Skip to content

AI Platform Architecture

This folder is the technical home for the Feoda AI Platform — the corporate-wide AI infrastructure that delivers on the Corporate AI Strategy. It contains both the current as-built design and the target-state architecture, plus per-subsystem designs that inherit from the platform.

Contents

Document Purpose Status
Current State As-built description of what exists today (Phase 2 documentation assistant on Vercel) Draft
Target Architecture Target-state design of the corporate AI Platform: identity tiers, Portkey + OpenRouter gateway, Pinecone RAG, Nango connectors, Cloudflare hosting, Axiom audit, vertical-slice delivery plan V1–V5+. Draft
subsystems/ Per-subsystem designs (documentation assistant, NetSuite/ARM Q&A, sales/support AI, client-facing channels). Each subsystem inherits foundations (auth, providers, audit, governance) from the parent target architecture. To be populated. Not started

Scope

  • In scope: internal AI for all Feoda departments, identity-bound access, provider gateway, audit logging, governance hooks, and any client-facing AI channel that runs on Feoda infrastructure.
  • Out of scope: product-embedded AI inside ARM/ERP/EPM that is delivered as part of those products to the client. Those are governed by separate product-architecture documents.

Relationship to other documents

  • AI Strategy — the why. This folder is the how.
  • Approved Provider List — every external AI provider used in any subsystem must be listed there.
  • Action plans AP-01, AP-04, AP-05 — execution work that builds the platform foundations described here.

Status log

Date Note
2026-04-22 Folder created. current-state.md captured. Target architecture and subsystem designs pending.
2026-04-23 target-architecture.md drafted: identity tiers, gateway, RAG, connectors, hosting, audit, vertical-slice plan.