# AI Periodic Table → Engram Architecture Mapping
A comprehensive tree-structure mapping of each AI Periodic Table element to Engram's
implementation, with a conceptualized roadmap for future development.
*Baseline: January 5, 2026*
[📊 Interactive Matrix](ai-periodic-table-matrix.html)
[📋 Business Plan](/00-strategy/business-plan-ai-periodic-table.html)
[📚 Wiki](wiki/ai-periodic-table.md)
Element Reference
Symbol
Element
Row
Status
Description
**Pr**
Prompts
R1
🟢
Agent system prompts (Elena, Marcus, Sage)
**Em**
Embeddings
R1
🟢
Zep automatic vectorization + Tri-Search
**Lg**
LLM
R1
🟢
Claude, Gemini, Azure APIM Model Router
**Fc**
Function Call
R2
🟢
20+ tools across agents (GitHub, Graph, memory)
**Vx**
Vector
R2
🟢
Zep built-in vector store
**Rg**
RAG
R2
🟢
Auto context injection every agent turn
**Gr**
Guardrails
R2
🟢
Azure Entra ID, RBAC, tenant isolation
**Mm**
Multimodal
R2
🟢
Imagen, VoiceLive, diagram generation
**Ag**
Agent
R3
🟢
LangGraph personas with tools + memory
**Ft**
Finetune
R3
🔴
Not implemented (roadmap: LoRA, Azure ML)
**Fw**
Framework
R3
🟢
Temporal durable workflows
**Rt**
Red-team
R3
🟡
Basic validation, Golden Thread
**Sm**
Small Models
R3
🔴
Not implemented (roadmap: Phi-4, edge)
**Ma**
Multi-agent
R4
🟢
Agent delegation via Temporal
**Sy**
Synthetic
R4
🟢
Story + diagram + image pipeline
**Gk**
Graph Knowledge
R4
⭐
Zep temporal knowledge graph (unique!)
**In**
Interpret
R4
🔴
Not implemented (roadmap: explainability)
**Th**
Thinking
R4
🟢
LangGraph multi-step reasoning
Visual Overview
AI PERIODIC TABLE → ENGRAM ARCHITECTURE ═══════════════════════════════════════════════════════════════════════════════
┌─────────────────────────────────────┐
│ ENGRAM PLATFORM │
│ Brain (Zep) + Spine (Temporal) │
└───────────────────┬─────────────────┘
│
┌───────────────┬───────────────┬───┴───┬───────────────┬───────────────┐
│ │ │ │ │ │ ┌────▼────┐ ┌─────▼─────┐ ┌─────▼─────┐ │ ┌───────────▼───────────┐ │ │ C1 │ │ C2 │ │ C3 │ │ │ C4 │ │ │ Reactive│ │ Retrieval │ │ Orchestr. │ │ │ Validation │ │ └────┬────┘ └─────┬─────┘ └─────┬─────┘ │ └───────────┬───────────┘ │
│ │ │ │ │ │
▼ ▼ ▼ ▼ ▼ ▼ ┌─────────┐ ┌─────────┐ ┌──────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ │Pr Em Fc │ │Vx Ft Sy │ │Rg Fw Gk │ │Gr Rt In │ │Lg Mm Sm │ │ Th │ │Ag Ma │ │ │ │ │ │ │ │ │ │ │ └─────────┘ └─────────┘ └──────────┘ └─────────┘ └─────────┘ └─────────┘
Summary Statistics
Status
Count
Elements
🟢 Strong
12
Pr, Em, Lg, Fc, Vx, Rg, Gr, Mm, Ag, Fw, Ma, Sy, Th
🟡 Emerging
1
Rt
🔴 Gap
3
Ft, Sm, In
⭐ Unique
1
Gk (Knowledge Graph)
Element-by-Element Tree Mapping
🟢 ROW 1: PRIMITIVES (Foundation Layer)
