Customer event control plane

The control plane for customer event truth.

Layarr governs, normalizes, routes, replays, and proves every customer signal before it becomes analytics, ads, automation, or AI context.

Source → Consent → Normalize → Stitch → Destinations → Proof → Replay / Archive
Install once. Govern every event. Prove what happened.
ControlD1 proof records
JobsQueues for delivery/replay
ArchiveR2 archives
MetricsAnalytics Engine metrics
AIAI suggestions audited

A checkout event should leave a trail.

Layarr turns one messy signal into a governed sequence of records: receipt, policy decision, canonical event, deterministic identity edge, destination attempt, replay eligibility, and archive reference.

checkout_completed
Source receives it

Web SDK or server intake records the event before any destination claim exists.

analytics allowed
Consent gates it

Policy allows analytics while advertising stays blocked when marketing consent is missing.

schema v1
Normalize maps it

The event is checked against the canonical contract with warnings preserved.

deterministic only
Stitch links it

Identity edges use declared identifiers, not probabilistic guesswork.

route manifest
Destinations fan out

Adapters receive only routes allowed by policy and config.

proof packet
Proof records it

Receipt, decision, attempt, retry, and failure states stay visible.

preview first
Replay recovers it

Backfill and replay jobs are previewed before operator approval.

The event layer, made operable.

Every capability is designed around visible state and durable records, not invisible promises.

Intake

Event intake and SDK

Capture browser and server events with account-scoped receipt IDs.

Proof detail: event row
Consent

Consent governance

Apply region and purpose gates before destination delivery.

Proof detail: consent receipt
Contract

Normalization

Map raw event names into canonical schemas and keep warnings.

Proof detail: schema warning
Graph

Identity stitching

Build deterministic identity links with consent basis attached.

Proof detail: identity edge
Adapters

Destination routing

Route to analytics, ads, email, webhook, warehouse, and proof lanes.

Proof detail: adapter config
Queues

Delivery jobs

Queue delivery and dead-letter paths without confusing receipt for delivery.

Proof detail: attempt status
Replay

Replay and backfill

Recover failed or changed routes through previewed operator jobs.

Proof detail: replay batch
Archive

R2 proof bundles

Keep large exports, raw archives, and backfill payloads outside D1.

Proof detail: archive object
Metrics

Analytics rollups

Separate high-cardinality metrics from recent control records.

Proof detail: rollup bucket
AI

Guarded assistants

Draft mappings, diagnostics, anomaly notes, and query plans from redacted context.

Proof detail: AI audit row
Ops

Launch readiness

Check D1, queues, R2, analytics sink, runtime, and billing mode in one view.

Proof detail: health signal

Cloudflare-first, with clear storage separation.

Layarr keeps the edge control plane fast while separating proof, jobs, archives, and analytics by responsibility.

Worker

Worker as edge/API/control surface

The Worker receives events, serves the operator app, applies routing decisions, and coordinates delivery/replay queues.

D1

Control and proof

D1 stores control records, proof, recent operational data, and audit rows.

Queues

Delivery work

Queues carry delivery, replay, retry, and backfill work.

R2

Archive plane

R2 holds archives, exports, replay snapshots, and proof bundles.

Analytics

Metric plane

Analytics Engine or an equivalent store handles high-cardinality event metrics.

Live

Future streams

Durable Objects, SSE, or WebSockets can coordinate live operator streams.

Guarded AI assistance

AI can accelerate the operator. It cannot replace proof.

Layarr uses AI where structured product context helps: mapping suggestions, diagnostic explanations, anomaly summaries, read-only query plans, and proof report drafting.

Redaction

AI sees redacted context.

Secrets, auth headers, raw PII, and unsafe payload fields stay out of prompts.

Approval

AI suggestions require operator approval.

Generated mappings or destination changes remain suggestions until an authorized operator applies them.

Truth

AI output never replaces persisted proof.

Events, decisions, deliveries, replay jobs, diagnostics, and audit rows remain canonical.

Audit

AI actions are audited.

Every assistant action records account, actor, redacted summary, model/provider, output summary, and timestamp.

Founder-led pilot access

Bring order to the event layer before it feeds everything else.

Use Layarr to see what fired, why it routed, where it landed, what failed, and what can be safely replayed.