AMODX Sprint 4: From Functional CMS to Commercial-Grade Operating System

By Andrei Roman

Principal Architect, The Foundry

Sprint 4 was massive. AMODX went from a functional CMS to a commercial-grade operating system. Complex visual features implemented. Infrastructure hardened against timeouts. Critical SEO flaws fixed. Research and Outreach module started.

Here is what shipped.

The Visual and Editorial Upgrades

The Coverflow Carousel

Started with raw CSS physics. It trembled on Safari. Tried complex JS math. It drifted on the last item. Arrived at Swiper.js. Result: butter-smooth, hardware-accelerated 3D carousel that snaps perfectly and looks native.

Dynamic Post Grid (Google Style)

Added List Layout option to mimic Google SERP snippets. Fixed Limit: 0 logic to show all posts.

Critical SEO fix: Moved data fetching from client-side (useEffect) to server-side (page.tsx). Ahrefs and Google now see the links immediately in raw HTML. Client-side rendering hid content from crawlers. Server-side rendering exposes everything.

Unified Editor UI

Refactored all 12 plugin editors (Hero, Pricing, Features, etc.) to share a clean card aesthetic. White background. Gray border. Header bar. Fixed padding issues and grid layouts in admin panel.

The Time Machine (Versioning)

Implemented copy-on-write architecture. Updating a page now snapshots the current LATEST state to version N.

Added History UI in admin panel. View previous versions. Restore them (forward rollback) with one click.

System routes (homepage and contact) now locked at backend level. Prevents accidental breakage of critical pages.

The X-Ray (Content Graph)

Integrated ReactFlow + Dagre for auto-layout visualization of site structure.

Orphan detection: System flags pages with zero incoming links. These pages exist but nobody can reach them. Either add links or delete them.

Taught backend crawler to understand PostGrid dynamic links and Navbar links. Removed false positives. Pages were being flagged as orphans when they were actually linked from navigation or blog grids.

Implemented pagination and try-catch blocks in crawler. One bad node no longer crashes the entire graph.

Infrastructure and Performance

Diagnosed why Content Graph and Content Lists were timing out. The infrastructure build was not applying. Lambda functions were running with 128MB memory and 3-second timeouts. Not enough for graph traversal on larger sites.

Forced tsc to compile the backend. Deployed 1024MB / 29s timeout Lambdas. llms.txt, Content Lists, and Content Graph now load instantly, even on larger sites.

Fixed TypeScript configuration in the build pipeline. tsconfig.json was missing in backend directory. Fixed ESM import errors (.js extensions). Code now actually updates on deploy instead of serving stale builds.

The Research Signals Module and Outreach Module

Initially I wanted to use Ayrshare and Serper but then rejected the expensive SaaS path. Ayrshare and Serper cost $50-150 per month. For what? Searching Reddit and posting replies. Not worth it.

New Architecture:

• Sensor: Brave Search API (free tier) integrated into backend. Search Reddit, Twitter, forums for signals.

• Execution: Local Playwright via MCP (free, undetectable). Automate posting without paying SaaS ransoms.

• Storage: Defined Signal schema and CRUD for tracking leads. Backend and UI ready. MCP integration implemented. Still testing.

Operational Hygiene

Audit Logs: Transformed from mangled JSON to clean, human-readable Activity Feed with proper Actor and Target tracking. You can now see who changed what and when.

Security: Tightened Cognito tokens to 1-week expiry. Prevents infinite sessions. Users must re-authenticate periodically.

AI Discovery: Added link rel alternate openai-feed headers to all pages. AI agents (ChatGPT, Perplexity) automatically discover your product catalog. Your content becomes LMO-native.

What Was Deferred (Intentionally)

Multi-language (I18n): High complexity, low immediate ROI. Nobody asked for it yet.

Tenant Export and Import: Pushed to Sprint 5. Safety net for backups and migrations.

Admin AI Integration: Nobody requested this. Claude integration in admin panel can wait.

Current State

AMODX is now SEO-native. Server-side rendering exposes content to crawlers immediately. OpenAI feed headers make product catalogs discoverable to AI agents.

Version-controlled. Copy-on-write snapshots every change. One-click rollback via History UI.

Visually polished. Coverflow carousel runs smooth on all browsers. Admin UI uses consistent clean card design. Post grids mimic Google SERP layouts.

Infrastructure hardened. Backend no longer gasping for RAM. 1024MB Lambdas handle graph traversal on large sites without timeouts.

Clear path to agentic features without paying monthly SaaS ransoms. Research Signals module uses Brave Search (free) and local Playwright (free) instead of Ayrshare and Serper ($50-150/month).

Sprint 4 shipped. AMODX moved from functional to commercial-grade.