Application enablement layer
AI Fabric Framework
Live application data, evidence-grounded retrieval and governed actions for Java and Spring Boot.
- Live data sync
- Governed RAG
- Controlled actions
Open-source AI engineering
We weave intelligence into real software.
Open-source products, live experiments and applied research for dependable AI-enabled applications.
Products from the lab
User experience layer
AI Fabric Chat UI
A reusable, framework-neutral chat interface for AI Fabric-enabled applications.
- Web Component
- React
- Arabic and RTL
- Four layouts
Built to work together
Runnable engineering proof
See the framework at work.
Each experiment exposes a bounded scenario, its application boundary and observable evidence.
AI Shopping Experience
A shopper explores synthetic products, asks grounded questions and inspects the evidence behind the response.
The health surface reports AI Fabric 0.4.0 and enabled chat, retrieval and data-sync capabilities.
Account Resolver
A user asks for account help, reviews the proposed operation and confirms only after the application exposes validation evidence.
Validation warnings can be shown without silently discarding user-confirmed values.
Live Data Sync
An operator creates, updates and removes synthetic application entities while watching the AI index converge.
Create, update and delete operations have visible indexing state.
Implementation-linked
Research that ends in inspectable software.
Engineering investigations with public implementation artifacts, explicit evidence levels and stated limitations.
Keeping Application Data and AI Evidence Aligned
This investigation treats AI indexing as an application data lifecycle rather than a one-time ingestion task. It links source mutations, queued operations, vector state and retrieval proof so an operator can inspect where alignment failed.
- Artifact
- Reproducible evaluation
- Maturity
- Demo verified
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
A Governed Lifecycle for AI-Proposed Actions
This investigation follows an action from model interpretation through validation, warning, confirmation and application execution. It distinguishes advisory validation from trusted target checks.
- Artifact
- Reference implementation
- Maturity
- Demo verified
Built in the open
Inspect the source. Run the proof. Question the boundary.
Products and experiments link directly to their implementation. Claims stop where public evidence stops.
Visit GitHubConnect
Bring a real application problem.
Discuss framework adoption, UI integration, a reproducible experiment or an engineering research question.