Security — Trellistry

What the visit includes.

Physical inspection — see where the server sits, how it is secured, and who has physical access.
Network architecture walkthrough — understand the data path from your device to the model and back, with no external hops.
Access controls — review who can access what, role-based permissions, and the audit trail.
Retention and deletion defaults — see how content is processed and discarded, what metadata is retained, and for how long.
Logging policy — confirm what is logged (usage metadata) and what is not (content).
Your questions — bring your safeguarding officer, your IT lead, your board member. We answer everything on the day.

The six-point readiness gate.

Before any customer data enters the system, we confirm:

Server is on our own premises, under our physical control.
No content is sent to any external AI service or API.
No content is logged — processing is ephemeral.
Usage metadata only: timestamps, token counts, user identifiers for billing.
Role-based access controls are configured and tested.
Your retention and deletion rules are implemented before go-live.

Where the server sits.

Our inference server is on our own premises in Singapore. Content is processed within our controlled environment — not sent to external AI services. For organisations requiring their own on-premise deployment, we design, install and manage a dedicated server on your premises.

Current infrastructure — verified August 2026.

Full-precision (BF16) inference with a 262,144-token context window, sufficient to read a whole case history or spreadsheet in one pass, plus a separate faster model class for lighter workloads. All inference runs on our own hardware with no external API calls. The specific models change as better ones ship; these details are dated for that reason.

Retention and deletion defaults.

Content: Processed and discarded. Not stored, not used for training, not available for retrieval unless you explicitly configure document retrieval (RAG) for your own materials.

Metadata: Timestamps, token counts, user identifiers. Retained for billing, support and capacity planning. Deletable on request.

Operational data: People records, rosters, registrations and follow-through records in your Trellistry account are exportable at any time and deletable on request.

How to arrange a visit.

Use the enquiry form on our contact page. We will schedule a visit at a time that works for you and whoever you want to bring. There is no charge and no obligation.