedded on every DCS-built site (per-site analytics product) PLUS the /app22 dashboard itself (user behavior, funnel analysis). Free tier sufficient for pre-launch; paid tier ($0-450/mo) at scale. NOT YET IMPLEMENTED — listed as upcoming launch-blocker in DCS Platform.' },
// ============ VENDOR CONSOLIDATION ANALYSIS ============
{ title: 'CONSOLIDATION 1: Hostinger → Cloudflare (DEFERRED — Hostinger prepaid through 2027)', status: 'upcoming', effort: '2-4 hrs (when ready)', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Consolidation', note: 'DECISION (May 13): DEFERRED — Hostinger is PREPAID for 3 years (1+ year remaining as of May 2026). No financial benefit to migrating now since the cost is already sunk. Will reconsider when Hostinger renewal approaches (~Q2 2027). CURRENT STATE: Hostinger hosts trdnetwork.com domain + email + visual-editor.html. NOTED FOR FUTURE: when Hostinger renewal date approaches, evaluate (1) transfer trdnetwork.com to Cloudflare Registrar, (2) set up Cloudflare Email Routing (free) to forward email to main inbox, (3) move visual-editor.html to Cloudflare Pages. NET RESULT when done: eliminate Hostinger entirely, save ~$36-60/yr ongoing, consolidate to one vendor for domains + DNS + email + static hosting. Set calendar reminder for 3 months before Hostinger renewal to start the migration.' },
{ title: 'CONSOLIDATION 2: Namecheap → Cloudflare for trdn.ai (NOTED for future consideration)', status: 'upcoming', effort: '15 min + 5-7 days transfer', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Consolidation', note: 'DECISION (May 13): NOTED for future consideration alongside Hostinger consolidation. Will revisit in coming weeks/months. Currently: Namecheap is registrar for trdn.ai with renewal April 2028. PROPOSAL TO REVISIT: Transfer trdn.ai to Cloudflare Registrar (at-cost domain pricing, typically 5-15% cheaper than Namecheap on .ai TLD). Benefit: all domains in one dashboard, one renewal calendar, one vendor. EFFORT: 15 min to initiate, 5-7 days transfer window. RISK: very low. SAVINGS: small ($5-15/yr). Bigger value is consolidation simplicity. Will batch with Hostinger consolidation when that becomes timely (Q2 2027).' },
{ title: 'CONSOLIDATION 3: Keep Vercel for customer sites — DECISION LOCKED', status: 'completed', effort: 'decision', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Consolidation', note: 'DECISION (May 13): LOCKED — KEEP Vercel for customer sites. Founder confirmed: "Keep Vercel for customer sites — faster builds, better for AI-generated React apps." This decision is final unless one of two conditions is met: (a) customer site volume scales to where Cloudflare\'s unlimited free tier would save $200+/month, AND (b) DCS agents reliably emit static HTML for 80%+ of builds (currently mixed React/Next.js + static). Until then, $20/month Vercel Pro is reasonable for what it delivers: native React/Next.js builds, per-deployment URLs for QA preview, instant rollback per project, mature programmatic deploy API used by agent_orchestrator.ts.' },
{ title: 'CONSOLIDATION 4: Do NOT consolidate Railway, Supabase, RunPod, Lighthouse, AI vendors', status: 'completed', effort: 'analysis', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Consolidation', note: 'These vendors have NO good alternatives at our scale: (1) Railway — best Node.js hosting with good DX. Render is comparable but switching cost not justified. (2) Supabase — Postgres + auth + storage + realtime in one. Firebase/PlanetScale require multi-vendor stacks. (3) RunPod — only real DePIN GPU marketplace with good API. Lambda/Vast.ai are alternatives but RunPod relationship is built. (4) Lighthouse — primary Filecoin SDK; no real alternative for permanence-as-a-service. (5) AI vendors (Anthropic, Cerebras, etc.) — these ARE the inference layer, not "vendors" to consolidate from. Each tier serves a specific cost/capability point. CONCLUSION: leave all 5 alone. Focus consolidation only on Hostinger (eliminate when prepay expires) and Namecheap (batch with Hostinger move).' },
// ============ MONTHLY VENDOR SPEND SUMMARY ============
{ title: 'Monthly vendor spend summary (May 2026)', status: 'completed', effort: 'snapshot', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Cost summary', note: 'Approximate monthly recurring vendor spend (pre-launch, low volume): Vercel ~$20, Cloudflare Pages/Workers ~$20, Railway ~$30-60 (3 backends), Supabase $0 (free tier), Hostinger ~$3-5 (email + domain), Anthropic ~$50-100 (Claude API), Cerebras ~$50-150 (inference), Other AI vendors combined ~$20-50, Resend $0 (free tier), Lighthouse ~$30-60, RunPod ~$200-400 (DCS Inference H200 + customer GPU rentals). TOTAL: ~$420-870/month. POST-LAUNCH at scale: Supabase will move to Pro $25/mo, Resend to $20/mo, Anthropic/Cerebras scale with usage. Forecast at 500 active users: ~$1.5k-3k/month vendor costs. Critical to maintain >70% gross margin (revenue - vendor costs - infrastructure).' },
{ title: 'Calendar reminders for all renewals', status: 'upcoming', effort: '30 min', owner: 'Deepak Dudi', addedDate: '2026-05-13', phase: 'Operations', note: 'Set Google Calendar reminders for: dcsai.ai renewal (Cloudflare, annually), trdn.ai renewal (Namecheap → Cloudflare after transfer, April 2028), trdnetwork.com renewal (Hostinger → eliminate via consolidation), domain transfer authorizations (in case any are needed). Also recurring monthly reminder to check: Anthropic spend, Cerebras balance, Railway invoice, Vercel invoice. Helps prevent surprise depletion of pre-paid credits or missed renewals.' },
],
tech: {
stack: 'Multi-vendor: Vercel (customer sites) · Cloudflare (DCS frontends + DNS + email routing future) · Railway (3 backends) · Supabase (DB) · Hostinger (email — to eliminate) · Namecheap (trdn.ai only — to migrate) · RunPod (GPU) · Lighthouse (Filecoin) · Resend (outbound email) · Stripe (fiat) · NOWPayments (crypto) · 5+ AI vendors (Anthropic, Cerebras, Together, Gemini, DeepSeek, SambaNova, Groq pending)',
hosting: 'See individual vendor items above for what each handles',
notes: 'Updated May 13, 2026. 3 consolidation opportunities identified: (1) eliminate Hostinger by moving domain + email to Cloudflare, (2) move trdn.ai from Namecheap to Cloudflare, (3) move visual-editor.html from Hostinger to Cloudflare Pages. Keep Vercel for customer sites, keep all other vendors as-is.',
},
},
// ============== 04 · STRATEGIC / FUTURE PRODUCTS (10–15 launches) ==============
{
id: 's-restaurant-os', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Restaurant OS',
kind: 'product', priority: 'high', launchReady: 'partial',
desc: 'Vertical AI Suite v1. Same engine as DCS Platform with restaurant-specific dashboard (Menu, Reservations, 86-list, OrderUp) + OpenTable integration. $149/mo vs generic $49. 80% of platform groundwork shipped.',
items: [
{ title: 'Restaurant audit complete', status: 'completed', effort: 'done', owner: 'Deepak Dudi', addedDate: '2026-05-08', note: 'Comprehensive audit of the restaurant industry: customer interviews (5-10 restaurant owners), competitor analysis (OpenTable, Resy, SevenRooms, Fresha, Square for Restaurants), workflow mapping (how restaurants currently handle: website, booking, online ordering, customer database, marketing, loyalty). Output: feature priority list for Restaurant OS based on real pain points.' },
{ title: '6 templates locked + copy bank ~70% tightened', status: 'completed', effort: 'done', owner: 'Deepak Dudi', addedDate: '2026-05-08', note: 'Six restaurant templates locked (noma-editorial, gucci-osteria, girl-goat, gramercy, sunday-brooklyn, nobu-signature). Copy bank — pre-written marketing copy variations for restaurant industry — at ~70% tightening (refining tone, removing generic phrases, adding restaurant-specific language). Tightening means each piece reviewed for: industry-appropriate vocabulary, conversion-optimized phrasing, no AI-generated tells.' },
{ title: 'Vertical dashboard prototype', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', deps: 'Edit Website reinject', addedDate: '2026-05-10', note: 'Prototype of Restaurant OS dashboard: bookings calendar, online orders queue, customer database with order history, menu management, daily sales summary, integration with POS (if any). Designed for restaurant owner who uses it on a phone behind the counter — must be touch-friendly, fast, no fluff. Goal: feels like a purpose-built restaurant tool, not a generic SaaS dashboard.' },
{ title: 'OpenTable / SevenRooms integration', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', deps: '50+ paying restaurants', addedDate: '2026-05-10', note: 'Integration path with major booking platforms (OpenTable for US/global, SevenRooms for premium restaurants). Bidirectional sync: bookings created in OpenTable show up in DCS Restaurant OS dashboard, bookings created in DCS push back to OpenTable. Allows restaurants to migrate gradually without losing existing customer pipeline. Implementation: REST API integration, OAuth for restaurant authorization.' },
{ title: 'Restaurant OS pricing $149/mo + GTM', status: 'upcoming', effort: '1–2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Restaurant OS pricing: $149/mo per location (single restaurant), $249/mo for multi-location chains (3-10 locations), custom enterprise pricing for 10+ locations. GTM (go-to-market) strategy: direct outreach to UAE/Dubai restaurants first (founder is based there, can do in-person demos), then expand to India, then global. Target: 50 paying restaurants in 90 days.' },
],
},
{
id: 's-salon-os', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Salon OS',
kind: 'product', priority: 'medium', launchReady: 'no',
desc: 'Vertical AI Suite v2. Salon-specific: consultation booking, membership, add-on services, Booksy/Square Appointments integration. Launch after Restaurant OS profitable 3+ months.',
items: [
{ title: 'Salon vertical audit', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', deps: 'Restaurant OS profitable', addedDate: '2026-05-10', note: 'Comprehensive audit of salon/spa/beauty industry: customer interviews with salon owners (5-10), competitor analysis (Fresha, Booksy, Square Appointments, Mindbody), workflow mapping (booking, stylist management, retail, customer reminders, loyalty programs). Output: feature priority list for Salon OS. Similar audit pattern as Restaurant OS but salon-specific pain points are different (stylist scheduling complexity, retail tracking, no-show problem).' },
{ title: 'Salon copy bank + 6 templates', status: 'upcoming', effort: '2 wk', owner: 'Cowork', addedDate: '2026-05-10', note: 'Salon-specific copy bank (marketing copy variations refined for beauty industry) plus 6 salon templates. Templates target: hair salons, nail salons, beauty spas, barbershops, brow bars, lash studios. Each template includes: service menu structure, stylist profiles section, before/after gallery (critical for visual industries), booking widget, customer reviews integration.' },
{ title: 'Booksy / Square Appointments integration', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Integration with Booksy (popular in EU/global) and Square Appointments (popular in US). Bidirectional sync of appointments. Allows salons to migrate gradually without losing existing customer pipeline. Implementation: REST API integration with each platform\'s appointment endpoints. OAuth for salon authorization. Estimated 1 week per integration.' },
],
},
{
id: 's-clinic-os', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Clinic OS',
kind: 'product', priority: 'medium', launchReady: 'no',
desc: 'Vertical AI Suite v3. Healthcare/dental: patient intake, Practice Fusion integration, HIPAA-aware copy. Pairs naturally with Sovereign Mode for compliance.',
items: [
{ title: 'Clinic vertical audit', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', deps: 'Salon OS shipped', addedDate: '2026-05-10', note: 'Comprehensive audit of medical/clinical industry: dentists, physiotherapy, general practice, mental health, alternative medicine. Higher regulatory burden than Restaurant/Salon (HIPAA in US, similar in UAE/India). Workflow includes: appointment booking, patient records, prescription tracking, lab results, billing, insurance claims. Slower path because clinical workflows require deeper customization + trust-building. Likely Year 2 product.' },
{ title: 'Healthcare copy bank + 6 templates', status: 'upcoming', effort: '2 wk', owner: 'Cowork', addedDate: '2026-05-10', note: 'Healthcare-specific copy bank with appropriate medical industry tone (professional, trust-building, regulation-compliant — no claims that imply medical advice). Six templates targeting: dental practice, physiotherapy clinic, general practice, mental health practice, dermatology clinic, alternative medicine practice. Each template carefully reviewed for medical accuracy + compliance.' },
{ title: 'Practice Fusion integration', status: 'completed', testResult: 'pass', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Practice Fusion is a popular cloud-based EMR (Electronic Medical Records) system for small clinics. Integration allows: clinic builds website with DCS, manages patient records in Practice Fusion, single sign-on between the two. Customer doesn\'t need to migrate EMR data to use DCS. Implementation: Practice Fusion API integration, OAuth for clinic authorization.' },
