The hardware
A dedicated on-site tonia sized for your team. Quiet enough for an office, compact enough for a rack. We spec and ship the hardware; you host it in your premises or with a Canadian hosting provider.
Coming soon
A custom configuration. One model. One audit trail.
The on-site tonia is installed preconfigured. Compact enough for a desk or a rack. You plug it in, you switch it on, it does the rest.
The model
A large language model ready out of the box. It keeps learning from the answers your team approves.
The audit log
Every approval, deletion, and export is recorded locally on-site. Your Privacy Officer can review what happened.
How a typical day plays out.
Marie-Ève is a paralegal at a Montreal-based law firm that switched to tonia six months ago.
- 8:45
Marie-Ève opens the tonia client on her workstation. The local model answers about the file she indexed yesterday.
- 10:20
She asks about a broad legal precedent. The request goes through Tonia and is sent as-is to the AI provider. The response comes back untouched. The event is logged locally on-site.
- 13:15
The firm's Privacy Officer opens the audit console. They see, in plain text, the 47 queries that went through Tonia this morning, their category, their decision, and the proof kept. None contained personal information.
- 22:00
Overnight, on-site tonia incorporates the 14 answers the partners approved today. Each update is versioned and logged. Tomorrow's model will be slightly better on the firm's domain.
Technical architecture
The solutions tonia uses.
| Component | Role | Status |
|---|---|---|
| Inference runtime | Local inference on dedicated hardware | Open source, audited |
| tonia | Orchestrator: models, policies, journaling | Proprietary tonia software |
| AI model | Base model + organisation-specific tuning | Licensed base model + client-owned adaptations |
| Audit service | Local audit log and export | Proprietary tonia software |