Loom AI Labs

live Tenant access

Tenant Guard

A knowledge-portal scenario testing explicit tenant context, retrieval filters and application-owned access boundaries.

Tenant Guard live demo interface
Live application capture · verified 2026-07-25
Status
live
Last verified
2026-07-25
AI Fabric
0.4.0
Data
Synthetic

Question

Hypothesis

Can tenant identity remain an explicit orchestration input instead of being inferred from natural language?

What this demonstrates

  • Tenant identifiers are supplied as application context.
  • Retrieval results remain attributable to the selected synthetic tenant.

What this does not demonstrate

  • A complete identity provider
  • Formal isolation certification

Architecture and data flow

  1. 01 Synthetic source data
  2. 02 Host application policy
  3. 03 AI Fabric capability
  4. 04 Observable UI result

Guided steps

  1. 01

    Choose the first synthetic tenant and run a knowledge query.

  2. 02

    Inspect the returned source ownership.

  3. 03

    Switch tenant context and repeat the query.

Observable evidence

The running application and its health surface are the public evidence for this experiment. The source repository exposes the implementation path behind the visible result.

Security and data boundary

The host authenticates the caller, resolves tenant membership and provides trusted retrieval filters. The model never chooses tenant scope.

Tenant memberships and knowledge records are synthetic.

Known limitations

  • The demo identity switcher is not an authentication system.
  • The dataset is intentionally compact.

Technologies and products used

  • Spring Boot
  • AI Fabric 0.4.0
  • Lucene
  • Tenant filters
  • OpenAI

Related research

The investigation behind this proof.

Context and grounding Implemented prototype

Explicit Application Context Instead of Model Guessing

This study separates trusted application context from model-extracted language. It focuses on tenant identity, allowed vector spaces and resource scope as explicit inputs rather than values inferred from a prompt.

Artifact
Architecture study
Maturity
Implemented prototype
Read investigation
Privacy and identity Implemented prototype

Tenant Identity as Orchestration Context

A focused security investigation into carrying resolved tenant identity as trusted orchestration metadata and applying it again at data and operation boundaries.

Artifact
Security investigation
Maturity
Ongoing evaluation
Read investigation

Connect

Inspect the live system and its source together.

The experiment is useful when the visible result, application boundary and implementation can all be questioned.