{ title: 'HIPAA + Sovereign Mode hookup', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', deps: 'Sovereign deployment package', addedDate: '2026-05-10', note: 'HIPAA-compliant infrastructure for US clinics: encrypted-at-rest patient records, encrypted-in-transit, audit log of every record access, role-based access controls, BAA (Business Associate Agreement) with all sub-processors. Also: Sovereign Mode option for clinics that want fully on-premise deployment — no patient data ever leaves their infrastructure. Premium pricing tier for compliance-sensitive customers.' },
],
},
{
id: 's-business-os', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'DCS Business OS',
kind: 'product', priority: 'medium', launchReady: 'no',
desc: 'One dashboard for the whole business — unified inbox + CRM lite + outbound marketing + finance lite + support automation + reviews. $199–499/mo. Replaces 3–5 SaaS subscriptions per customer.',
items: [
{ title: 'Unified inbox + CRM lite (P1)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 1 of Business OS: unified inbox (incoming WhatsApp, email, contact form, social DMs all in one view) + lightweight CRM (contact database, deal pipeline, task management). Not competing with Salesforce/HubSpot on depth — competing on ease of use and price for SMBs that find those tools overkill. Goal: replace Excel spreadsheets that most small businesses use for customer tracking.' },
{ title: 'Outbound marketing module (P2)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 2: outbound marketing — email campaigns (Resend integration), WhatsApp broadcast (subject to compliance), social media scheduler (post to Instagram/Facebook/Twitter from one dashboard), customer segmentation (filter contacts by behavior/tags). Replaces Mailchimp + Buffer + WhatsApp Marketing tool stack for SMBs.' },
{ title: 'Finance lite — invoicing + payments (P3)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 3: invoicing module — generate professional invoices in 30 seconds, send via email with Stripe payment link, recurring invoices for retainers, automatic overdue reminders. Replaces QuickBooks + separate Stripe for the invoicing flow. Includes basic expense tracking (upload receipt, categorize, run report). Not full accounting — handoff to accountant for tax filing.' },
{ title: 'Support automation (P4)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', deps: 'F6 Memory', addedDate: '2026-05-10', note: 'Phase 4: customer support tools — shared inbox for support team, auto-routing of tickets by topic, canned responses, knowledge base / FAQ for self-service, customer satisfaction surveys. Replaces Intercom + Zendesk for small teams. Lighter feature set but enough for most SMBs.' },
{ title: 'Reviews + reputation (P5)', status: 'upcoming', effort: '1–2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 5: review management — request reviews from happy customers automatically post-purchase, monitor Google/Facebook reviews, respond from one dashboard, get alerted to new reviews, showcase best reviews on website automatically. Critical for service businesses where Google reviews drive new customer acquisition. Replaces BirdEye / Podium.' },
{ title: 'Business OS pricing $199–499/mo (P6)', status: 'upcoming', effort: '1–2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 6: pricing tiers and GTM. Starter $199/mo (CRM + Inbox), Growth $349/mo (+ marketing + invoicing), Pro $499/mo (+ support + reviews + analytics). Annual billing = 2 months free. Target: 100 paying customers within 90 days of launch.' },
],
},
{
id: 's-framework-export', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'DCS Framework Export',
kind: 'product', priority: 'medium', launchReady: 'no',
desc: 'HTML → React / Vue / Node / PHP exporters. Pro+ tier $149/mo OR $49 per-export. Big wedge — freelancers cannot do this. Customer owns the codebase.',
items: [
{ title: 'Phase 1: React converter (post-build agent)', status: 'upcoming', effort: '2–3 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 1 of Framework Export: convert DCS\'s HTML output to React/Next.js project structure. Pages become app/ directory routes, sections become components, styles become CSS modules or Tailwind config. Generates package.json, next.config.js, ready to deploy. Output is opinionated (matches React best practices) — produces idiomatic React that a developer would actually maintain.' },
{ title: 'Phase 2: Vue + Node/Express (shared AST)', status: 'upcoming', effort: '2–3 wk', owner: 'Deepak Dudi', deps: 'P1 AST layer', addedDate: '2026-05-10', note: 'Phase 2: Vue/Nuxt export (similar to React, different framework) + Node/Express backend export (for sites that need server-side logic). Shared AST (Abstract Syntax Tree) layer means the conversion logic is reusable across framework targets — easier to add new frameworks in future. Implementation: parse HTML to AST, transform AST per target framework, serialize.' },
{ title: 'Phase 3: PHP / Laravel', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', deps: '≥30 PHP requests', addedDate: '2026-05-10', note: 'Phase 3: PHP and Laravel export. Targets the large WordPress/Laravel developer ecosystem who want DCS\'s quality without leaving their preferred stack. Lower priority than React/Vue but significant market (PHP still ~75% of web). Output: clean Laravel project structure or WordPress theme.' },
{ title: 'Phase 4: Export pricing + UI', status: 'upcoming', effort: '1–2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 4: pricing and UI for framework export. Pricing options: bundle into Pro tier free (positions as \'no lock-in\' marketing wedge), OR sell per-export at $19-49 (revenue from each conversion). Initial bet: bundle into Pro tier as feature. UI: \'Export\' button in /app22, choose framework, click to download zip with all files.' },
{ title: 'Phase 5: Template marketplace auto-export', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 5: every template in the Marketplace (when launched) gets auto-exported to all framework variants. Marketplace shoppers see \'Available as: React, Vue, PHP, HTML\' before buying. Increases marketplace value because templates are framework-agnostic. Implementation: CI/CD pipeline that runs export on every new template.' },
],
},
{
id: 's-marketplace', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'DCS Marketplace',
kind: 'product', priority: 'low', launchReady: 'no',
desc: 'Two-sided marketplace: agencies/freelancers list DCS-built templates + agents for sale. Revenue split 70/30. Premature before 10k+ users — re-evaluate M9+.',
items: [
{ title: 'Internal agent versioning + isolation (P1)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', deps: '10k+ users', addedDate: '2026-05-10', note: 'Phase 1: technical foundation for marketplace. Each agent (designer-agent, content-agent, etc.) gets versioned — version 1.0, 1.1, etc. New community-submitted templates can use specific agent versions for stability. Isolation: marketplace templates run in sandboxed environment so a bad template can\'t break the main platform. Engineering work, no user-facing value yet.' },
{ title: 'Submission + review pipeline (P2)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 2: submission flow for community contributors. Designer uploads .zip with HTML+CSS+JS, fills metadata (industry, style tags, demo URL, screenshots), passes automated quality check (lighthouse score, mobile responsive, no broken links), then manual review by DCS team. Approved templates appear in template picker. Time from submission to listing: target <72 hours.' },
{ title: 'Marketplace UI + Stripe Connect splits (P3)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 3: customer-facing marketplace browse/buy UI in /app22. Filters by industry, price, style, rating. Stripe Connect handles payment splitting: 70% to template creator, 30% to DCS (industry-standard creator royalty). Revenue tracked per creator + paid out monthly via Stripe Connect.' },
{ title: 'Discovery + recommendation (P4)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', deps: '100+ agents', addedDate: '2026-05-10', note: 'Phase 4: smart discovery + recommendation engine. \'Customers who liked this template also liked...\', \'Trending in restaurants this week\', \'Most-installed in your industry\'. Increases marketplace activity and creator revenue. Implementation: collaborative filtering, behavior tracking, rankings algorithm.' },
],
},
{
id: 's-agency', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'DCS Agency · White-Label',
kind: 'product', priority: 'medium', launchReady: 'no',
desc: 'Agencies resell DCS builds under their own brand. 95% margin for agency. DCS handles build, agency handles client. B2B distribution multiplier — one agency = 20–50 clients without DCS doing direct sales.',
items: [
{ title: 'White-label dashboard (custom domain + branding)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Full white-label dashboard for agencies. Includes: custom domain for the agency\'s platform login (clients see agency-platform.com, not dcsai.ai), agency logo throughout interface, agency colors and fonts in client builds, custom email templates from agency address. The DCS brand is invisible to end clients. Agency keeps 100% of client-facing relationship.' },
{ title: 'Agency-specific pricing + billing', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Agency tier pricing: $499/mo for 10 client sites, $799/mo for 25 client sites, $1499/mo for 100 client sites. Agencies pay DCS wholesale and resell to clients at their own markup (typically 3-5x). Strong unit economics for DCS — one paying agency drives 10-100+ active sites. Billing: monthly subscription via Stripe, agency manages their own client billing separately.' },
{ title: 'First 5 agency partners signed', status: 'upcoming', effort: 'ongoing', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Target: first 5 agency partners signed within 60 days of launch. Acquisition: direct outreach to web design agencies in UAE/India (founder\'s network), Founders Slack/Discord communities, content marketing about \'building 100 sites with one tool\'. Each agency partner = ~$500 MRR + 10-50 active sites = $5k-25k worth of MRR pipeline.' },
],
},
{
id: 's-partner', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'DCS Partner Program',
kind: 'product', priority: 'low', launchReady: 'no',
desc: 'Freelancer affiliate program. 20–30% recurring commission for 12 months per referred customer. Turns freelancers who were replaced by DCS into DCS distributors.',
items: [
{ title: 'Referral dashboard + tracking links', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Partner-facing dashboard showing: unique referral link, traffic + conversion metrics, list of referred customers, commission earnings (current month + lifetime), payout history (paid monthly via Stripe Connect). Self-service so partners can monitor their own performance without contacting support. Tracking link uses signed token to attribute referrals correctly.' },
{ title: 'Stripe Connect payout flow', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Monthly payout flow via Stripe Connect: at end of each month, calculate commissions earned per partner (based on subscribed referred customers), trigger Stripe Connect payout to partner\'s bank account. Automated — no manual processing. Includes commission statement (CSV export) for partner\'s records. Tax forms (1099 for US partners) generated annually.' },
{ title: 'Commission rules engine (20–30% × 12 months)', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Commission rules: 30% recurring commission for first 12 months of customer\'s subscription, then 0% (lifetime would be too expensive long-term). Higher than typical 20% to attract serious partners. Anti-fraud: partners cannot refer themselves; multiple accounts from same IP get flagged. Commission only on PAID plans (free tier referrals not commissioned). Cookie window: 90 days from first click.' },
],
},
{
id: 's-enterprise-api', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'DCS Enterprise API',
kind: 'product', priority: 'low', launchReady: 'no',
desc: 'REST API for agencies building 100+ sites/month programmatically. Custom rate limits, dedicated support, SLA. $999/mo minimum. Scales revenue non-linearly — one enterprise = 50–100 individual subs.',
items: [
{ title: 'Public API spec + auth (API keys)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Public REST API documented via OpenAPI 3.0 spec. Authentication via per-customer API keys (issued in customer dashboard). Endpoints: create site, list sites, get site details, update content, publish, get analytics, webhook config. SDKs in TypeScript, Python, Ruby (auto-generated from OpenAPI spec). Comprehensive docs at api.dcsai.ai/docs.' },
{ title: 'Custom rate limits per customer', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Rate limits per customer tier: Starter (1000 calls/day, 100 calls/hour burst), Growth (10k/day, 1k/hour burst), Enterprise (100k+/day, custom burst rates). Rate limit headers (X-RateLimit-*) included in every response for client-side awareness. Customers hitting limits get clear error messages explaining how to upgrade.' },
{ title: 'SLA + dedicated support tier', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Enterprise tier includes Service Level Agreement: 99.9% uptime guarantee, response time guarantee for support tickets (2-hour for critical), dedicated support contact (named individual, not generic email), quarterly business reviews. Higher pricing ($4999/mo+) covers the higher-touch support model. Critical for landing larger customers who require SLAs in their procurement process.' },
],
},
{
id: 's-verified', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'DCS Verified Badge (F5)',
kind: 'product', priority: 'medium', launchReady: 'no',