R1 PRIMITIVES ├── Pr (Prompts) ──────────────────────────────────── 🟢 STRONG │ ├── Current Implementation │ │ ├── backend/agents/elena/agent.py → ElenaAgent.system_prompt │ │ ├── backend/agents/marcus/agent.py → MarcusAgent.system_prompt │ │ ├── backend/agents/sage/agent.py → SageAgent.system_prompt │ │ └── backend/llm/claude_client.py → SAGE_SYSTEM_PROMPT │ │ │ ├── Key Features │ │ ├── Context-First Requirements Framework (Elena) │ │ ├── Calm in the Storm Leadership (Marcus) │ │ ├── Sage Meridian Storyteller persona │ │ └── Enterprise context injection via EnterpriseContext.to_llm_context() │ │ │ └── Roadmap │ ├── [ ] Prompt versioning and A/B testing │ ├── [ ] Dynamic prompt composition based on user role │ └── [ ] Prompt performance analytics dashboard │ ├── Em (Embeddings) ───────────────────────────────── 🟢 STRONG │ ├── Current Implementation │ │ ├── Zep Cloud → Automatic embedding on message ingestion │ │ ├── backend/memory/client.py → search_memory() │ │ └── backend/etl/ingestion_service.py → Document chunking + embedding │ │ │ ├── Key Features │ │ ├── Tri-Search (Keyword + Vector + Graph) │ │ ├── Automatic entity extraction │ │ └── Cross-session knowledge compounding │ │ │ └── Roadmap │ ├── [ ] Custom embedding models for domain-specific terms │ ├── [ ] Embedding drift detection │ └── [ ] Multi-modal embeddings (text + image) │ └── Lg (LLM) ──────────────────────────────────────── 🟢 STRONG
├── Current Implementation
│ ├── backend/agents/base.py → FoundryChatClient
│ ├── backend/llm/claude_client.py → ClaudeClient (Anthropic)
│ ├── backend/llm/gemini_client.py → GeminiClient (Google)
│ └── Azure APIM Gateway → Model Router pattern
│
├── Key Features
│ ├── Multi-model orchestration (Claude for text, Gemini for visuals)
│ ├── Automatic fallback (Anthropic API → Azure APIM)
│ ├── Model Router for dynamic model selection
│ └── Token-based authentication (Azure AD + API keys)
│
└── Roadmap
├── [ ] Cost-aware model routing (use cheaper models for simple tasks)
├── [ ] Latency-optimized routing (edge deployment)
└── [ ] Model performance comparison dashboard
🟢 ROW 2: COMPOSITIONS (Integration Layer)
R2 COMPOSITIONS ├── Fc (Function Call) ────────────────────────────── 🟢 STRONG │ ├── Current Implementation │ │ ├── backend/agents/github_tools.py │ │ │ ├── create_github_issue_tool │ │ │ ├── update_github_issue_tool │ │ │ ├── get_project_status_tool │ │ │ ├── list_my_tasks_tool │ │ │ └── close_task_tool │ │ │ │ │ ├── backend/agents/elena/agent.py │ │ │ ├── search_memory_tool │ │ │ ├── send_email_tool (Microsoft Graph) │ │ │ ├── list_onedrive_files_tool │ │ │ ├── save_to_onedrive_tool │ │ │ ├── trigger_ingestion_tool │ │ │ ├── run_golden_thread_tool │ │ │ ├── analyze_requirements │ │ │ ├── stakeholder_mapping │ │ │ ├── create_user_story │ │ │ └── delegate_to_sage │ │ │ │ │ ├── backend/agents/marcus/agent.py │ │ │ ├── create_project_timeline │ │ │ ├── assess_project_risks │ │ │ ├── create_status_report │ │ │ ├── estimate_effort │ │ │ ├── delegate_to_sage │ │ │ ├── start_bau_flow_tool │ │ │ └── check_workflow_status_tool │ │ │ │ │ └── backend/agents/sage/agent.py │ │ ├── generate_story │ │ ├── generate_diagram │ │ └── generate_visual │ │ │ └── Roadmap │ ├── [ ] MCP (Model Context Protocol) server implementation │ ├── [ ] Tool discovery and dynamic registration │ └── [ ] Tool execution analytics and cost tracking │ ├── Vx (Vector) ───────────────────────────────────── 🟢 STRONG │ └── Zep Cloud → Built-in vector store + Tri-Search │ ├── Rg (RAG) ──────────────────────────────────────── 🟢 STRONG │ └── Automatic context injection in every agent turn │ ├── Gr (Guardrails) ───────────────────────────────── 🟢 STRONG │ └── Azure Entra ID + RBAC + Tenant isolation │ └── Mm (Multimodal) ───────────────────────────────── 🟢 STRONG
