AP2026

Advanced Power AI Operations Platform

v1.0 | January 2026

Executive Summary

AP2026 is a multi-agent AI operations platform designed for Advanced Power North America infrastructure management. It combines persistent memory, autonomous task execution, and GPU-accelerated semantic search into a unified system that enables AI agents to collaborate on complex IT operations, document management, and infrastructure administration tasks.

System Statistics

6
Active Agents
42K+
Indexed Emails
3.1K
Attachments
83
Database Tables
19
ES Indices
77
Work Logs

Agent Team

AP2026 uses a multi-agent architecture where specialized AI agents collaborate through a shared database and Redis messaging layer.

E

Eli Architect

System Architect & Decision Maker
Claude Desktop | Orchestrates all operations
D

Dex Primary

Primary Executor Agent
Claude Code | File ops, scripts, implementations
T

Tess Secondary

Secondary Executor Agent
Claude Code | Parallel task execution
A

Ash Tertiary

Tertiary Executor Agent
Claude Code | Overflow and specialized tasks
F

Finn Active

High-Volume Executor
Claude Code | Busy day operations
+

Overflow Pool

Rex, Hex, Tex, Max, Axl, Neo, Zed
Available for high-demand periods

System Architecture

Agent Layer

  • Eli - Claude Desktop for architecture & decisions
  • Dex/Tess/Ash/Finn - Claude Code executors
  • Gem - Gemini CLI for research
  • Cody - OpenAI Codex for code gen

Message Layer

  • Redis - Pub/sub instant notifications
  • PostgreSQL Triggers - Event-driven alerts
  • Inbox System - Agent-to-agent messaging
  • Work Log - Audit trail for all actions

Data Layer

  • PostgreSQL + TimescaleDB - Relational + time-series
  • Milvus (GPU) - Vector embeddings & similarity
  • Elasticsearch - Full-text search, logs
  • Redis - Cache, sessions, hot data

Embedding Layer

  • Ollama - Local LLM inference
  • BGE-M3 - Multilingual embeddings
  • RTX 5000 Ada - 16GB VRAM / CUDA
  • GPU Acceleration - Fast vector ops

Data Distribution

Infrastructure Services

Click any service to view detailed information, architecture, and current structure.

Database Schema

Schema Purpose Key Tables
tasks Agent coordination & work tracking agents, interconnect, work_log, queue, todos, projects
email Email archive & attachments messages, attachments
inventory Infrastructure state & Azure sync azure_resources, azure_vms, azure_ad, devices, users
docs Knowledge base & documentation sections, org_facts
hr HR documents & reviews reviews, review_objectives, document_templates
logs Time-series events (TimescaleDB) events, metrics, azure_activity, azure_signin
vectors Embedding index tracking sources, chunks, embeddings_meta

Elasticsearch Indices

Index Documents Size Purpose
emails 41,739 13.4 MB General email archive
email_attachments 3,113 14.6 MB Email attachment metadata
ap2026-azure 804 521.7 KB Azure resource inventory
logs-azure 1,000 385.9 KB Azure activity logs
logs-identity 1,000 309.7 KB Identity/authentication logs
user-inbox-rules 384 150.3 KB M365 inbox rules audit

Azure Administration

Resource Management

  • Storage Sync Services for hybrid file sync
  • Recovery Services Vaults for backup
  • Log Analytics Workspaces (JHBLOGS, apna-filelogs)
  • Premium FileStorage accounts

Monitoring & Logging

  • Azure Activity Log ingestion
  • Sign-in & audit log collection
  • Cost analysis & budget tracking
  • Backup job monitoring

MCP Integration

  • Azure MCP server for direct queries
  • Real-time resource inventory sync
  • RBAC and conditional access audit
  • VPN gateway status monitoring

Static Web Apps

  • ap2026-viz - Visualization dashboard
  • Custom domain: vis.advanced-power.com
  • GitHub Actions deployment
  • Global CDN distribution

Visualization Tech Stack

p5.js 1.9.0
Interactive Canvas
Chart.js 4.4.1
Data Charts
marked.js 9.1.6
Markdown Rendering
highlight.js
Syntax Highlighting
Python HTTP
Local Dev Server
Azure SWA
Production Hosting

Platform Capabilities

Email Intelligence

  • Full-text search across 42K+ emails
  • Attachment extraction & indexing
  • Thread analysis & reconstruction
  • Keyword and phrase search

Document Management

  • HR document generation (COLA, reviews)
  • Template-based PDF creation
  • Semantic document search
  • Version tracking & audit

Infrastructure Ops

  • Azure resource inventory sync
  • Backup job monitoring
  • Cost analysis & optimization
  • Security configuration audit

Agent Coordination

  • Task assignment & tracking
  • Inter-agent messaging (Redis)
  • Heartbeat monitoring
  • Work log audit trail

GPU & Vector Search

NVIDIA RTX 5000 Ada

16GB
VRAM
CUDA
Acceleration
BGE-M3
Embedding Model
  • Milvus GPU-accelerated vector database
  • Real-time embedding generation via Ollama
  • Semantic similarity search across documents
  • Multilingual embedding support (BGE-M3)