desc: 'Quality + permanence cert on every DCS site. Storage stamps Filecoin proof, Platform renders badge. Public verification page. Trust signal for end visitors. M9–12.',
items: [
{ title: 'Quality scoring of completed builds', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', deps: 'smoke <20 majors', addedDate: '2026-05-10', note: 'Automated quality scoring for every completed DCS build. Checks: contact form actually works (test submission), mobile-responsive (no horizontal scroll on iPhone width), no broken internal links, SSL active, page load <3 seconds, lighthouse score >85. Sites scoring above threshold get verified badge. Sites below get a \'needs improvement\' notice with specific items to fix.' },
{ title: 'DCS Verified Badge UI on sites', status: 'upcoming', effort: '1–2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Trust badge displayed in footer of DCS-built sites: \'✓ DCS Verified — Built professionally\' with link back to dcsai.ai. Small but real marketing wedge: adds credibility for customer-facing sites, drives traffic back to DCS for new customer acquisition. Standard tier sites cannot remove the badge; premium tier ($19/mo or bundled with Pro) can hide it for white-label appearance.' },
{ title: 'Public verification page', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Public verification page at dcsai.ai/verified/[site-id]. Anyone can verify a site\'s authenticity. Shows: when site was verified, what checks passed, last quality score. Useful for: building trust with customers checking who built the site, SEO (inbound links from DCS), social proof for DCS marketing (\'verified 10,000+ sites\').' },
],
},
{
id: 's-memory', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Memory & Intelligence Layer (F6)',
kind: 'internal', priority: 'medium', launchReady: 'no',
desc: 'Long-term customer memory + industry memory graph. Every restaurant build feeds the next. Self-improving moat — no funded team can replicate in 12 months. M3–6.',
items: [
{ title: 'Schema + storage (pgvector decision)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Decision: use pgvector (PostgreSQL extension) for embedding-based similarity rather than dedicated vector DB (Pinecone, Weaviate). Reason: already using Supabase Postgres for everything else, pgvector is good enough at our scale, avoids another infrastructure dependency. Schema: memory_embeddings table with vector column (1536 dims for OpenAI embeddings), metadata jsonb, user_id + project_id for scoping.' },
{ title: 'Identity resolution across channels', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Cross-channel identity resolution: when a customer interacts via WhatsApp, email, web chat, phone — figure out it\'s the same person. Required for cross-channel memory (\'we discussed pricing yesterday on WhatsApp\'). Implementation: probabilistic matching on phone, email, name. Privacy: identity resolution stays per-account (DCS customer A\'s identity graph stays separate from customer B\'s).' },
{ title: 'Retrieval into agent_orchestrator', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Wire memory retrieval into agent_orchestrator.ts so agents can access relevant memories during build. Example: building a second site for a customer → agent retrieves their preferences from first site (favorite colors, brand voice, target audience). Implementation: vector search in pgvector based on current context, top-K results passed to agent as context.' },
{ title: 'Industry Memory Graph proper', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', deps: '500+ builds per industry', addedDate: '2026-05-10', note: 'Year 2 vision: aggregated industry-level memory graph. Patterns extracted from all DCS-built sites in an industry (which sections work, which copy converts, which images get clicks). New builds in that industry inherit collective intelligence. Premium \'Prestige\' tier feature. Privacy: only AGGREGATED PATTERNS shared across accounts, never customer-specific data.' },
{ title: 'GDPR + CCPA compliance', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'GDPR (EU) + CCPA (California) compliance for memory features. User rights: access to all their stored memories, export of their data, deletion of their data, opt-out of memory features entirely. Implementation: API endpoints for data access + deletion, audit log of consent decisions, encrypted-at-rest storage, retention policies. Required for enterprise customers and EU users.' },
],
},
{
id: 's-predict', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Predictive Outcome Engine (F7)',
kind: 'internal', priority: 'medium', launchReady: 'no',
desc: 'Cheap GPU pre-flight (~$0.005) predicts quality BEFORE $0.16 build runs. Real margin lever — saves DCS inference cost on bad builds. M6–9.',
items: [
{ title: 'Pre-flight model (inputs → quality)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', deps: 'F1 quality scoring', addedDate: '2026-05-10', note: 'Predictive model: given user\'s prompt + chosen industry + content provided, predict the quality score of the build BEFORE running it. Flag low-confidence builds for human review or prompt refinement. Saves compute cost on low-quality builds. Implementation: train classifier on historical (input, output_quality) pairs from past builds.' },
{ title: 'Suggest-fix flow (never block)', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'When the model predicts a low-quality build, suggest fixes BEFORE the user submits: \'Add 2-3 more sentences to the about section for better content\', \'Upload your logo for better branding consistency\', \'Specify your target audience for sharper copy\'. Never blocks the user — just suggests. Removes \'why is my site bad?\' frustration after the fact.' },
{ title: 'User-facing savings dashboard', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Dashboard showing: \'DCS\'s predictive engine saved you $X by suggesting fixes that prevented Y low-quality rebuilds.\' Quantifies the value of the predictive feature in dollar terms. Important for premium tier justification — customers paying $99/mo want to see what they\'re getting beyond base features.' },
],
},
{
id: 's-multimodal', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Multimodal Content Engine',
kind: 'product', priority: 'medium', launchReady: 'partial',
desc: 'Image gen IN FLIGHT (covered by DCS Image Bank). Voiceover (ElevenLabs) + video gen (Runway/Sora) future. Brand kit aggregation for every site.',
items: [
{ title: 'AI image gen for hero + services', status: 'pending', effort: 'in flight', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'AI-generated images for hero sections and service cards on DCS-built sites. Replaces stock photos (which feel generic) with industry-specific custom imagery. Uses SDXL via Replicate (already integrated for Image Bank). Per-site cost: ~$0.30-0.50 in compute. Quality: indistinguishable from professional photography for most use cases.' },
{ title: 'Brand consistency layer (LoRAs)', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'LoRAs (Low-Rank Adaptations) trained on each customer\'s brand allow brand-consistent image generation. Example: a hotel brand uploads 20 of their photos → LoRA trained → all future AI images for that brand have consistent style/aesthetic. Premium feature — $19/mo add-on for brand-conscious customers. Compute-intensive but high willingness to pay.' },
{ title: 'Voiceover (ElevenLabs)', status: 'upcoming', effort: '1–2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'AI voiceover via ElevenLabs for video content + audio messaging on DCS-built sites. Customer types script → AI voices it in chosen voice. Useful for product demo videos, welcome messages on homepage, audio versions of long-form content (accessibility). Compute cost ~$0.10-0.30 per minute of audio. Premium add-on.' },
{ title: 'Video gen (Runway/Sora) — premium tier', status: 'upcoming', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'AI-generated short videos (10-60s) via Runway or Sora for product demos, homepage hero videos, social media content. Compute cost much higher than images (~$5-20 per minute). Premium tier only ($39/mo+) due to cost. Quality improving rapidly with new model releases. Year 2 feature once costs drop.' },
{ title: 'Brand kit aggregation panel', status: 'upcoming', effort: '1–2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Unified brand kit per customer: logo, colors, fonts, voice, image style. All AI generation features pull from this brand kit for consistency. Customer uploads logo + brand guide once → applies to all images, videos, audio, copy automatically. Replaces having to specify brand preferences for every generation. Foundation feature for the multimodal vertical.' },
],
},
{
id: 's-orch-split', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'F4 · agent_orchestrator Split',
kind: 'internal', priority: 'low', launchReady: 'na',
desc: 'Architectural refactor. 16,156 lines → modular files. DEFER until pain signal (compile >10s OR 2+ devs join). Today 2.4s compile, no urgency.',
items: [
{ title: 'Defer until pain signal hit', status: 'upcoming', effort: '3 days when triggered', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'agent_orchestrator.ts has grown to 16,156 lines. At some point it needs to be split into smaller files for maintainability — orchestrator-core.ts, orchestrator-agents.ts, orchestrator-deploy.ts, orchestrator-storage.ts. But: file currently works, splitting risks breaking Sentinel protected zones, and tests don\'t fully cover the orchestrator. Decision: defer until file exceeds 25k lines OR team grows to 3+ developers OR clear pain signal hits. Until then, accept the monolith.' },
],
},
// ============== 04 · HYBRID AUTOMATION ROADMAP (May 15 2026 research) ==============
// 10 features researched May 15 2026 across 3 web search passes — agentic-browser landscape,
// SMB AI tools, what is unsolved. Mapped against existing DCS code: many already have
// substrate (D4 Review Drafter, D3 VoC Digest, Journey Engine, POS integration shipped).
// Build order: #9 first (foundation), then 5/7/8-alt (unique moats), then 2/4/6 (hybrid),
// then 1 (catch-up bundle), then 10 (orchestrator). Skip/partner 3+8 as originally written.
{
id: 's-hybrid-roadmap', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid Automation Roadmap (10 features)',
kind: 'internal', priority: 'high', launchReady: 'no',
desc: '10 hybrid browser-agent + LLM features for Commerce OS, researched May 15 2026 across 3 web searches. 4 UNIQUE moats (#5/7/9/10), 3 hybrid distinctive (#2/4/6), 3 catch-up parity (#1/3/8). 9 of 10 live in OS, 1 (#5 Onboarding Bot) is Platform→OS handoff. Builds on top of existing D-features (D1/D3/D4 shipped, D6/D9/D10/D14/D15/D16 in roadmap). 11 new tables, 8 new workers, 3–4 months focused work across Claude+Cowork sessions. Defensible window 12–18 months before enterprise tools move down-market.',
items: [
{ title: 'Research basis — 3 web search passes May 15 2026', status: 'completed', effort: 'done', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Three-pass research established the agentic-browser landscape: Browser Use (89.1% WebVoyager benchmark), Stagehand, Skyvern, Cloudflare Browser Run, Anthropic Claude for Chrome, Opera Neon, Perplexity Comet, ChatGPT Atlas, Microsoft Copilot Studio, Vercel Agent Browser, Browserbase, Bright Data Agent Browser. OpenAI Operator shut down Q1 2026 — being first doesn\'t mean winning. Browser-agent INFRASTRUCTURE is saturated; the moat is vertical SMB integration. Amazon vs Perplexity injunction (March 2026) sets legal precedent that agent actions on third-party platforms need explicit per-action consent — shapes feature #9 (Approval Inbox) as legal moat. FCC one-to-one consent rule effective Jan 27 2026 amplifies same requirement.' },
{ title: 'Build order (sequenced by moat strength + dependency)', status: 'upcoming', effort: 'sequencing', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Recommended sequence: (1) #9 Approval-First Inbox FIRST — foundation that every other feature uses, also the legal moat post Amazon-v-Perplexity ruling. (2) #5 Smart Onboarding Bot + #7 Deep Research Agent + #8-alt Customer Memory Web Lookup — the 3 unique wins, any order. (3) #2 Cross-Channel Auto-Poster + #4 Competitor Intelligence + #6 Booking Sync Watcher — the 3 hybrid distinctives. (4) #1 Multi-Platform Review Responder — catch-up bundle for pricing pressure. (5) #10 Predictive Action Engine — long-game meta-orchestrator, requires the bench depth of all above. SKIP or PARTNER for #3 GBP Sync + #8 Aggregator Listings — Yext/Synup do these via direct APIs at $34–250/mo, browser-agent versions strictly inferior.' },
{ title: 'Product placement — 9 of 10 live in Commerce OS', status: 'upcoming', effort: 'architecture', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Mapping to DCS\'s 4-product architecture: Platform (build-time) — only #5 Onboarding Bot has Platform component (kickoff UI handing build-form data to OS). Commerce OS (run-time) — features 1, 2, 3, 4, 5-runner, 6, 7, 8-alt, 9, 10 all live here. Compute + Storage — no new features short-term, but Storage could anchor #9 approval-inbox audit trails to Filecoin for regulated verticals (clinic/legal/finance) in Phase 3. Strategic conclusion: Commerce OS becomes the centre of the product family; tagline shifts to "The AI ops layer that runs your business across every platform you sell on."' },
