AIPUSULA Enterprise Proof v9

Independent AIPUSULA proposal for Siemens • Synthetic industrial data • Not endorsed by Siemens • No real OT actuation
Ready. API-backed demonstration, not production-certified software.

Why AIPUSULA?

Govern AI-assisted industrial decisions without giving agents unrestricted access to industrial control systems.

Ingest → investigate → cite evidence → enforce approval → verify audit.

Architecture

Simulated OT source → authenticated ingestion API → PostgreSQL-backed events + Redis Streams worker (enterprise mode) → evidence-based decision engine → supervisor approval → tamper-evident audit chain.

Siemens value hypothesis

Reduce fragmented investigations, inconsistent approval practices and missing decision provenance across industrial sites.

Value must be validated during a customer pilot.

Demonstration flow

1. Generate a high-vibration event. 2. Inspect AI evidence. 3. Attempt an unauthorized approval. 4. Approve as supervisor. 5. Verify audit chain. 6. Run a measured synthetic workload.

Industrial event ingestion

Event investigations & evidence

Measured local queue experiment

Synthetic local workload; not a distributed scalability certification.

No workload run yet.

SHA-256 chain + Ed25519 signed anchors

Each record commits to its predecessor and canonicalized event fields, with a signed anchor. Verification detects ordinary tampering, but database administrators could rewrite the entire chain. Local signed anchors detect ordinary database rewriting; independent key custody and WORM storage remain required.

Evidence & AI decision limitations
Evidence-retrieval plus deterministic decision rules using a vibration threshold of 7 mm/s. This is NOT a trained AI model; no calibrated confidence is claimed. Each decision references versioned source documents.
OT / IT integration boundary
Example REST ingestion endpoint: POST /api/events with X-Demo-Token. The endpoint receives synthetic telemetry; it does not connect to Siemens equipment or issue PLC commands.
MQTT and OPC UA synthetic integration adapters are included in the Docker lab; no live Siemens connection is provided.
Synthetic payload; Docker lab also includes MQTT broker and OPC UA simulator.
Enterprise module explorer — API-backed
17 operational views are mapped to backend endpoints. Their modes distinguish measured functionality from simulation.
Select a module to inspect live API data.
Original enterprise showcase
The legacy interface is retained only as a visual reference; its buttons do not call this backend. Its controls use client-side simulation and are not wired to the new backend.
Open original v3 showcase ↗
Enterprise Evidence Center
Demonstrate the operating model and its measurable acceptance gates. These are synthetic local metrics, not Siemens production results.
Multi-site overview
Select Refresh to load.
Acceptance evidence
Select Inspect to load.
Customer pilot decision gates
Identity federation • read-only OT DMZ connectivity • incident provenance • operator approval enforcement • measured p95 and recovery time • independent audit custody • model safety evaluation.
Positioning: AIPUSULA complements industrial operations as a governance and evidence layer. It does not replace PLCs, SCADA, MES or existing Siemens systems.
Technical readiness — verified vs not proven
Persistent investigation notes
Notes are stored in the backend database and written to the signed audit trail. Enter an existing event ID from the Event & approvals screen.