Skip to main content
Skip to main content
Launching Dec 31142 days
Japan Toll Receipts
Technical Engineering / Architecture

Our Engineering Architecture

How Japan Toll Receipts works with the ETC Usage Inquiry Service (MEISAI)

The engineering behind the platform — not a customer walkthrough. It explains, layer by layer, how a centralized control plane, a single safety governor, and private workers coordinate controlled, scalable access to the ETC Usage Inquiry Service (MEISAI).

View the System Diagram Safety Controls

For the cleanest PDF, disable browser “Headers and footers” in the print dialog. (A4 landscape · 4 pages)

Share this architecture overview
Control points
Centralized
Concurrency
Governed
External access
Authorized
Scaling
Layered

See it live

Smart scaling, simulated live

Watch, in real time, how the M-Level Governor raises capacity as report demand rises and safely returns toward baseline once the backlog clears.

Live smart-scaling simulation
in sync · at capacity

Watch how the M-Level Governor increases capacity as report demand rises, then safely returns toward baseline after the backlog clears.

Incoming traffic5/tick
Queue backlog5
More backlog raises capacity; clearing it lowers capacity.
M-Level ladder (available capacity)M10
M4
M6
M8
M10
M12
M20
M30
M40
M48
Worker slots5 busy · 5 ready

At M10, every available slot is working the backlog — demand and capacity are in sync.

Demand story
1
Calm / Baseline
2
Demand Spike
3
Scaling Response
4
Backlog Clearing
5
Cooldown / Return to Baseline

Demand Spike: Report requests increase and the queue begins to grow.

Watch it live

The AI pipeline, directed by Sora — in real time

A customer request enters at the top-left and flows right through the queue, scheduler, database, report engine, safety governor and a worker (A–D) to MEISAI — then loops back left through central processing, analysis, status fan-out (PDF / Email / Dashboard) and consolidation, landing as an email back with the customer. Sora — Director of AI Operations — assigns each step and narrates progress live.

AI pipelineSora (AI advisor)
JTR AI Pipeline · Live
SSora
SoraDirector

A customer just requested a report — let’s get to work.

Tanaka, H. · ETC card #3018 · Monthly report
Customer requestYouworking

A customer just requested a report — let’s get to work.

Queuequeued
Schedulerqueued
Databasequeued
Report enginequeued
Safety governorqueued
Worker A–Dqueued
MEISAI sitequeued
Central processingqueued
Analyze · 31 checksqueued
Status fan-outqueued
PDFEmailDash
Consolidatequeued
Customer receivesYouqueued
5 AI teamSSoraSoraYYukiYukiHHanaHanaAAoiAoiRRinRin
Log stream · live polling
[02:01:14] ▸ Inbound request · ETC card #3018 · monthly
※ Demo visualization. Sora directs the pipeline and hands each step to an AI analyst. No real customer data is shown.

System Diagram

How a request flows through the system

L01
Customer
USERS
Users / Customers
Report requests · schedules · settings
L02
Application
WEB
JapanTollReceipts Website
Dashboard · admin · settings
L03
Control plane
Centralized System
QUEUE
Queue
SCHED
Scheduler
DATA
Database
ENGINE
Report Engine
L04
Governor
GATE
Safety Governor
Limits simultaneous activity · the single control point for external work
L05
Workers
Private Worker Processes
W-A
Worker A
W-B
Worker B
W-C
Worker C
W-+
Future
L06
External
EXTERNAL
ETC Usage Inquiry Service (MEISAI)
Customer-authorized access only
L07
Processing
PROCESS
Central Report Processing
Validation · PDF / CSV generation
L08
Delivery
Customer Delivery
PDF
PDF / CSV
MAIL
Email
DASH
Dashboard
STATUS
Status

// Customers never access external systems directly — everything routes through Queue → Governor → Private Workers.

Safety Governor

Controls simultaneous activity, avoids spikes

JapanTollReceipts is designed with a safety governor that limits how many background sessions run at once. Rather than unlimited parallel activity, requests pass through controlled scheduling — designed to avoid traffic spikes and keep processing stable.