{ title: 'Defensible window 12–18 months', status: 'upcoming', effort: 'strategic', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Honest reality check: the 4 unique-moat features (#5, #7, #9, #10) are unique RIGHT NOW but two factors erode that fast: (1) Anthropic Claude for Chrome + OpenAI Atlas + Perplexity Comet are commoditising browser-agent capabilities — infrastructure becomes free in 2026. What stays defensible is vertical data + workflow integration (DCS\'s CRM + bookings + memory schema). (2) Enterprise tools (Salesforce Agentforce, Microsoft Copilot Studio, Aisera) will move down-market as pricing pressure forces SMB tiers. Recommendation: ship features 5, 7, 9, 10 in the next 6 months. Spread effort across all 10 = no moat anywhere.' },
{ title: 'Cost economics — sustainable inside $49 Pro tier', status: 'upcoming', effort: 'modeling', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Cost per automation action ~$0.02–0.05 (Browser Use Cloud or Cloudflare Browser Run + LLM call for reasoning). At $0.04/action × 50 actions/owner/month = $2/owner/month variable cost. Sustainable inside the $49/mo Pro tier with 95%+ gross margin. Engineering: each feature 1–3 weeks for a competent solo dev who already has DCS\'s infra. Full 10 features = 3–4 months dedicated work OR 6–8 months alongside other tracks. Numbers estimated from current public pricing for Browser Use Cloud, Cloudflare Browser Run, existing DCS inference chain — verify before committing.' },
],
},
{
id: 's-auto-1-review-responder', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #1 · Multi-Platform Review Responder',
kind: 'internal', priority: 'medium', launchReady: 'partial',
desc: 'Catch-up parity feature — extend D4 Review Drafter from on-site reviews to Google / Yelp / TripAdvisor via browser agent. Owner approves replies on WhatsApp through Approval Inbox (#9). Standalone competitors (EmbedSocial $34/mo, Birdeye, Vendasta, Shapo) are mature — DCS wins via bundling, not innovation. Reviews tab already exists in WhatsApp sidebar (per A7 hotfix May 15).',
items: [
{ title: 'BASE ALREADY SHIPPED — D4 Review Drafter (Phase 42, Zone 82)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-14', note: 'src/services/review-drafter.ts already shipped May 14 2026 (Phase 42, Zone 82). Polls reviews lacking drafts, picks tone from rating (>=4 gracious_thanks, ==3 neutral, <=2 apology_recovery), composes deterministic opener (hash of review_id), upserts into whatslink_review_drafts table (migration 012). Exports pollPendingReviews / draftReplyForReview / saveDraftReply / listDrafts / runReviewDrafter. THIS IS THE DRAFTING ENGINE. What\'s missing for #1: the browser-leg that posts the approved drafts to Google Business Profile + Yelp + TripAdvisor.' },
{ title: 'BASE ALREADY SHIPPED — Reviews endpoints (Shot 3 BE)', status: 'completed', effort: 'done', owner: 'Cowork', addedDate: '2026-05-13', note: 'POST /api/reviews (public, customer submits star + text) and owner GET /api/reviews (dashboard list) shipped May 13 as Shot 3 BE (commit 332ec9a). On-site review flow live. Backend foundation for the review CRUD that #1 extends.' },
{ title: 'BASE ALREADY SHIPPED — Reviews dashboard UI', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-15', note: 'reviews-dashboard.html exists as page module (per A7 dashboard hotfix May 15 listing 23 modules including reviews-dashboard). Reviews tab in WhatsApp sidebar with sub-items already wired. UI shell exists; #1 adds the multi-platform source rows + browser-posting actions.' },
{ title: 'NEW — review_sources table (Google/Yelp/TripAdvisor creds)', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New table review_sources(business_id, platform[google_gbp/yelp/tripadvisor/fb], encrypted_session_profile, last_synced_at, status). Stores the encrypted browser-session profile per platform so the agent can resume authenticated sessions without re-logging-in. OAuth where available (Google Business Profile has API), browser-session-profile fallback for the rest. Session profiles encrypted at rest with per-tenant key.' },
{ title: 'NEW — review-poll-and-draft worker', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Hourly Railway worker: for each business, pull new reviews from connected sources (Google API + browser sessions for Yelp/TripAdvisor), normalize into whatslink_reviews schema, fire existing D4 review-drafter. Draft enters whatslink_review_drafts as today, plus a new browser_post_pending=true flag for #1-scoped drafts. Owner approves via Approval Inbox (#9); approval triggers browser-posting worker.' },
{ title: 'NEW — D4 prompt extension for per-platform tone', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Small extension to review-drafter.ts: accept platform param. Google = formal, professional tone. Yelp = casual, conversational. TripAdvisor = travel-customer tone with destination references. Tone variants stored in INDUSTRY_VOCABULARY (already exists per Wave V). Single config change, no new architecture.' },
{ title: 'BUNDLE PRICING — wedge vs EmbedSocial $34–60/mo', status: 'upcoming', effort: 'pricing', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Standalone review-management tools cost $34–60/mo (EmbedSocial, Birdeye SMB tier, Shapo) for what is essentially D4 + browser posting. DCS bundles into existing $49/mo Pro tier — customer saves a subscription. Don\'t market #1 as innovation; market as "you don\'t need a separate review tool." Catch-up parity, not unique. Verdict from May 15 audit: BUILD as bundle play, not as moat.' },
],
},
{
id: 's-auto-2-channel-poster', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #2 · Cross-Channel Auto-Poster',
kind: 'internal', priority: 'medium', launchReady: 'partial',
desc: 'Hybrid distinctive feature — extend D10 Auto-Broadcast Copywriter from WhatsApp/IG/TG/SMS (API-driven) to TikTok / YouTube Shorts / LinkedIn / Pinterest (browser-driven). One upload, six channels, six different captions. Schedulers like Buffer/Hootsuite/SocialPilot only queue posts; DCS also generates the per-channel variant copy via D10 — that\'s the rare bundle.',
items: [
{ title: 'BASE ALREADY SHIPPED — Bulk broadcast (Wave 1B)', status: 'completed', effort: 'done', owner: 'Cowork', addedDate: '2026-05-13', note: 'Owner selects customer segment, picks template, schedules broadcast. Rate-limited per WA Cloud API tier (250/day → 1k/day → 10k/day → 100k/day). Wave 1B May 13 2026. Foundation for marketing campaigns. THIS IS THE BROADCAST ENGINE for WhatsApp; #2 extends to non-API channels.' },
{ title: 'BASE ALREADY SHIPPED — WhatsApp Status auto-poster (Wave 2)', status: 'completed', effort: 'done', owner: 'Cowork', addedDate: '2026-05-13', note: 'Daily auto-post to owner\'s WhatsApp Status (visible to all customers). Promotes new menu items, offers, hours changes. Configurable schedule and content templates. Wave 2 May 13 2026. THIS IS THE AUTO-POSTER PATTERN — #2 generalizes to other channels.' },
{ title: 'BASE ALREADY SHIPPED — Outbound marketing module (P2 spec)', status: 'pending', effort: '1 month', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Phase 2 spec already includes "social media scheduler (post to Instagram/Facebook/Twitter from one dashboard)" — this IS #2 as originally scoped. #2 extends to TikTok / YouTube Shorts / LinkedIn / Pinterest via browser agent for channels without public posting APIs.' },
{ title: 'NEW — channel_credentials table (encrypted browser profiles)', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New table channel_credentials(business_id, channel[tiktok/youtube/linkedin/pinterest], encrypted_session_profile, last_used_at, status). Same encryption pattern as review_sources. Browser-agent fleet (Cloudflare Browser Run or Browser Use Cloud) reuses the session per channel.' },
{ title: 'NEW — browser-posting-fleet worker', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Per-channel browser-session pool. On owner approval (via Approval Inbox #9), worker picks the right session, opens platform compose page, types caption, uploads media, submits. Failure modes: rate-limit (back off + retry), 2FA prompt (escalate to Human-in-the-Loop), captcha (escalate). All actions logged to journey_runs for audit.' },
{ title: 'NEW — D10 prompt extends channel list', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'D10 Auto-Broadcast Copywriter already handles WA/IG/TG/SMS per-channel tone. Extend channel list to include tiktok/youtube/linkedin/pinterest with their own tone variants (TikTok = hooks + trends, LinkedIn = thought-leadership, etc). Same prompt structure, new channels in the variants array.' },
],
},
{
id: 's-auto-3-gbp-sync', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #3 · GBP Auto-Sync (PARTNER not build)',
kind: 'internal', priority: 'low', launchReady: 'no',
desc: 'Catch-up parity feature — Google Business Profile + Bing Places sync from DCS site changes. ORIGINAL PLAN: browser agent posts diffs. REVISED PLAN after May 15 audit: PARTNER with Yext or use Google\'s own Business Profile API. Browser-agent version would be strictly inferior to Yext\'s 200+ direct publisher integrations at $199+/mo (or Synup at $34.99/mo). Don\'t fight that battle.',
items: [
{ title: 'AUDIT VERDICT — partner or skip, do not build browser version', status: 'upcoming', effort: 'decision', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'May 15 audit found Yext / Birdeye / Synup / Uberall / BrightLocal all do this with direct API integrations to 150–200+ publishers. Browser-agent version would be slow and unreliable next to direct API. Recommended path: (a) integrate with Yext or Synup as backend partner and pass cost through to customers, OR (b) just handle Google Business Profile via Google\'s own free Business Profile API and skip the long-tail directories. Yext-style listing management is its own business; trying to clone it is a side-quest.' },
{ title: 'GBP API minimal version (4 weeks)', status: 'upcoming', effort: '4 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'If we ship anything for #3, do the bare minimum: Google Business Profile API only. On Platform site republish, OS reads the new content via webhook, diffs against current GBP listing, fires GBP API update for hours / services / posts / photos. Free API, no browser-agent infra, no $199/mo Yext bill. Ships in 4 weeks. Long-tail directories (Yelp / TripAdvisor / 200 others) defer to Phase 3 partnership.' },
],
},
{
id: 's-auto-4-competitor-intel', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #4 · Competitor Intelligence Agent',
kind: 'internal', priority: 'medium', launchReady: 'no',
desc: 'Hybrid distinctive feature — extend D6 Daily Briefing with weekly competitor scraping. Owner names 3–5 local competitors at onboarding; browser agent runs weekly, scrapes competitor site + Google reviews + IG profile, output delivered as part of D6\'s already-daily-routine email. Standalone competitor-intel tools (SimilarWeb / SpyFu / Birdeye enterprise) cost $300+/mo and require owners to log in — DCS wins because owners already open D6.',
items: [
{ title: 'BASE — D6 Daily Briefing (in roadmap)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'D6 Daily Briefing is in the existing 17-D-feature roadmap but not yet shipped (May 15 2026). When D6 ships, the briefing email/WhatsApp template is the perfect distribution channel for competitor intel — slots in alongside booking stats and customer issues without adding a new owner habit.' },
{ title: 'NEW — competitor_targets table (per business, max 5)', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New table competitor_targets(business_id, competitor_url, competitor_gbp_id, competitor_ig_handle, added_at, active). Owner enters 3–5 at onboarding via Settings → Track Competitors. Hard cap of 5 to keep compute cost bounded. Browser-agent cost per business per week ~$0.10.' },
{ title: 'NEW — competitor-weekly-scan worker', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Weekly Railway worker. Browser agent reads each competitor\'s site (homepage + services + pricing if visible), public Google reviews (last 7 days, sentiment classification via heuristic), public IG bio + recent posts. Outputs structured changes since last scan into competitor_scans table. D6 prompt template includes "This week\'s competitor moves: ..." section pulling from latest scan.' },
{ title: 'CATCH-UP CHECK — Birdeye competitor benchmarks at $300+/mo enterprise', status: 'upcoming', effort: 'positioning', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Standalone competitor-intel tools exist but at enterprise pricing ($300+/mo Birdeye, SimilarWeb $1k+/mo, SpyFu $99+/mo). No SMB-priced product delivers competitor intel through a daily owner habit. DCS wins on distribution (D6 already opened daily) and bundling (no separate subscription). Position as bundled feature, not headline differentiator.' },
],
},
{
id: 's-auto-5-onboarding-bot', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #5 · Smart Onboarding Bot ★ UNIQUE MOAT',
kind: 'product', priority: 'high', launchReady: 'no',
desc: 'GENUINELY UNIQUE feature. Auto-fills Toast / Square / DoorDash / Zomato / UberEats / Talabat signups using DCS\'s build-form data. Owner enters business info once into DCS build form (Step 4 content editor already collects this), bot fires sequential browser sessions on each chosen platform, fills the common 80%, prompts owner for the unique 20%. End-to-end onboarding for 5 platforms in 30 minutes instead of 3 hours. First-week wow experience. Existing POS integration (Square + Toast) already shipped as Shot 33 May 12 — this builds on that auth scaffold.',
items: [