├── Text → Image generation (Gemini/Imagen)
├── Voice → Text → Voice (VoiceLive)
└── Story + Image + Diagram bundling
🟡 ROW 3: DEPLOYMENT (Production Layer)
R3 DEPLOYMENT ├── Ag (Agent) ────────────────────────────────────── 🟢 STRONG │ ├── Elena (Business Analyst), Marcus (PM), Sage (Storyteller) │ └── LangGraph StateGraph + Memory-enriched context │ ├── Ft (Finetune) ─────────────────────────────────── 🔴 GAP │ └── Roadmap: LoRA adapters, Azure AI Fine-tuning │ ├── Fw (Framework) ────────────────────────────────── 🟢 STRONG │ └── Temporal Server → Durable workflow orchestration │ ├── Rt (Red-team) ─────────────────────────────────── 🟡 EMERGING │ └── Basic validation, Golden Thread checks │ └── Sm (Small Models) ─────────────────────────────── 🔴 GAP
└── Roadmap: Phi-4, Granite, edge deployment
⭐ ROW 4: EMERGING (Innovation Layer)
R4 EMERGING ├── Ma (Multi-agent) ──────────────────────────────── 🟢 STRONG │ └── Agent-to-agent delegation via Temporal workflows │ ├── Sy (Synthetic) ────────────────────────────────── 🟢 STRONG │ └── Story + diagram + image generation pipeline │ ├── Gk (Graph Knowledge) ──────────────────────────── ⭐ UNIQUE DIFFERENTIATOR │ ├── Zep Cloud → Temporal Knowledge Graph │ ├── Automatic entity/relationship extraction │ ├── Knowledge compounds across sessions │ └── WHY UNIQUE: Dynamic context orchestration vs static RAG │ ├── In (Interpret) ────────────────────────────────── 🔴 GAP │ └── Roadmap: "Why did you say that?", attention viz, source attribution │ └── Th (Thinking) ─────────────────────────────────── 🟢 STRONG
└── LangGraph multi-step reasoning + tool-augmented thinking
Implementation Matrix (Effort vs. Impact)
quadrantChart
title AI Periodic Table Roadmap Matrix
x-axis Low Effort --> High Effort
y-axis Low Impact --> High Impact
quadrant-1 High Impact / High Effort
quadrant-2 High Impact / Low Effort
quadrant-3 Low Impact / Low Effort
quadrant-4 Low Impact / High Effort
"Sm (Small Models)": [0.8, 0.3]
"Ft (Finetune)": [0.7, 0.4]
"In (Interpret)": [0.6, 0.8]
"Rt (Red-team)": [0.6, 0.7]
"Pr (Prompts)": [0.2, 0.2]
"Em (Embeddings)": [0.3, 0.8]
"Lg (LLM)": [0.2, 0.9]
"Gk (Graph)": [0.4, 0.95]
"Ag (Agents)": [0.4, 0.85]
"Sy (Synthetic)": [0.3, 0.6]
Recommendation
- Quadrant 2 (High Impact, Low Effort): Focus on Embeddings (Em), LLM (Lg), and Synthetic (Sy) pipelines. (🟢 Completed)
- Quadrant 1 (High Impact, High Effort): The Graph Knowledge (Gk) and Agents (Ag) layers are the core differentiators. (⭐ In Progress)
- Quadrant 4 (Low Impact, High Effort): Small Models (Sm) and Finetuning (Ft) should be deferred until strict cost/latency requirements emerge.
Element-by-Element Tree Mapping
(See full breakdown below)