Incoming
GATE
Active · limited
Queued

Even with many requests, only a small controlled number run at once — the rest wait safely. (Tap or hover any part for details.)

Adding workers never automatically increases simultaneous external access. The governor is always in control.

Queue & Scheduling

Why requests are queued

Controlled timing

Work runs at the most appropriate time.

Duplicate prevention

Duplicate requests can be detected.

Spike avoidance

Requests are spread out, not run all at once.

Spread schedules

Daily reports are not all pulled at once.

Prioritization

Manual/urgent requests can be prioritized.

Status tracking

Report status is easier to track.

Private Workers

Private worker processes

Workers are not customer-facing websites.
Workers process approved background jobs only.
Workers report status back to the centralized system.
More workers can increase internal processing capacity.
Workers → Governor → External
W-A
W-B
W-C
W-+
Safety Governor

Internal capacity can grow; external access is always governed.

Data Integrity

Designed for traceability and integrity

QueuedScheduledWaitingProcessingValidatingDeliveredDelayedRetryingNeeds ReviewFailed · with reason
Each job is tied to an account, card, and report period.
Delivery is validated before reports are sent.
Status and errors are tracked.
Uncertain jobs can be marked for review.
Sensitive credentials are never shown publicly.
Workflows are recorded centrally.

Autonomous Self-Repair

Self-healing delivery that never stops at scale

A dedicated recovery lane runs continuously.
Heavy actions run in the background.
One job can never freeze the whole run.
An AI learns from errors and auto-repairs.
Main run + 2 dedicated recovery → all at once when done
Main run · M10 → M12 → M14…
+2 recovery (always reserved)
When done · process all errors at once (no limit)

During the main run, at least two workers stay reserved solely for errors; once it finishes, every queued error is processed at once with no fixed limit, draining to zero. The safety governor still caps external concurrency at all times.

Scaling Model

Scale in layers, keep the controls

L1
Website Layer

Customer traffic, dashboards, settings, requests.

L2
Coordination Layer

Queue, scheduling, Database, status monitoring.

L3
Worker Layer

Background processing, governed external access, delivery.

JapanTollReceipts is designed to scale internal capacity without removing the controls that limit simultaneous external activity.

Smart Scaling with the M-Level Governor

Scales gradually and safely with demand

The scheduler tunes its parallel capacity in controlled stages called the "M-Level". It always holds at least M10 (10 in parallel) and steps up — carefully, two rungs at a time — only when the queue grows and MEISAI responses stay healthy. As the backlog clears, it steps gradually back down toward the M10 baseline.

JapanTollReceipts uses controlled automation, account isolation, traffic monitoring, and governed scaling to process ETC report requests responsibly as customer demand grows.

M4
M6
M8
M10
M12
M14
M16
M20
M24
M30
M40
M48
ceiling 50 sessions

Baseline M10 → up to M48, one controlled step at a time. Concurrent sessions always stay under the 50-session ceiling so MEISAI/NEXCO is never overloaded.

Scales up when
  • The queue backlog is growing
  • Estimated clearance time exceeds the target window
  • MEISAI response times remain stable
  • Login, download & delivery error rates stay low
Scales down when
  • The backlog has cleared
  • Work fits comfortably inside the target window
  • MEISAI responses start to slow (it slows down to protect them)
  • Error rates rise

Adding capacity never makes one customer’s report faster. Each customer’s MEISAI work (log in → pull cards → build report → log out) stays isolated and sequential so data can never mix between accounts. The M-Level only controls how many customers are processed in parallel.

Smart scaling loop
Report queue
M-Level Governor
Worker pool
MEISAI pulls
Reports delivered
Metrics & alerts
back to Governor

The “Sora” AI advisor analyses trends and recommends adjustments — final execution stays controlled by the deterministic governor that enforces the safety rules. AI recommendations do not bypass safety rules; final execution is controlled by the M-Level Governor.

Safety guardrails
Per-account locking
One account is never processed twice at once
Hard session ceiling
Concurrent MEISAI sessions stay capped
Rate & error monitoring
Response time, errors & retries watched
Alert notifications
The team is paged when thresholds cross
Emergency baseline (M10) control
Admins can instantly return to baseline
Audit logs
Every scaling decision is recorded
Data integrity
Accounts stay isolated; records never mix

The system is designed to avoid uncontrolled traffic spikes by using session ceilings, queue management, retry limits, and automatic throttling.