{ title: 'BASE ALREADY SHIPPED — POS integration scaffold (Shot 33)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-12', note: 'Square Connect OAuth + Toast token-based auth shipped May 12 as Shot 33 (Launch-priority). /pos-integration dashboard live with sync errors surfaced inline. Dummy mode default for demo/beta; real mode is one-env-var flip when first paying restaurant onboards. THIS IS THE AUTH FOUNDATION for #5 — extends from "connect existing account" to "create account on signup."' },
{ title: 'BASE ALREADY SHIPPED — Build form Step 4 content editor', status: 'completed', effort: 'done', owner: 'Deepak Dudi', addedDate: '2026-04-30', note: 'Build form Step 4 already collects business name, address, hours, contact, services/menu, photos during the DCS site build. THIS IS THE DATA SOURCE for #5. No new data entry needed — reuse what\'s already collected. Critical for the "one form, five signups" wedge.' },
{ title: 'BASE ALREADY SHIPPED — Square + Toast scaffolds (Wave V May 14)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-14', note: 'Wave V May 14 shipped src/services/integrations/booksy-stub.ts + practicefusion-stub.ts with HIPAA-aware dummy data, plus industry-vocabulary.ts for restaurant/beauty/healthcare. Substrate pattern for adding more platform stubs. #5 follows this pattern for DoorDash / Zomato / UberEats / Talabat / etc.' },
{ title: 'NEW — onboarding_targets table (per-business platform list)', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New table onboarding_targets(business_id, platform[toast/square/doordash/zomato/ubereats/talabat/etc], status[pending/in_progress/completed/needs_human], browser_session_id, screenshots jsonb, errors text). Owner picks which platforms after build form → table populates → bot picks up from queue.' },
{ title: 'NEW — onboarding-bot-fleet worker', status: 'upcoming', effort: '3 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Browser-session worker pool per platform. On Platform publish + owner platform-picks, worker queues a session per platform. Reads build-form data (already in agent_sessions), fills the common 80% (business name, address, hours, contact, generic photos), screenshots progress at each step. When unique 20% needed (ID upload, bank account, payout details), pauses session and prompts owner on WhatsApp via Approval Inbox (#9): "DoorDash needs your tax ID — paste here." Owner replies → bot continues.' },
{ title: 'NEW — Platform → OS handoff (build form data as JSON)', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Single cross-product wiring. On Platform build publish, fire webhook to OS with build-form data as JSON. OS stores in agent_sessions.build_form (already exists per s-content-editor roadmap). #5 onboarding-bot-fleet reads from there. This is the ONLY cross-product wiring in the 10 features — everything else is OS-internal.' },
{ title: 'MOAT POSITIONING — first-week wow + ~$5–15k consulting equivalent', status: 'upcoming', effort: 'positioning', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'No SMB tool does this. Enterprise consulting firms do it manually for $5–15k per restaurant onboarding. Yext does listing distribution but you still set up Toast/Square/DoorDash manually. Vertical SaaS (Toast, Square) onboard you to THEIR system only — won\'t auto-fill competitors. The moat: "you signed up at 10 AM, you\'re live on 7 platforms by 10:30 AM" is the marketing demo that sells the whole DCS product. Pull forward in roadmap — worth de-prioritising other tracks for.' },
],
},
{
id: 's-auto-6-booking-sync', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #6 · Booking Sync Watcher',
kind: 'internal', priority: 'medium', launchReady: 'partial',
desc: 'Hybrid distinctive feature — browser agent pulls from Fresha / Booksy / OpenTable / Mindbody every 15 min, writes new appointments into whatslink_bookings. Lets D14 No-Show Recovery + D16 Reschedule Wizard operate on FULL data including legacy bookings. Lets owners try DCS without abandoning Fresha — earns the migration over months. Substrate already shipped Wave V May 14.',
items: [
{ title: 'BASE ALREADY SHIPPED — Booksy + Practice Fusion stubs (Wave V)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-14', note: 'Wave V May 14 2026 shipped src/services/integrations/booksy-stub.ts + practicefusion-stub.ts with HIPAA-aware dummy data. Pure substrate — real OAuth not wired yet. #6 lights up the real OAuth + the polling worker on top of the stub pattern.' },
{ title: 'BASE IN ROADMAP — OpenTable / SevenRooms / Booksy / Square Appointments integration', status: 'upcoming', effort: '1 month each', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Already in roadmap under s-restaurant-os (OpenTable/SevenRooms, $50+ restaurants gate) and s-salon-os (Booksy/Square Appointments). #6 generalizes these into a single browser-driven sync watcher + extends to legacy systems without public APIs (smaller regional booking tools).' },
{ title: 'BASE ALREADY SHIPPED — whatslink_bookings table + Google Calendar sync (Shot 27)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-12', note: 'whatslink_bookings is the canonical booking table — already wired to D14 reminders, D16 reschedule, journey triggers (BOOKING_CONFIRMED / BOOKING_NO_SHOW / etc per Wave J). Google Calendar OAuth + ICS fallback shipped Shot 27 May 12. #6 writes external bookings INTO this table with source=fresha/booksy/etc — D14+D16 work without modification.' },
{ title: 'NEW — external_booking_sources table (per business, credentials)', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New table external_booking_sources(business_id, source[fresha/booksy/opentable/sevenrooms/mindbody/regional], encrypted_session, last_synced_at, status). Per-tenant encryption for session profiles. Polling cadence per source (15min default, configurable).' },
{ title: 'NEW — booking-sync-poll worker (15-min cadence)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Railway worker. Every 15 min: for each active external_booking_source, headless browser pulls appointment list from the legacy system, normalizes to whatslink_bookings schema (using INDUSTRY_VOCABULARY mappings from Wave V), upserts with source field. Idempotent on external_id+source unique index. New rows trigger journey events via existing spawnExtendedTriggers (Wave J-2). NO changes needed to D14/D16/journey engine — they already operate on whatslink_bookings.' },
{ title: 'WEDGE — "try DCS without abandoning Fresha"', status: 'upcoming', effort: 'positioning', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Most owners refuse to migrate booking systems day 1. #6 lets DCS sit alongside, eat the data, earn the migration over months. Standalone "booking sync" tools (Cronofy, Reclaim, Zapier connectors) have no purpose — DCS\'s sync gives D14 + D16 + journey engine real customer data to operate on. Low-risk to ship: it\'s read-only initially. Migration earned when owner sees DCS\'s automations work better than the legacy system.' },
],
},
{
id: 's-auto-7-deep-research', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #7 · Deep Research Agent ★ UNIQUE MOAT',
kind: 'internal', priority: 'high', launchReady: 'no',
desc: 'GENUINELY UNIQUE feature. When customer asks WhatsApp inbox a question whose answer is NOT in owner\'s content + memory + FAQ, browser agent runs deep research (Claude Computer Use or Browser Use), drafts a cited reply, asks owner "OK to send this?" via Approval Inbox (#9). Owner taps yes → message sends. The unknown becomes known automatically. Perplexity + ChatGPT Atlas can do deep research IN THEIR OWN UI; no tool does the research-to-message-delivery loop tied to a specific business\'s inbox. DCS owns both ends.',
items: [
{ title: 'BASE ALREADY SHIPPED — Phase 42 D3 VoC Digest (Zone 81)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-14', note: 'src/services/voc-digest.ts shipped May 14. Reads recent customer signal from memory_events + whatslink_reviews + whatslink_bookings, classifies into 8 issue categories. Pure heuristic — zero LLM cost on default path. PATTERN PRECEDENT for #7\'s question classifier (decides when memory/D15 isn\'t enough → triggers research).' },
{ title: 'BASE IN ROADMAP — D15 Auto-FAQ', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'D15 Auto-FAQ is the first-line answer-generator (in 17-D-feature roadmap). #7 is the backstop when D15 fails. Architectural relationship: customer message → check memory → check D15 FAQ → if both miss, fire #7 research. Three-tier fallback.' },
{ title: 'BASE ALREADY SHIPPED — Unified Inbox (Day 6 sprint)', status: 'completed', effort: 'done', owner: 'Cowork', addedDate: '2026-05-13', note: 'Multi-channel unified inbox shipped Day 6 of WA expansion sprint May 13 (commits c446013 + 7ac08a9). #7 sits inside the unified inbox — every channel\'s unknown question gets the same backstop.' },
{ title: 'NEW — question-classifier worker', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Fires on every inbound message. Heuristic: keyword overlap with stored memories + D15 FAQ entries + content from build-form. If overlap below threshold (~30%), flags as "unknown" and queues for #7 research. Heuristic-first like D3 VoC Digest — no LLM cost on the default path. Threshold tunable per business.' },
{ title: 'NEW — research-agent worker (Claude Computer Use or Browser Use)', status: 'upcoming', effort: '3 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Browser-agent fleet for research tasks. Source priority: (1) the business\'s own website (canonical), (2) public Google reviews, (3) nearby Google Maps for venue questions, (4) public news. Drafts a candidate reply with citation block ("found this on your website at /menu, and from your last 5 Google reviews ..."). Cost per research ~$0.05–0.10. Output goes through Approval Inbox (#9).' },
{ title: 'NEW — research_runs audit table', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Audit table research_runs(business_id, message_id, question_text, sources_consulted jsonb, draft_reply text, owner_decision[approved/edited/rejected], approved_at, sent_at). Every research action logged. Enables (a) feedback loop to improve question-classifier accuracy, (b) compliance audit for high-stakes responses, (c) "approve-all from this business for 24h" patterns in #9.' },
{ title: 'MOAT POSITIONING — best demo in the product', status: 'upcoming', effort: 'positioning', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'This is the feature where one demo sells the whole product. "Customer asked something the salon owner didn\'t know. DCS researched the answer from their own website + reviews. Drafted a reply. Asked permission. Sent." Goes from "managing the inbox" to "approving the inbox." No SMB tool does this loop. Defensible for 12–18 months before enterprise tools move down-market.' },
],
},
{
id: 's-auto-8alt-memory-web-lookup', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #8-alt · Customer Memory Web Lookup ★ UNIQUE',
kind: 'internal', priority: 'medium', launchReady: 'no',
desc: 'GENUINELY UNIQUE feature (replaces original #8 Auto-Listing on Aggregators, which is parity-with-Yext and not worth building). When a returning customer messages, browser agent quickly checks public sources (their Instagram bio, LinkedIn role, prior reviews they\'ve left elsewhere) and surfaces it to the owner inside the chat — owner walks into the conversation knowing context. PRIVACY-FENCED: only public info, only on inbound message, refreshed not stockpiled. Pairs with D9 Customer Segment Tagger (already shipped Wave 1B).',
items: [
{ title: 'BASE ALREADY SHIPPED — D9 Customer Segment auto-tags (Wave 1B)', status: 'completed', effort: 'done', owner: 'Cowork', addedDate: '2026-05-13', note: 'D9 Customer Segment Tagger shipped Wave 1B May 13. Auto-tags customers based on behavior: VIP (5+ bookings), Recent (last 30 days), Lapsed (no activity 90+ days), Ad-converted (came via click-to-WA ad). Powers targeted broadcasts. #8-alt enriches D9 with public-context tags (industry / role / interests inferred from public LinkedIn or IG bio).' },
{ title: 'BASE ALREADY SHIPPED — C1 inbox tables (memory_events + contact_identities + memory_summaries)', status: 'completed', effort: 'done', owner: 'Cowork', addedDate: '2026-05-10', note: 'memory_events + contact_identities + memory_summaries tables shipped May 10 — the memory schema that powers customer context. #8-alt writes to a new contact_public_context table that JOINs to contact_identities by phone. Same schema family.' },
{ title: 'NEW — contact_public_context table (TTL-based, refreshed not stockpiled)', status: 'upcoming', effort: '1 day', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New table contact_public_context(contact_id, source[ig/linkedin/google_reviews], data jsonb, fetched_at, expires_at). TTL 30 days — refresh on each new inbound message if expired. NEVER stockpile. Privacy-first: only stores PUBLIC info (bio text, public role, public reviews), never DMs, never private data. GDPR-compliant: contact_public_context purged when contact_identities purged.' },
{ title: 'NEW — contact-context-lookup worker (fires on first inbound)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Fires on first inbound message from a contact (or when contact_public_context expired). Pulls phone → matches to public IG/LinkedIn via lightweight search (Twilio Lookup-style public-data API where available, browser-agent for the rest). Strict privacy fence: only stores what\'s publicly visible without login. Cost per lookup ~$0.02. Cached 30 days.' },
{ title: 'NEW — D9 prompt extension to include context tags', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Small extension to D9 segmentation prompt: include public-context as additional tag inputs ("works in tech", "vegan based on IG", "leaves detailed reviews"). Helps owners walk into conversations with context. Tags shown in unified-inbox contact card panel.' },
