✓
Swappable AI provider
One code base, three providers, chosen by config: **self-hosted Gemma** (your DGX Spark) · **Claude API** (cloud) · **rule catalog** (no LLM). No vendor lock-in to a single AI model.
AI fault analysis for your charging network — on your own model, inside your own network. The logs never leave the building.
Evaluate in your own data center: Pipelet as a signed container bundle (Helm or single-node appliance), the AI model on an NVIDIA DGX Spark next to it. No outbound internet required — we support the setup, remotely if you wish.
Other charging portals send your OCPP logs and operational data to a third-party cloud for fault analysis. Pipelet does the opposite: the **AI Co-operator** analyses right at the OCPP log and the database — and you decide **which model** does it.
In the independent variant everything runs **entirely inside your network**: charge points → broker/server → Postgres → AI Co-operator → a **self-hosted Gemma model on an NVIDIA DGX Spark**. No cloud vendor, no outbound internet, no data sharing — not even with Pipelet.
Still want the cloud? The same code uses the Claude API. Want no LLM at all? The deterministic rule catalog is enough. One code base, three providers — switchable per install via config.
For inspection bodies, public authorities, bus and truck depots and large fleets that won't negotiate on data sovereignty: that's exactly what Air-Gapped AI is built for.
One code base, three providers, chosen by config: **self-hosted Gemma** (your DGX Spark) · **Claude API** (cloud) · **rule catalog** (no LLM). No vendor lock-in to a single AI model.
OCPP logs, session data and support cases are analysed locally — no outbound internet, no transfer to a model vendor or to Pipelet. Data sovereignty for GDPR and public-sector requirements.
Instead of model training: RAG over your own wallbox and support history (pgvector, no new infra). Every diagnosis **cites** the matching past case — fresh, evidenced, trustworthy. Personalised without training.
Pipelet as a signed container bundle (Helm or single-node appliance), the model on an **NVIDIA DGX Spark** next to it. From an edge-box depot to your own data center — same software, same AI.
Data sovereignty is non-negotiable: no charging or operational data may leave the building. Air-Gapped AI still gives operators AI fault analysis — entirely local, auditable, cloud-free.
The depot must charge 24/7, even when the WAN link drops. Fault analysis keeps running on the on-site DGX Spark — no cloud LLM that's unreachable at night blocks the fix.
Call-center co-pilot: for every incident the agent sees raw data, a cited likely-cause and one-click actions (Reset · TriggerMessage · Session-Stop). Faster first-fix, fewer technician dispatches.
Evaluate in your own data center: Pipelet as a signed container bundle (Helm or single-node appliance), the AI model on an NVIDIA DGX Spark next to it. No outbound internet required — we support the setup, remotely if you wish.
Home charging, charging history, CarData integration — the EV driver UI of the Pipelet suite.
Fleet management, driver management, RFID and billing for commercial EV fleets.
The new operator UI: sites, charger commands, vouchers, OCPI, billing.