Deployment
Run sensitive research where it is allowed to run
Local and hybrid deployment is for labs that need stronger data-control guarantees: keep documents, embeddings, retrieval, and model calls inside approved infrastructure.
Source-grounded answersWorkspace boundariesDeployment choice
Who reviews this page
Labs handling sensitive, unpublished, regulated, or institutionally restricted research data.
How it is used
- Run local models with Ollama for development or vLLM for more production-grade local inference.
- Keep the private data plane inside your environment when documents cannot leave institutional control.
- Use cloud coordination only where your review process allows it, or keep the deployment fully local.
Best fit
- Unpublished work, sensitive collaborations, or data governed by institutional review.
- Labs with access to local GPU capacity or an approved private model platform.
- Security reviewers who need a credible path beyond public-cloud AI defaults.
Controls to review
- Tenant and workspace isolation remain required in local and hybrid modes.
- Answers must remain tied to source evidence, not general model memory.
- Backups, health checks, audit events, and operator review remain part of the deployment posture.
- Local infrastructure reduces data movement; it does not remove the need for access control.
Before uploading sensitive documents, review the known limitations and confirm which AI operating model your lab can approve.
Start with a corpus your lab already trusts
A focused pilot is the cleanest way to evaluate citations, refusal behavior, access boundaries, and deployment fit.