{ title: 'MOAT POSITIONING — Crystal Knows for hospitality', status: 'upcoming', effort: 'positioning', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'No SMB tool does this for hospitality/services. Closest competitor: Crystal Knows (B2B sales personality profiling, enterprise pricing). #8-alt is the consumer-services equivalent at SMB pricing, fenced for privacy. Marketing positioning: "Walk into every conversation knowing who you\'re talking to." Pairs naturally with D9.' },
{ title: 'PRIVACY GUARANTEE — public sources only, TTL refresh, never private DMs', status: 'upcoming', effort: 'policy', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Hard policy constraints baked into worker: (1) NEVER reads anything requiring login (no scraping private DMs, no scraping behind paywall). (2) Only public bio text + public reviews + public IG posts visible to anyone. (3) TTL 30 days then refreshed, never long-term stockpiled. (4) Disclosed in DCS privacy policy. (5) Customer can opt-out: their contact_id flagged no_public_lookup=true. Conservative legal stance post Amazon-v-Perplexity ruling.' },
],
},
{
id: 's-auto-9-approval-inbox', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #9 · Approval-First Inbox ★ FOUNDATION + LEGAL MOAT',
kind: 'internal', priority: 'critical', launchReady: 'no',
desc: 'GENUINELY UNIQUE FOR SMB. Foundation layer that every other automation depends on. New dashboard tab showing every queued agent action as a card with: (a) what the agent wants to do, (b) visible diff/preview, (c) which platform, (d) which customer affected, (e) Approve / Edit / Reject / Approve-all-from-this-agent-for-24h. High-stakes actions ping owner on WhatsApp; low-stakes batch into daily digest. Build this FIRST — every other feature uses it. Also the LEGAL MOAT post Amazon-v-Perplexity (March 2026) + FCC one-to-one consent rule (Jan 27 2026).',
items: [
{ title: 'BASE IN ROADMAP — Phase 1 architecture lock (append-only audit log)', status: 'completed', effort: 'design', owner: 'Deepak Dudi', addedDate: '2026-05-09', note: 'Phase 1 enterprise-readiness architecture lock #3 already identified: "Append-only audit log — every action logged to immutable table, never UPDATE/DELETE, for compliance." This IS the substrate #9 builds on. Audit log spec exists; #9 makes it user-facing.' },
{ title: 'BASE ALREADY SHIPPED — Journey engine state persistence (Shot 29 + Wave J)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-12', note: 'Journey engine MOAT shipped May 12 + Wave J expansion May 14 (22/22 triggers active). Engine persists multi-step state between actions — that\'s the same pattern #9 uses for queued actions. journey_runs table is the precedent for agent_actions_queue.' },
{ title: 'NEW — agent_actions_queue table (every pending action across all features)', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New table agent_actions_queue(action_id, business_id, feature[#1/#2/#4/#5/#6/#7/#8alt/#10], action_type, payload jsonb, preview text, platform, customer_id, stakes[low/medium/high], status[pending/approved/edited/rejected/executed], queued_at, approved_at, executed_at, error text). Every feature 1–8alt + 10 writes here. Single funnel for all owner approvals.' },
{ title: 'NEW — agent_approval_policies table (per owner, per feature)', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New table agent_approval_policies(business_id, feature, auto_approve_threshold[never/low_only/all], blanket_approval_until timestamp, last_changed_at). Owner can set: "auto-approve all review replies under 50 words for 24h" or "require approval on every browser-post" or "never auto-approve onboarding-bot final-step." Per-feature granularity.' },
{ title: 'NEW — Approvals dashboard tab (top-level OS UI)', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'New top-level tab in OS dashboard sidebar (sits with Inbox / Bookings / Templates / etc per the 25-tab list from A7 hotfix). Renders cards from agent_actions_queue. Each card: action description, diff preview, platform badge, customer name, [Approve] [Edit] [Reject] [Approve all from #N agent for 24h] buttons. Sortable by feature, stakes, age.' },
{ title: 'NEW — WhatsApp delivery for high-stakes actions', status: 'upcoming', effort: '3 days', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'High-stakes actions (browser-posting to platforms, sending to customers, onboarding-final-steps) ping owner directly on WhatsApp with action preview + Approve/Reject buttons. Low-stakes (review reply drafts <50 words, internal context lookups) batch into a daily digest at 8 AM local time. Stakes classifier is per-feature config.' },
{ title: 'NEW — audit_log table (immutable, every approval decision)', status: 'upcoming', effort: '2 days', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Append-only audit_log(action_id, business_id, owner_phone, decision, decision_at, executed_at, original_payload jsonb, final_payload jsonb). Never UPDATE, never DELETE. Compliance grade. Phase 1 architecture lock #3 (already designed) cashed in here. Storage layer could Filecoin-anchor these for regulated verticals (clinic/legal/finance) in Phase 3.' },
{ title: 'LEGAL MOAT — Amazon-v-Perplexity ruling + FCC Jan 27 consent rule', status: 'upcoming', effort: 'policy', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Amazon vs Perplexity injunction (March 2026) blocks agentic browsing on third-party platforms without per-action consent. FCC one-to-one consent rule (effective Jan 27 2026) eliminates blanket consent for customer messaging. Both legal precedents require: per-action owner consent, audit trail, customer opt-out. #9 IS the mechanism that makes features #1, #2, #4, #5, #6, #7, #8-alt, #10 legally shippable in 2026\'s regulatory climate. Build this FIRST before any other automation fires.' },
{ title: 'MARKETING POSITIONING — "DCS shows you every decision before it\'s made"', status: 'upcoming', effort: 'positioning', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Frame as: "DCS shows you every decision before it\'s made — built for the new consent era." Position as trust feature, not bureaucracy. While Salesforce Agentforce / Microsoft Copilot rush autonomy and lose customer trust, DCS ships approval-first and earns it. Trust scales with owner comfort: start with all-approve-required, owner toggles auto-approve as confidence grows.' },
],
},
{
id: 's-auto-10-predictive-action', project: '04 · STRATEGIC / FUTURE PRODUCTS', name: 'Hybrid #10 · Predictive Action Engine ★ LONG-GAME MOAT',
kind: 'internal', priority: 'high', launchReady: 'no',
desc: 'GENUINELY UNIQUE for SMB. Meta-orchestrator over the 17 D-features + features 1–8-alt. Reads all D-feature outputs each morning, generates top-3 owner actions for the day, drafts each action using features 1–8-alt, presents via Approval Inbox (#9). Goes from "operator running tools" to "owner approving actions." Multi-agent orchestration exists at enterprise scale (Salesforce Agentforce, Microsoft Copilot, Aisera) — no SMB-tier product has the BENCH DEPTH of agents to orchestrate. DCS has 17 D-features + 8 hybrid automations = 25 specialist agents. The orchestrator is the long-game category lead. Note: distinct from existing s-predict (which is build-quality prediction, different feature).',
items: [
{ title: 'BASE ALREADY SHIPPED — D1 Owner Dashboard via WhatsApp (Zone 79)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-14', note: 'src/services/owner-command-router.ts shipped May 14 Phase 42 Zone 79. 7 deterministic commands (today/confirm N/no-show N/revenue/customer/reminder/help). 1h session-scoped cache for positional commands. PATTERN PRECEDENT for #10 — the morning-orchestrator extends this with "today\'s top 3 actions" as a deterministic+LLM hybrid.' },
{ title: 'BASE ALREADY SHIPPED — 17 D-features in roadmap, several already live', status: 'partial', effort: 'tracking', owner: 'chat-Claude', addedDate: '2026-05-14', note: 'D1 Owner Dashboard SHIPPED (Zone 79). D3 VoC Digest SHIPPED (Zone 81). D4 Review Drafter SHIPPED (Zone 82). D2/D5 SHIPPED (Zones 80/83). The other D-features (D6 Daily Briefing, D9 Customer Segment Tagger SHIPPED Wave 1B, D10 Auto-Broadcast Copywriter, D11 Heatmap, D12 Demand Forecaster, D13 LTV Tier, D14 No-Show Recovery, D15 Auto-FAQ, D16 Reschedule Wizard, D17 LTV Forecaster) ship over the next 3 months. #10 reads from ALL of them once they\'re live.' },
{ title: 'BASE ALREADY SHIPPED — Journey engine 22/22 triggers (Wave J → J-2-E)', status: 'completed', effort: 'done', owner: 'chat-Claude', addedDate: '2026-05-14', note: 'Journey engine with 22/22 triggers active by Wave J-2-E. Already a meta-engine: triggers fire → actions chain → state persists. #10 extends this from REACTIVE (trigger-based) to PREDICTIVE (read state + propose actions before triggers fire). Same engine, new direction of dataflow.' },
{ title: 'NEW — morning-orchestrator worker (8 AM local per owner)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Runs at 8 AM local time per owner (timezone from agent_sessions). Reads: all D-feature outputs from prior 24h (D3 VoC digest, D11 heatmap deltas, D12 demand forecast for today, D17 LTV trend), all feature 1–8alt status, booking trends, no-show patterns. Generates: top 3 action proposals using DCS inference chain. Routes: every proposal goes through Approval Inbox (#9). Cost per owner per day ~$0.10 (~$3/owner/month, sustainable in $49 Pro tier).' },
{ title: 'NEW — Today\'s Actions widget (OS dashboard homepage)', status: 'upcoming', effort: '1 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Top widget on OS dashboard homepage. 3 cards each morning, each a proposed action: ("Demand forecast shows Tuesday low — broadcast 20% Tue discount to lapsed VIPs?"), ("Negative review trend — D4 drafted 3 apology-recovery replies for your approval"), ("Onboarding bot finished UberEats step 1 — owner ID upload needed"). Tap Approve → routes through Approval Inbox (#9).' },
{ title: 'NEW — action-feedback-loop (tuning future suggestions)', status: 'upcoming', effort: '2 wk', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Track which proposals get approved vs rejected. Feed back into the morning-orchestrator prompt as "this owner prefers low-touch broadcasts over discount offers." Self-improving per business. Over 3–6 months, suggestions become more relevant. Same RLHF-on-rails pattern as F1 Quality Scoring substrate (Shot 20).' },
{ title: 'DISTINCTION FROM s-predict (F7) — different feature', status: 'upcoming', effort: 'clarification', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Existing s-predict (Predictive Outcome Engine F7) is about PRE-FLIGHT BUILD QUALITY prediction on the Platform side — predicts whether a build will be low-quality before running it, saves compute. #10 Predictive Action Engine is about MORNING OPERATIONAL ACTION recommendations on the OS side — predicts what the owner should do today. Different products, different surfaces, different timing. Both valid; keep both in roadmap.' },
{ title: 'MOAT POSITIONING — category leader for SMB multi-agent orchestration', status: 'upcoming', effort: 'positioning', owner: 'Deepak Dudi', addedDate: '2026-05-15', note: 'Multi-agent orchestration is THE hot topic for 2026 (per PwC, Deloitte). Salesforce Agentforce, Microsoft Copilot Studio, Aisera all have "orchestrator agents" that coordinate sub-agents. All enterprise-priced. SMB-tier products like Apollo, NoimosAI, Robylon do ONE thing autonomously, not a meta-orchestrator across verticals. #10 is genuinely unique at the SMB tier — possible only because DCS has 17 D-features + 8 hybrid automations = the bench depth no competitor has. Defensible 18+ months.' },
],
},
// ============== 05 · INTERNAL TOOLS / OPS ==============
{
id: 't-vendor', project: '05 · INTERNAL TOOLS / OPS', name: 'Vendor Spend Tracker',
kind: 'internal', priority: 'medium', launchReady: 'partial',
desc: 'v3 + v4 endpoints, autofetch worker, daily snapshots, burn_rates view. Owner-only auth. DROPPED FROM LAUNCH-BLOCKING per Deepak Dudi May 12 — week 2+ post-launch.',
items: [
{ title: 'v3 endpoints preserved', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-10', note: 'Vendor tracker version 3 endpoints kept available for backward compatibility while v4 rolls out. Endpoints: GET /api/admin/vendor-spend (sum per vendor per period), GET /api/admin/vendor-list (active vendors), GET /api/admin/vendor-detail/:id (per-vendor breakdown). v3 used by current dashboard; v4 will replace once verified stable.' },
{ title: 'v4 auto-fetch worker + daily snapshots', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Vendor tracker v4 includes a background worker that auto-fetches latest balances from each vendor API (Anthropic Admin API, Replicate, GCP, etc.) once per day. Snapshots stored in vendor_balance_snapshots table for trend analysis. Replaces manual monthly check of each vendor portal. Commit: 6121845.' },
{ title: 'v4 backend deployed on api.dcsai.ai', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Vendor tracker v4 backend deployed at api.dcsai.ai with endpoints for the dashboard. Authenticated via existing DCS auth (only founder + future admin team can access). Returns: per-vendor spend, period comparisons, projected monthly spend based on trend, alert thresholds. Frontend dashboard reads from these endpoints.' },
{ title: '8 AI vendor balances locked ($1,262.10)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Total locked balance across 8 AI vendors as of last check: $1,262.10. Vendors: Anthropic (~$X), Cerebras, Together, Gemini, DeepSeek, Groq, SambaNova, Replicate. Locked balance = remaining credit on each platform. Helps with cash flow planning — knowing how many builds we can run before needing to top up.' },
