AIBAI Behavioral Provenance EngineCustomer deployment
CUSTOMER · PRODUCTION · ON REQUEST

Provision the product
only when needed.

The production product is not a shared public runtime. Each customer engagement receives an isolated deployment boundary sized to its storage, security, retention and operational requirements.

IsolatedAuthenticatedDurable storageCustomer-owned boundary

Recommended topology

Runtime → durable evidence platform

The repository's production guide describes a collector/SDK boundary feeding durable storage, with explorer access layered above the recorded evidence.

AI services
↓
SDK / collector
↓
Durable queue
↓
Event store
↓
Explorer API
STORAGE

Evidence fan-out

Object storage for evidence artifacts, relational indexing for queries, and an analytical store can sit behind the same storage contract.

Production controls

What is kept separate from the public showcase.

C1

Data boundary

Redact or tokenize sensitive payloads before persistence; isolate tenants and projects.

C2

Security

Encrypt stored artifacts and authenticate ingestion and read APIs.

C3

Lifecycle

Version schemas and enforce retention/deletion workflows.

C4

Integrity

Hash chains provide tamper evidence; signed manifests are used when evidence crosses trust boundaries.

C5

Replay safety

Exact replay uses recorded outputs. Live model/tool re-execution is explicitly enabled, sandboxed and observable.

C6

Operations

Monitor ingestion latency and event loss, and keep the deployment environment observable.

Request path

No public customer runtime.

Start the engagement from the portfolio. The deployment is then scoped and provisioned separately from the public synthetic explorer.

CUSTOMER DEPLOYMENT

Request an isolated deployment.

Include your workload, expected event volume, data boundary, retention requirements and preferred hosting model.