RAGuardian supports three LLM deployment modes. Choose based on your risk level, compliance requirements and budget — and switch later without effort.
European AI inference provider with zero data retention. Inference runs on EU infrastructure, with a strong focus on Italy, aligning with European data sovereignty requirements.
RAGuardiangpt-oss-120b · gpt-oss-20b · Llama-3.3-70B · qwen3.5-122b · qwen3-coder-next · mistral-small-4-119b · more...
Enterprise controls: Customer prompts and outputs are handled in memory by the inference layer and are not persisted. Regolo does not use customer request content to train or fine-tune shared models. Logging and monitoring are designed to avoid storing full payloads, while operational access is limited to what is necessary for reliability and support under clear contractual terms.
Leading French open-weight AI provider. EU data residency, transparent data usage policies.
mistral-medium · mistral-large · mistral-small · mistral-tiny · codestral
Run LLMs on your server. No data ever leaves your network. Maximum privacy, zero external dependencies.
Example config: Ollama with Qwen 3.5 122B on your server. Local embedding with sentence-transformers. No data ever leaves your network. Maximum compliance.
| Feature | Regolo.ai | Mistral AI | Local Model |
|---|---|---|---|
| Data residency | ✓ European | ✓ European | ✓ Your datacenter |
| Zero data retention | ✓ Yes, guaranteed | ✓ Yes, with conditions* | ✓ Zero transfer |
| No training on data | ✓ Guaranteed | ✓ Except specific cases | ✓ Impossible |
| GDPR compliance | ✓ By design | ✓ Compliant | ✓ Total |
| Operating costs | Pay-per-use API | Pay-per-use API | GPU infrastructure |
| Hardware required | None | None | GPU with 16GB+ |
| Air-gapped mode | ✗ | ✗ | ✓ Yes |
| Cloud embedding | ✓ Qwen3-Embedding | ✗ | ✓ sentence-transformers |
* See Mistral AI Data Usage Policy for full details.
Go to /admin/config, select the provider and model for the instance. This acts as the central policy for all workspaces.
For Regolo: REGOLO_API_KEY. For Mistral: MISTRAL_API_KEY. For local: OPENAI_BASE_URL=localhost:11434. Individual user API keys remain separate.
Chats and connectors keep using each user's personal workspace. No data migration needed.
Our team can help evaluate the best scenario for your use case, compliance requirements and budget.