{ title: 'Cerebras tracker patches in code (commit 568c8ea)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Cerebras-specific usage tracker patches shipped May 12 (commit 568c8ea). callCerebras function now writes every API call to vendor_usage_events table with model, tokens, cost, session_id. First vendor with full per-call tracking. Will extend pattern to other 7 vendors in coming phases (Phase 43.4 next).' },
{ title: 'Verify Cerebras writes after Railway redeploy', status: 'pending', effort: '5 min', owner: 'Deepak Dudi', addedDate: '2026-05-12', testResult: 'partial', note: 'After Railway redeploy of the Cerebras tracker patch, verify that real API calls are correctly writing to vendor_usage_events. Test: run a small build, check Supabase for new rows in vendor_usage_events with correct values, confirm cost calculations match Cerebras invoice. Critical sanity check before extending pattern to other vendors.' },
{ title: 'Apply 7 other callXXX tracker patches (POST-LAUNCH)', status: 'upcoming', effort: '60 min batch', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'POST-LAUNCH work: extend the Cerebras tracker pattern to the other 7 vendors. callAnthropic, callTogether, callGemini, callDeepSeek, callGroq, callSambaNova, callTrdInference. Each takes ~30 min. Total ~4 hours. Result: full per-call cost visibility across entire inference chain. Critical for understanding unit economics at scale.' },
{ title: 'Anthropic Admin / Replicate / GCP fetchers (POST-LAUNCH)', status: 'upcoming', effort: '4 hrs', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'POST-LAUNCH: build daily fetchers for vendor portals that don\'t auto-track. Anthropic Admin API (for total org spend), Replicate (image generation cost), GCP (if any GCP services used for analytics or backup). Each fetcher runs daily, pulls remaining balance + period spend, writes to vendor_balance_snapshots. Removes manual portal check work.' },
],
},
{
id: 't-cost-dash', project: '05 · INTERNAL TOOLS / OPS', name: 'F1 · Cost Dashboards',
kind: 'internal', priority: 'low', launchReady: 'na',
desc: 'Cost-per-build · cost-per-agent · cost-per-user. Owner-only. DROPPED to week 2+ post-launch.',
items: [
{ title: 'Cost-per-build tab', status: 'upcoming', effort: '4 hrs', owner: 'Deepak Dudi', deps: 'tracker verified', addedDate: '2026-05-11', note: 'Dashboard tab showing cost breakdown per individual build attempt: which agents ran, how many tokens each used, which tier each was routed to, total cost in micro-USD, session_id for traceback. Helps debug expensive outliers (a build costing $0.50 when avg is $0.16 is worth investigating). Sortable by date, cost, session.' },
{ title: 'Cost-per-agent tab', status: 'upcoming', effort: '2 hrs', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'Dashboard tab aggregating cost by agent type: architect-agent, designer-agent, frontend-agent, backend-agent, qa-agent, devops-agent, etc. Helps identify which agents are cost drivers and which optimizations (e.g. moving qa-agent off Claude to Cerebras) had biggest impact. Visualized as bar chart with month-over-month comparison.' },
{ title: 'Cost-per-user attribution', status: 'upcoming', effort: '4 hrs', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'Per-user cost attribution: how much vendor cost did each DCS customer generate this month? Critical for unit economics — identifying free-tier users costing >$5/mo (unsustainable), Pro-tier users with healthy margin (>70%), enterprise users that need custom pricing. Drives pricing decisions and tier enforcement.' },
],
},
{
id: 't-knowledge', project: '05 · INTERNAL TOOLS / OPS', name: 'Confidential Knowledge Tab (this!)',
kind: 'internal', priority: 'high', launchReady: 'yes',
desc: 'You are looking at it right now. Single HTML file dropped into trd-frontend/owner/. Same token-secured auth as vendor dashboard.',
items: [
{ title: 'v1 build (29 features, 141 items)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Initial version of the Knowledge dashboard built earlier this week. 29 features, 141 items. Foundation for what eventually grew into the current 54-feature, 500+ item dashboard. v1 lacked: tech/stack tab, blockers+bugs tab, timeline view, FAQ-style item expand. Useful reference point for what changed since.' },
{ title: 'v2 build (DCS Holding parent, 15 strategic products)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', testResult: 'pass', note: 'v2 expanded the Knowledge dashboard scope: added DCS Holding parent entity (legal structure context), 15 strategic future products (Restaurant OS, Salon OS, etc.), and reorganized into 5 project groups (Holding & Entities, Core Products, Internals, Strategic, Tools/Ops). This is the structure that the current v4 builds on.' },
{ title: 'URL field + PRODUCT badge for top-level surfaces', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Added URL field to product items so any product with a live URL (dcsai.ai, compute.dcsai.ai, storage.dcsai.ai, etc.) shows it prominently. Also added kind: \'product\' vs \'entity\' badge to differentiate revenue-generating products from legal/organizational entities. Visual clarity for what\'s a product vs what\'s structure.' },
{ title: 'Search + 5 filters (All / Products / Ready / Blockers / New)', status: 'completed', effort: 'shipped', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Top-bar search across all 500+ items by title, note, phase, status. Plus 5 quick filters: All (everything), Products only (hide entities + internals), Launch ready (yes/partial), Blockers (failing tests + pending bugs), New (added in last 7 days). Helps quickly find what\'s relevant for a given context.' },
{ title: 'PDF download per feature', status: 'upcoming', effort: '1 hr', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Future feature: button on each product page to download a PDF report. Useful for: investor updates (here\'s what\'s shipped for Storage), retrospectives (what did we ship in April), team handoff (when scaling beyond solo founder). Generated server-side via pdfkit from the same data.' },
{ title: 'Wire to backend for live edits', status: 'upcoming', effort: '3 hrs', owner: 'Deepak Dudi', addedDate: '2026-05-12', note: 'Future feature: move FEATURES array from inline JS to a Supabase table (knowledge_features). Each Claude session can read/write via API instead of editing the HTML file. Avoids re-uploading chat batches + rebuilding the file every session. Allows live edits from any device. Decision pending — current static file is simpler to ship for launch.' },
],
},
{
id: 't-ops', project: '05 · INTERNAL TOOLS / OPS', name: 'Ops & Infra Hygiene',
kind: 'internal', priority: 'low', launchReady: 'na',
desc: 'Token rotation, cleanup, renewals. Mostly non-Claude work.',
items: [
{ title: 'Generate Anthropic Admin API key', status: 'upcoming', effort: '15 min', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'Anthropic Admin API provides org-level spend data (vs the per-API-key data from regular API). Required for the vendor tracker to show total Anthropic spend. Key generation: Anthropic Console → Admin → Generate new Admin Key. Store in Railway env vars. Used by daily fetcher worker (post-launch).' },
{ title: 'Rotate Supabase service_role JWT', status: 'upcoming', effort: '30 min', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'Supabase service_role JWT is the master key for backend operations. Should be rotated periodically (every 3-6 months) to limit blast radius if leaked. Process: generate new JWT in Supabase dashboard → update Railway env vars (DCS-CN backend + storage backend + main DCS backend) → verify nothing broken → invalidate old JWT. ~30 min total.' },
{ title: 'Rotate Replicate token', status: 'upcoming', effort: '15 min', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'Replicate API token used for SDXL image generation in Image Bank. Should be rotated periodically. Process: generate new token in Replicate dashboard → update Railway env vars → verify image generation still works → invalidate old token. ~15 min. Lower urgency than Supabase rotation since Replicate access is more contained.' },
{ title: 'Cleanup src/index.ts.bak* files', status: 'upcoming', effort: '5 min', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'Multiple .bak backup files left in src/index.ts from various Cowork edits over the past weeks. Files like src/index.ts.bak.1778316620599 etc. take up space in repo and clutter the file tree. Cleanup: delete all .bak files older than 30 days. Could automate via cron job or pre-commit hook to prevent buildup.' },
{ title: 'Vendor data Deepak Dudi to provide (renewal dates, last-4 CCs, etc.)', status: 'pending', effort: 'Deepak Dudi to compile', owner: 'Deepak Dudi', addedDate: '2026-05-11', note: 'Founder to provide: renewal dates for all vendor subscriptions (Anthropic, Cerebras, Together, etc.), last-4-digits of credit cards used (HDFC 0206, Indusind 1258), domain renewal calendar entries. Centralized record helps with: forgetting renewals, replacing cards when they expire, business continuity if founder unavailable. Document stored in encrypted vault.' },
],
},
// ============== 06 · COWORK SESSION SHIP LOG — MAY 2026 ==============
{
id: 'cowork-ship-log-may-2026', project: '06 · COWORK SHIP LOG (May 2026)',
name: 'DK × chat-Claude · Cowork Ship Log (May 2026)',
kind: 'internal', priority: 'high', launchReady: 'yes',
desc: 'Chronological log of all work shipped through the Cowork (DK paste error log → chat-Claude writes patcher .cjs + merge .sh → DK runs → green) workflow during the May 2026 launch sprint. Three threads: (a) Wave 2 + Tier substrate (Shots 22-29, 33), (b) Late-night batch (Shots 30, 32, 34, 36, 37, 42), (c) Final shots + RCA architectural fixes + critical hot-fixes. Output of about ~30 patcher/merge script pairs, several RCA docs, and an alphabet-soup of sentinel zones (Zone 49-68). Each entry below has a one-paragraph explanation of what it does and why it matters. Filter by status to see what is done vs still pending verification.',
items: [
// ── WAVE 2 + TIER SUBSTRATE (Shots 22-25) ──
{ title: 'Shot 22 — Permanence footer link injector (Zone 49)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Wave 2 substrate', note: 'Post-emit injector adds a small "Permanently archived on Filecoin" link to every customer page footer, deep-linking to the DCS storage certificate for that build. Wires up the Phase 4 Living Site / Permanence promise so customers can see the proof, not just take it on faith. Sentinel zone 49 protects the injector function so future agents do not accidentally strip it. Backward-compatible — pages without a DCS storage CID skip the link instead of erroring.' },
{ title: 'Shot 23 — Replicate SDXL provider swap', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Wave 2 substrate', note: 'Switched the hero/feature-image generator from a placeholder pipeline to Replicate SDXL (Stable Diffusion XL) via the official Replicate Node SDK. Adds a robust retry loop with exponential backoff and webhook polling. Cost-aware: rejects requests over a per-build image budget so a runaway agent cannot burn 50 USD on hero variants. Falls back gracefully to a curated stock pool if Replicate is rate-limited.' },
{ title: 'Shot 24 — Real Lighthouse + Copy judge', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Wave 2 substrate', note: 'Build-quality scoring pipeline that ACTUALLY runs Lighthouse on the deployed site (Performance, Accessibility, Best Practices, SEO) plus a Claude-Haiku-backed Copy judge that grades hero/feature copy for clarity, tone, and tier-appropriateness. Scores write to the build_quality_scores table and surface on the build review card in /app22. Replaces the previous stub that returned a fixed score of 87 regardless of build quality.' },
{ title: 'Shot 25 — F1 cost dashboard (/admin/costs.html)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Wave 2 substrate', note: 'Admin-only page that pulls per-vendor spend from the daily vendor snapshot table and renders a top-line "today vs trailing-30-day average" with sparkline. Shows Cerebras, Together, Replicate, Stripe-fee, Railway, Vercel, Anthropic. Used by founder to spot a runaway agent within the same day instead of finding out at month-end. Auth-gated to admin user IDs (hardcoded allowlist for now, RLS-policy upgrade scheduled F1.5).' },
// ── LAUNCH-PRIORITY QUEUE (Shots 26-29, 33) ──
{ title: 'Shot 26 — Saved views (filters)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Launch-priority', note: 'Customers can save and re-apply named filter sets on the Bookings + Reviews dashboards (e.g. "This week, confirmed, party >= 4"). Each user can have up to 10 saved views per dashboard. Stored in a saved_views table with (user_id, dashboard, name, filter_json). UI: dropdown beside the filter bar, plus a Save Current View link. Foundation for future "team views" feature (shared between teammates) without schema migration.' },
{ title: 'Shot 27 — Google Calendar OAuth + ICS sync', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Launch-priority', note: 'Two-way integration with Google Calendar: OAuth flow to grant access, then per-booking calendar event creation + updates + cancellations. Falls back to ICS file download for users who decline OAuth. Stores OAuth refresh token encrypted in calendar_integrations table. Refresh token is rotated automatically when expired. Owner sees their bookings on their phone calendar without copying them by hand — fixes a top customer complaint from the beta program.' },