“M10” does not mean 10 workers are always busy. It means 10 safe processing slots are ready. When the queue is empty the system stays healthy and idle; when reports are waiting, those slots activate automatically.

Capacity growth example
M10
Early access — baseline
M8–M12
Moderate demand
M20–M40
High daily volume
M48
Max — under 50-session ceiling
Admin visibility & accountability
Current M-Level
Queue backlog
Estimated clearance
Active sessions
Alerts
Last scaling event
Scheduler proof
Audit history

Admins can pause, override, return to M10, or adjust thresholds when needed. Every scaling event is recorded.

JapanTollReceipts was built with controlled automation, account isolation, audit visibility and smart scaling — so ETC report processing can grow safely without creating uncontrolled traffic or compromising customer data integrity.

Production currently runs an in-process elastic worker pool, and the architecture is ready for future Kubernetes/HPA/KEDA or other external autoscaler integration — so the platform can grow from early access to large-scale daily processing without redesigning its core.

JapanTollReceipts is not affiliated with NEXCO or the official ETC Usage Inquiry Service.

Risk Controls

How we prevent common risks

Risk
Traffic spike
Prevention

Queueing, scheduling, and safety-governor controls limit how much background activity can proceed at one time.

Risk
Duplicate processing
Prevention

The system checks for matching queued, active, or recently completed work before creating another job.

Risk
Too many workers running
Prevention

Workers still require approval from the safety governor before starting controlled background work.

Risk
Report status confusion
Prevention

Each workflow has status tracking so customers and administrators can see whether work is queued, processing, delivered, delayed, or needs review.

Risk
Failure loops
Prevention

Failures can trigger delay, retry, pause, or review behavior instead of uncontrolled repeated attempts.

Risk
Customer data mix-up
Prevention

Report workflows are associated with customer, account, card, period, and delivery settings before reports are delivered.

Engineering FAQ

How do you prevent…?

10 questions

Ready to see how this system can support your ETC reporting workflow?

Organized ETC usage records and processing history for the customer’s own internal review.

Independent Service Notice

JapanTollReceipts is an independent service. It is not the ETC Usage Inquiry Service (MEISAI) operator, NEXCO, or a toll road operator. The platform helps customers organize authorized ETC usage records through a structured reporting workflow.

Try FreeEarly Access
JapanTollReceipts
System Architecture Map
Controlled, Queue-Based Report Processing — Engineering Briefing
japantollreceipts.jp
August 11, 2026
Page 1 of 4
Customer Layer
Users / Customers
Report requests
Dashboard
Status & history
Scheduled reports
Recurring delivery
Settings
Cadence · recipients
Control Layer — Central System
Queue
Orders all requests
Scheduler
Runs at the right time
Database
Tracks status
Status tracking
Organized ETC usage records and processing history for the customer’s own internal review.
SAFETY GOVERNOR
Central Gate
Only approved, limited work passes through. The rest waits safely.
Worker / Delivery Layer
Private Workers
Not public websites
ETC Usage Inquiry Service (MEISAI)
Controlled access
Report Processing
Validation · PDF / CSV
Delivery
PDF · CSV · email · dashboard
The correct request flow
CustomerCentral SystemQueueGovernorPrivate WorkerETC Usage Inquiry Service (MEISAI)Report ProcessingDelivery
Customers never access the ETC Usage Inquiry Service (MEISAI) directly — everything routes through the central system first.
JapanTollReceipts· Engineering BriefingConfidentialPage 1 of 4japantollreceipts.jp
JapanTollReceiptsSafety Controls & Queue Management2 / 4 · japantollreceipts.jp
Throughput Control
Many Requests
Inbound report requests
Managed Queue
Requests wait in order
Safety Governor
Controls simultaneous activity
Approved Worker Activity
Limited background processing
Controlled External Access
Measured volume only
Remaining requests wait safely in the queue.
The control model, at a glance
✕ The model we avoid
CustomerWorkerETC Usage Inquiry Service (MEISAI)
✓ The JapanTollReceipts model
CustomerCentral SystemQueueGovernorWorkerETC Usage Inquiry Service (MEISAI)Report Processing
Every request passes through the queue and the governor before reaching any external system. This is the core of the JapanTollReceipts trust model.
Controlled Timing
Work does not start just because a customer clicks. Jobs are scheduled centrally so timing can be delayed, spread, or retried safely.
Spread Scheduling
Scheduled reports are distributed across time windows instead of all running at the same time.
Duplicate Prevention
The system checks whether the same customer, card, and period already has queued, active, or recent work.
Prioritization
Urgent, first-time setup, business, government, and support jobs can be ranked ahead of lower-priority work.
Status Tracking
Each report moves through visible states: queued, scheduled, processing, validating, delivered, delayed, retrying, or needs review.
Spike Avoidance
Queueing, scheduling, governor controls, and retry backoff work together so bursts do not become uncontrolled parallel activity.
JapanTollReceipts· Engineering BriefingConfidentialPage 2 of 4japantollreceipts.jp
JapanTollReceiptsM-Level Governor & the M-Process3 / 4 · japantollreceipts.jp

