Architecture Reference · ADI D1–D4 · Three-SSOT Doctrine

ADI — Complete Architecture Reference

AI Delivery
Infrastructure

The complete four-layer architecture reference: D1 Discovery through D4 Semantic Corpus, three-SSOT doctrine, and Sacred Architecture L1–L6 implementation map.

Architecture Summary · ADI v1.0

AI Delivery Infrastructure (ADI) operates across four progressive layers (D1–D4) and three Single Sources of Truth (GitHub / Cloudflare KV / Cloudflare Workers). The Sacred Architecture L1–L6 maps these layers to business value. Layers D1–D3 constitute the machine infrastructure (complete). Layer D4 — the Semantic Corpus is the public content layer now entering production. Created by AiVenture S.R.L. (CUI 51415878, Bucharest, Romania).

📅 May 2026
🏗 D1 · D2 · D3 complete — D4 active
Bitcoin-anchored · 7 innovations verified

The Four Layers

D1 → D4: the complete
ADI layer architecture

ADI is organized into four progressive layers. Each layer addresses a distinct aspect of AI system understanding — from raw crawl accessibility to semantic embedding reinforcement.

D1 — DISCOVERY
AI Crawler Access Layer
Makes the domain discoverable and crawlable by AI systems. Primary files: robots.txt (per-crawler Allow rules), llms.txt (LLM governance declaration), ai.json (master signal file). Status: ✅ COMPLETE
ClaudeBot · GPTBot · Google-Extended · PerplexityBot · CCBot
D2 — DECLARATION
Entity & Intent Declaration
Declares who the entity is and what it does. Files: entities.json (entity registry), intents.json (intent mapping schema v1.1), governance.json (AI policy + EU AI Act). Status: ✅ COMPLETE
entities · intents · governance · aliases · session
D3 — DECISION
Proof & Permission Layer
Cryptographic proofs and granular access decisions. Files: ai-proof.json (OTS-verified SHA-256), allow-lane-matrix.json (per-AI permission matrix). Status: ✅ COMPLETE
SHA-256 · OTS · Bitcoin blocks 944442–949191
D4 — SEMANTIC CORPUS
Public Semantic Content Layer — NOW ENTERING PRODUCTION
Canonical definition pages (1,200–2,500 words each), FAQ blocks for AI answer extraction, JSON-LD schemas embedded in HTML, semantic internal linking network, repeated entity association across all pages. This layer directly improves: Vectorial Brand Representation (ofs5), LLM Embedding Proximity (ofs4), Perplexity Citation Score (ofs7), Gemini Grounding (ofs2), AI Answer Layer Coverage (fp26). Target: Phase 1 — 5 canonical pages. Phase 2 — 25 pages. Phase 3 — full semantic corpus.

Infrastructure Doctrine

The three-SSOT
doctrine

ADI uses three Single Sources of Truth with strict separation of concerns. No duplication across SSOTs is permitted. Each SSOT owns its layer exclusively.

SSOT 1: GitHub = CONTENT
The GitHub repository is the canonical source for all site files, pages, and static assets. Every HTML page, JSON file, and configuration is authored in GitHub. No content is authored anywhere else. Cloudflare Pages deploys directly from the GitHub main branch. This is the single source of truth for what the site says.

SSOT 2: Cloudflare KV = SIGNALS
Cloudflare Workers KV is the single source of truth for all AI signal data — the real-time values served to AI crawlers at the edge. KV stores are hydrated from GitHub on deploy but serve independently at 0ms latency globally. No signal data is authored directly in KV; it is always pushed from GitHub.

SSOT 3: Cloudflare Worker = EXECUTION
The Cloudflare Worker is the single source of truth for all business logic — routing, permission enforcement, signal serving, rate limiting, and EU AI Act header injection. Workers read from KV (signals) and serve to AI crawlers. No business logic lives in GitHub files or KV values; it is always in the Worker.

The three-SSOT doctrine eliminates the most common ADI failure mode: signals drifting out of sync because they are authored in multiple places. When every AI crawler request is served by a Worker reading from KV hydrated by GitHub, the entire system is deterministic, auditable, and cacheable at Cloudflare edge nodes globally in under 50ms.

Sacred Architecture

L1–L6 mapped to
D1–D4 layers

The Sacred Architecture L1–L6 is the business value framework. The D1–D4 layers are the technical implementation framework. Here is how they map.

Layer Sacred Architecture ADI Layer Status
L1Edge · AI Signal DeliveryD1 Discovery + D3 Decision✅ Complete
L2Semantic · Entity DeclarationD2 Declaration + D4 Semantic Corpus⚠️ D2 Complete · D4 Active
L3Voice-CRM · Conversational AID4 Semantic (voice schema)🔲 Planned
L4KAI369 · Vectorial CRMCommercial product layer🔲 Planned
L5Economic Twin · AI ProspectingCommercial product layer🔲 Planned
L6Predictive Mastery · BICommercial product layer🔲 Planned

Frequently Asked Questions

ADI Architecture — Questions
& Answers

What are the four ADI layers (D1–D4)? +
D1 Discovery (robots.txt, llms.txt, ai.json — AI crawler access), D2 Declaration (entities.json, intents.json, governance.json — entity and intent declaration), D3 Decision (ai-proof.json, allow-lane-matrix.json — cryptographic proof and permissions), D4 Semantic Corpus (canonical definition pages, FAQ content, long-form semantic pages — public content that reinforces entity embedding in AI systems).
What is the three-SSOT doctrine? +
The three-SSOT doctrine defines three Single Sources of Truth: GitHub = CONTENT (canonical source for all files), Cloudflare KV = SIGNALS (real-time edge delivery of AI signal values), Cloudflare Worker = EXECUTION (all business logic for routing, permissions, and signal serving). No duplication across SSOTs is permitted.
Why is D4 the current bottleneck? +
D1, D2, and D3 are complete — the machine infrastructure is fully deployed. D4 (Semantic Corpus) is the bottleneck because it requires producing long-form canonical content pages that reinforce brand entity associations in LLM embedding spaces. Without D4, AI systems receive the correct machine signals but encounter insufficient public semantic content to build strong vectorial representations of the brand.
What specific AUDIT-AI signals does D4 improve? +
D4 Semantic Corpus directly improves: ofs4 LLM Embedding Proximity (currently 78), ofs5 Vectorial Brand Representation (76), ofs7 Perplexity Citation Score (80), ofs2 Gemini Grounding Verified (82), fp26 AI Answer Layer Coverage (84). These are the five lowest-scoring off-site signals — all improvable through consistent D4 content deployment.
How does ADI FACTORY scale to multiple sites? +
ADI FACTORY v1.0 defines a tenant-first KV key structure (tenant:domain:file) enabling multi-tenant signal delivery from a single Cloudflare Worker. The COBOL-style GET /v1/{tenant}/{file} pattern provides edge delivery at the 12-million-site scale. Each tenant onboards via git add + CNAME — no per-tenant Worker deployment required.
What is the edge.5thelement.ai endpoint? +
edge.5thelement.ai is the ADI Edge Central — the Cloudflare Worker endpoint that serves AI signals for all ADI-managed domains. It implements the W-L1-EDGE pattern: RFC 9309 compliant robots.txt, programmatic blank-line generation, allow-lane-matrix enforcement, and Cloudflare KV-backed signal delivery with ZERO LLM in the delivery path.

Related

Explore the ADI architecture

Audit your ADI readiness — 167 signals
AIVENTURE S.R.L. · eu-ai-audit.eu · [email protected] · WhatsApp +40 737 123 540
💬 Order via WhatsApp Run Free Audit →