{ title: 'Shot 28 — Push notifications via web-push (VAPID)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Launch-priority', note: 'Native browser push notifications for owners: new-booking, booking-cancelled, review-arrived, urgent-message. Uses the web-push npm package with VAPID keys (rotated quarterly). Permission flow asks once on first dashboard load and stores the resulting endpoint per device, so an owner can be subscribed across phone + laptop + tablet. Required @types/web-push for TypeScript builds — that dep got added by the patcher.' },
{ title: 'Shot 29 — Journey automation MOAT (the multi-step engine)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Launch-priority', note: 'Core differentiator that no SMB-marketing-tool competitor has: a multi-step customer journey engine. Owners define triggers (booking confirmed, deposit paid, review left) and a sequence of actions (WhatsApp message, email, wait 24h, branch on tag, send another message). Engine runs as a Railway worker, executes one step at a time, persists state between steps. Foundational MOAT — adds to DCS what Customer.io / Iterable / Klaviyo charge $500/mo for, and ours is WhatsApp-native instead of email-native.' },
{ title: 'Shot 33 — POS integration (Square + Toast scaffold)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Launch-priority', note: 'POS-side scaffold for restaurant verticals: Square Connect OAuth + Toast token-based auth. Lets the dashboard see live orders flowing through the POS in /pos-integration, with sync errors surfaced inline. Dummy mode is the default (no API key needed) for the demo and beta. Real Square/Toast live mode is a one-env-var flip when the first paying restaurant onboards.' },
// ── LATE-NIGHT BATCH (Shots 30, 32, 34, 36, 37, 42) ──
{ title: 'Shot 30 — F1 cost alerts (founder text on threshold breach)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-12', phase: 'Late-night batch', note: 'Cron worker that compares todays cumulative spend (from the F1 vendor snapshot) against per-vendor thresholds (e.g. Anthropic > 30 USD/day, Replicate > 15 USD/day). On breach, sends a WhatsApp message + email to the founder with a one-tap link to the cost dashboard. Prevents the "I woke up to a 200 USD bill" failure mode. Thresholds live in env vars so they can be tuned without a deploy.' },
{ title: 'Shot 32 — Permanence dashboard (/admin/permanence.html)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Late-night batch', note: 'Admin view of every customer build that has been mirrored to Filecoin. Shows: site name, build date, Filecoin CID, mirror status, last successful verification time. Lets the founder confirm at a glance that the Permanence promise is real for every customer, not just the demo ones. Pulls from the trd-storage backend via a server-side API call so the admin browser never sees Filecoin credentials.' },
{ title: 'Shot 34 — PWA support (manifest.json + service worker)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Late-night batch', note: 'Every customer-built site is now installable as a Progressive Web App. Adds manifest.json (icons, theme color, start URL) and a stub service worker that caches the shell for offline access. Customers on iOS Safari and Android Chrome can now Add to Home Screen and the site opens fullscreen like a native app. Critical for the Restaurant OS positioning — owners want their staff to "install the restaurant" on every shared phone.' },
{ title: 'Shot 36 — Audit log (/audit-log.html)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Late-night batch', note: 'Per-account audit log of high-significance events: who logged in from where, who changed pricing tier, who deleted a booking, who exported customer data. Backed by an audit_events table written-to from middleware on the critical routes. Required for SOC2 readiness and for any future enterprise contracts that ask "show me the audit log feature".' },
{ title: 'Shot 37 — F2 revenue dashboard (/admin/revenue.html)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Late-night batch', note: 'Founder-only view of subscription revenue (Stripe MRR, ARR projections, churn rate, ARPU). Mirrors what Stripe itself shows but pre-aggregated for fast reads and split by tier (Free / Hobby / Pro / Team). Used for board-update prep and to spot churn spikes within hours instead of the end of the month. Falls back to dummy data in non-prod environments so the demo is safe to share.' },
{ title: 'Shot 42 — Journey presets (one-click templates)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Late-night batch', note: 'Pre-built journey templates that owners can clone in one click: "Reservation reminder 24h + 2h before", "Post-visit review request 4h after seating", "Win back 30 days inactive", "Birthday discount on file". Removes the cold-start problem of customers staring at an empty journey builder. Each preset is a JSON template that the cloning flow customizes with the owner business name. Critical for "5 minutes from signup to first running automation" goal.' },
// ── FINAL SHOTS (knowledge base + marketing + retention + affiliate + public help) ──
{ title: 'Shot 50 — FAQ knowledge base + restoration incident', status: 'completed', effort: 'shipped (after recovery)', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Final shots', testResult: 'pass', note: 'Built a customer-facing FAQ knowledge base at /faq.html — searchable, categorised, 60+ Q&A entries covering onboarding, billing, WhatsApp setup, booking flows. INCIDENT: my initial patcher overwrote the pre-existing /knowledge.html (a 2,138-line founder document) with the FAQ shell, losing the contents. RECOVERY: restored from a .bak file in the repo, and moved the FAQ to its own /faq.html. Lesson: every patcher now greps the destination file for existing meaningful content before writing. The .bak recovery file is what you are reading right now.' },
{ title: 'Shot 49a — Marketing scaffold (restaurant flagship)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Final shots', note: 'Industry-specific landing page generator. First flagship: /marketing/restaurant.html with hero, anti-frame comparison (DCS vs Resy/OpenTable/Toast), outcome cards, pricing tiers, FAQ accordion. Shared theme in /marketing/_styles.css so future verticals (salon, clinic, etc.) reuse the chrome and only swap content. SEO-tuned with meta tags + OG tags so each vertical can rank for "AI website for restaurants" etc.' },
{ title: 'Shot 28b — Retention emails (Resend templates + cron)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Final shots', note: 'Lifecycle email campaigns for retaining users: day-3 "are you stuck?" check-in, day-7 "your first booking is X days away from showing impact" pitch, day-14 "build velocity dropping — need help?" intervention. Each email is a Resend template, triggered by a Railway cron worker that scans user activity. Easy to disable per-user via the unsubscribe link. Counts toward Resend free-tier quota so monitored against monthly cap.' },
{ title: 'Shot 28c — Affiliate emails (commission tracking)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Final shots', note: 'Email touchpoints for the affiliate program (20% recurring for 12 months, shipped in Shot 15 — Wave 1C): welcome email on signup, monthly commission report, payout-day "we sent you X USD" confirmation. Also includes a one-click link to the affiliate dashboard showing live click → trial → paid conversion funnel. Critical for keeping affiliates engaged so they keep promoting.' },
{ title: 'Shot 51 — Public /help SEO landing', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Final shots', note: 'Public /help page that serves both as a search-index target ("DCS Network help") and as a single jumping-off point to /faq.html, /booking-status, /careers, and our affiliate/partner pages. Includes a contact-us form and a "ask Claude" search box (calls Claude Haiku with the question + FAQ context, returns 1-paragraph answer with citation back to the FAQ entry). Indexable by Google so we capture organic support search traffic.' },
// ── CRITICAL BUG FIXES (login + cache + sidebar) ──
{ title: 'Shot 52 — /login.html restoration (DCS-branded sign-in)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'Critical hotfix', testResult: 'pass', note: 'Critical incident: dcsai.ai/login was serving an OLD crypto/DePIN homepage instead of the DCS Network sign-in page, blocking sign-ups. Root cause: an early-April static-site fallback rule. Fix: rebuilt /login.html with proper DCS branding, Sign In + Sign Up tabs, Google + GitHub OAuth buttons (via supabase.auth.signInWithOAuth), magic-link fallback, and post-login redirect to /app22.html. Two follow-ups: Shot 52b hardcoded the Supabase URL + anon key (placeholder values broke the OAuth redirect) and Shot 52c restored the OAuth buttons that had been silently removed during a previous patcher run.' },
{ title: 'app22 cache-busting — Cloudflare Pages _headers', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-14', phase: 'Critical hotfix', note: 'Kills the "I shipped the fix but the cache is still serving the old HTML" loop that wasted 30+ minutes per dashboard fix. New /_headers file at trd-frontend root sets Cache-Control: no-cache, must-revalidate on every .html, plus extensionless clean URLs (/app22, /login, /app22/, /login/). Hashed static assets (JS, CSS, images) keep aggressive caching. Also added baseline security headers (nosniff, SAMEORIGIN, strict-origin, no-feature-permissions, HSTS) to match the trd-storage-frontend defaults. Verified live with curl -I https://dcsai.ai/app22.html showing the no-cache header.' },
{ title: '_headers cleanup — drop redundant .html entries', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-14', phase: 'Critical hotfix', note: 'Cosmetic follow-up to the cache-busting fix. Cloudflare Pages merges (instead of picks most-specific) matching rules — so listing both /app22.html and /*.html produced a doubled Cache-Control header value. Functionally harmless (browsers dedupe directives) but ugly. Removed the explicit .html entries from _headers, kept only the extensionless aliases (which /*.html does NOT match). Result: single clean Cache-Control: no-cache, must-revalidate, max-age=0 header on every request.' },
{ title: 'WhatsApp sidebar toggle — hide-lock fix', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-14', phase: 'Critical hotfix', note: 'Chevron rotated on click but the sub-items (Forms, Bookings, AI Assistant, Templates, Settings, Reviews, Carousels, Ads, Flows, Channels, Unified Inbox, Calendar, Broadcasts, Integrations, A/B Tests, Voice Calls, WA Status, Activity, Costs, Calendar Sync, Journeys, POS, Audit Log, Revenue, Embed, Help) stayed visible. Root cause: the inline display:none was being overridden by some downstream CSS or DOM rewrite, losing a specificity fight. Fix: added a class-driven CSS rule with !important (.sb-sub.sb-wa-hidden { display:none !important; }) AND have toggleSbGroup add the class alongside the inline style. Belt-and-suspenders so whichever interference is happening, one of the two wins. Marked with SB-SUB-HIDE LOCK sentinel comments.' },
// ── VISUAL-BUGS RCA · ARCHITECTURAL FIX (Zones 64-67) ──
{ title: 'Visual Bugs RCA · Diagnosis doc', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'RCA architecture', note: 'TRD_VISUAL_BUGS_RCA.md authored after a ~month of recurring "footer text invisible / dull body text / contrast looks wrong" reports. Root cause: a build pipeline of 17 sequential HTML post-processors, 3 disagreeing palette sources, and !important warfare. Same set of agents produced different visual output on different runs because each downstream consumer resolved colors independently. The architectural pattern that solved it: lock the answer once, freeze, parameter-thread to every consumer. Doc covers the full diagnosis + phased fix plan (Phase A → A2 → B → C-1) and serves as the template for any future RCA work (e.g. the Theme RCA that followed).' },
{ title: 'Phase A — Palette Lock (Zone 64)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'RCA architecture', note: 'src/services/palette-lock.ts — single source of truth for template colors. Exports lockPalette(template) which returns an Object.freeze() SealedPalette with provenance tracking (each token tagged with whether it came from css_vars, root_parse, or a fallback). The first sealed object produced for a session is then parameter-threaded to every downstream consumer. Replaces the previous "every injector parses css_vars again from scratch" pattern. Sentinel zone 64 protects the file.' },
{ title: 'Phase B — WCAG contrast correction (Zone 65)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'RCA architecture', note: 'Added to palette-lock.ts: at the moment of locking, compute WCAG 2.1 contrast for each text-on-bg pair and substitute a high-contrast token from a curated TEXT_FALLBACK_DARK/_LIGHT array whenever the original drops below 4.5:1 (AA standard for body text). Provenance log records each correction so the founder can see what changed. Railway logs now print [contrast-correction] tpl=... token=text #ffffff→#1a1a1a (ratio 1.16→15.03) etc. Hooked into Phase A so corrections happen at-lock-time, before any downstream consumer reads the palette.' },
{ title: 'Phase A2 — Palette injector (Zone 66)', status: 'completed', effort: 'shipped', owner: 'chat-Claude', addedDate: '2026-05-13', phase: 'RCA architecture', note: 'src/services/palette-injector.ts — first downstream consumer of the locked palette. Injects a LATE