The M-Level Governor decides how much parallel capacity is safe to use. Starting at the M10 baseline, it steps capacity up or down with demand and MEISAI health — never exceeding a hard 50-session ceiling.

M-Level Scaling Ladder
M4
baseline
M6
M8
M10
M12
M20
M30
M40
M48
max
Controlled capacity growth from baseline to high-volume — always under the 50-session ceiling.
The M-Process (control loop)
Report QueueM-Level GovernorWorker PoolMEISAI PullReport DeliveryMetrics & Alerts
Metrics flow back to the governor to inform the next decision. The “Sora” AI advisor analyses trends and recommends adjustments, but never bypasses safety rules — final execution is controlled by the deterministic governor.
Capacity is not “always busy”

“M10” does not mean 10 workers are always busy — it means 10 safe processing slots are ready. When the queue is empty the system stays healthy and idle; when reports are waiting, those slots activate automatically.

How it scales
Baseline: always holds at least M10
Up: steps up two rungs at a time as backlog grows and MEISAI stays healthy
Down: steps gradually back toward baseline as the backlog clears
Ceiling: never exceeds 50 concurrent sessions
Per-Account Locking
Each customer’s MEISAI work stays isolated and sequential, so data never mixes between accounts.
Hard Session Ceiling
Concurrent MEISAI sessions stay capped at 50 regardless of internal capacity.
Emergency Baseline (M10) Control
Admins can instantly return to the M10 baseline; every scaling decision is recorded for audit.
See the live smart-scaling animation and full details at japantollreceipts.jp/architecture
JapanTollReceipts· Engineering BriefingConfidentialPage 3 of 4japantollreceipts.jp
JapanTollReceiptsData Integrity, Scaling & FAQ4 / 4 · japantollreceipts.jp
Data Integrity Flow
1Report Request
2Customer Association
3Card Association
4Report Period
5Validation
6Approved Delivery

Each report workflow is associated with the correct customer, account, card, report period, and delivery setting before delivery. Jobs that cannot be validated are delayed, retried, or marked for review instead of being sent blindly.

Scaling Model
Internal capacity can grow
Worker A
Worker B
Worker C
+
Safety Governor remains the gate
More private workers can increase internal processing capacity, but they do not create unlimited external access. The Safety Governor remains the control point for simultaneous external activity.
Compact Engineering FAQ
Q1. Do customer requests directly access MEISAI?
No. Requests enter the queue and must pass through scheduling and the Safety Governor.
Q2. What prevents too many sessions?
The Safety Governor limits simultaneous background activity.
Q3. Do more workers mean more MEISAI traffic?
Not automatically. More workers add internal capacity, but the governor still controls external activity.
Q4. Why use a queue?
Queueing absorbs bursts, prevents duplicate work, and tracks status.
Q5. Are workers public websites?
No. Workers are private background processors, not customer-facing websites.
Q6. Is JapanTollReceipts official MEISAI or NEXCO?
No. JapanTollReceipts is an independent service.