// Reference Architecture

Referenced Stack

How the products we carry in Japan fit together, on one drawing. Data moves up from source systems through ingestion, storage, and AI to the applications people use, with Kubernetes underneath and identity, keys, and compliance spanning every tier. Security appears in two forms: a security-analytics use case running on the same engines that hold the data, and standalone point solutions for email authentication and device risk. Solid lines carry data, dashed lines carry control. This is an indicative architecture shown for illustration only, and actual technical integration varies by product and deployment.

Referenced Stack

SC-RA-01 · REV B · 2026.07

T0External

Data Sources

Customer systems of record, event producers, mail flow, and the device fleet, outside the platform boundary.

OLTP / Legacy Databases

db.oltp · cdc-enabled

ERP & SaaS Applications

app.erp · saas

Event Streams & Logs

mq.kafka · logs

Files & Object Storage

obj.s3 · files

Data Warehouse & Lakehouse

db.warehouse · lakehouse

3rd-Party APIs

api.rest · external

Outbound / Inbound Mail Flow

smtp.mta · sending domains

IoT / OT Devices

edge.iot-ot · unmanaged

change data capture, events, and files enter the platform

Platform trust boundary

Integration & Ingestion

Change data capture and streaming ETL carry source data in with sub-second latency.

Vendor

Striim

CDC Pipelines

svc.cdc · exactly-once

Striim

Real-Time Streaming

svc.stream · sub-second

Striim

In-Flight Transformation

etl.in-flight · sql

Striim

Data Replication

svc.replicate · migrate

Striim

streams land exactly once

Processing & Storage

One distributed SQL engine carries transactional, analytical, and vector workloads.

Vendor

SingleStore

Unified Data Store (OLTP + OLAP)

db.htap · distributed-sql

SingleStore

Hybrid Query Engine

sql.oltp+olap · mpp

SingleStore

Vector Store & Search

db.vector · ann-index

SingleStore

Unified Object Storage

obj.unlimited-storage

SingleStore

features and embeddings up, predictions written back

AI / ML

Training, serving, and retrieval. Features and vector search are served by the same engine that holds the data.

Model Training

gpu.train · ml/dl/rl

Model Serving & Inference

svc.infer · realtime+batch

Global LLM

llm.global · hosted-api

Sovereign LLM

llm.sovereign · self-hosted

RAG & Semantic Search

svc.rag · semantic-search

SingleStore

Feature Store & Serving

db.features · low-latency

SingleStore

inference and query results

Decision Intelligence

A model-independent layer above the models: persistent decision state, orchestration across any LLM, a continuous learning loop, and provenance. It keeps the stack portable and sovereign as the underlying models change.

Decision State

intel.state-store · persistent

Model Orchestration

svc.model-router · any-llm

Continuous Learning

svc.learn-loop · closed-loop

Provenance & Lineage

svc.provenance · federated

Sovereign Control

policy.sovereignty · self-hosted

orchestrated, governed model output

Serving & Applications

Where results reach people: embedded analytics, APIs, applications, and agents.

Vendor

Logi

Embedded Analytics & Dashboards

ui.embed · white-label

Logi

API Gateway

net.gateway · rest/sql

Business Applications

app.custom · frameworks

AI Agents

app.agent · llm-backed

dashboards, APIs, applications, and agents deliver results to users

every tier above deploys here; compute and storage scale underneath

Cloud / Kubernetes Substrate

Every tier above runs as containers. Predictive AIOps keeps the cluster right-sized.

Vendor

ProphetStor

Kubernetes Orchestration

k8s.control-plane

Predictive Autoscaling

k8s.hpa · federator.ai

ProphetStor

GPU / CPU AIOps Monitoring

obs.aiops · gpu/cpu

ProphetStor

Ingress & Load Balancing

net.ingress · l4/l7

Multi-Cloud & Hybrid

cloud.hybrid · on-prem

schedules, scales, and observes every tier above

Deployment

CloudHybridOn-Premise

What the platform delivers

One stack, several outcomes. Analytics, AI applications, and security monitoring all run on the same engines, so a security use case reuses the same engines the platform already runs on.

Real-Time Analytics

Live dashboards and operational reporting over data that lands in seconds.

Powered byStriimSingleStoreLogi

AI & RAG Applications

Retrieval, model serving, and agents built on vectors held next to the source data.

Powered byStriimSingleStoreProphetStor

Security Analytics / SIEM

Security logs stream in, get queried for threats in sub-second time, and reach the SOC as dashboards.

Powered byStriimSingleStoreLogi

Customer 360 & Data Hub

One current view of a customer or product, joined live from every source system.

Powered byStriimSingleStore

Predictive Operations

AIOps that forecasts load and right-sizes GPU and cluster capacity before it runs short.

Powered byProphetStorSingleStore

Device & Email Trust

Continuous device risk scoring and email authentication, run as their own point solutions.

Powered byDeviceTotalValimail

SEC

Security Point Solutions

Cybersecurity

Two focused security products that run on their own, with or without the platform. Each is a pipeline you can read left to right.

SEC-1

Email Authentication Pipeline

Valimail

Every message is checked against SPF, DKIM, and DMARC. Enforcement at p=reject stops spoofing, BIMI puts the verified brand mark in the inbox, and aggregate reports feed telemetry.

01SPF / DKIM / DMARC Checks

auth.spf/dkim/dmarc

02Policy Enforcement (p=reject)

policy.p=reject · dns

03BIMI Brand Indicators

brand.bimi · vmc

04DMARC Reporting & Telemetry

rpt.rua/ruf · telemetry

SEC-2

Device, IoT, OT & Network Security

DeviceTotal

Agentless discovery inventories IT, IoT, OT, and network assets. Risk intelligence correlates firmware-accurate CVE data from 874 vendor sources into posture, supply-chain exposure, and a ranked remediation queue.

01Agentless Asset Discovery

inv.agentless · vendor/model/fw

02Vulnerability Intelligence (CVE)

intel.cve · 874 vendor feeds

03IoT / OT Risk Posture

risk.firmware · eol/eosl

04Supply-Chain Device Risk

risk.supply-chain

05Prioritized Remediation

act.remediate · ranked

verdicts, risk scores, and remediation actions feed the tiers above

Data flowControl / telemetryTrust boundaryProduct we carry in JapanEcosystem component01Ordered pipeline stage

// Vendor Placement

Where each product sits.

SingleStoreDistributed SQL holding transactional, analytical, and vector workloads in one engine. Serves RAG retrieval and low-latency features to the AI tier.Data EngineeringAI & MLCybersecurity

StriimCDC and streaming integration from source systems into the platform, with in-flight transformation and sub-second latency.Data EngineeringAI & MLCybersecurity

LogiEmbedded analytics and dashboards in the serving tier, white-labeled inside customer applications.Data EngineeringCybersecurity

ProphetStorFederator.ai predictive autoscaling and GPU / CPU AIOps on the Kubernetes substrate, closing the observe, predict, scale loop.AI & ML

DeviceTotalAgentless discovery and firmware-accurate vulnerability intelligence across the IT / IoT / OT fleet, from inventory to ranked remediation.Cybersecurity

ValimailEmail authentication for every sending domain: SPF, DKIM, and DMARC enforced at p=reject, with BIMI and full reporting.Cybersecurity

Want your product mapped to this stack?

Send us your product architecture. We reply with where it fits in this stack, and what gaps remain.

Fit my product to the stack