Products / AI Fabric Framework Open source · Early release

Build intelligent Java applications without surrendering application control.

Connect live application data, evidence-grounded retrieval and governed actions to Spring Boot while your application retains authority over identity, authorization, validation, persistence and execution.

Explore live demos View on GitHub

Java
Spring Boot
Spring AI
Provider-neutral
Apache 2.0

@AICapable(entityType = "product")
class Product { ... }

@AIProcess(operation = AIProcessOperation.UPDATE)
Product save(Product product) { ... }

Application-owned

Identity & Policy Users, tenants, roles, capabilities

Domain Services Validation, business rules, workflows

Application DB Relational DB, source of truth

AI Fabric enablement

Orchestration Coordinate retrieval, actions and policy

Live Evidence Keep approved evidence aligned with app data

Governed Actions Propose, verify and act within policy

Providers

Models LLMs and inference

Embeddings Embedding models

Vector Stores Vector databases and search

Entity changeApp data updated

@AIProcessHook triggered

Evidence updatedRetrieval store aligned

Action proposedIntent generated

Policy checkRules evaluated

ConfirmApproved to proceed

Application serviceExecution in app

The application remains the authority

An enablement layer built around application authority

Annotation-driven integration

Declare searchable entities, context and lifecycle hooks using familiar Java annotations.

Live data synchronization

Keep approved retrieval evidence aligned with application changes.

Governed retrieval & actions

Ground responses in evidence and require policy checks or confirmation before execution.

Provider flexibility

Integrate models and vector stores behind richer framework contracts.

Application data, identity & policy Source of truth
AI Fabric orchestration & evidence flow Enablement layer
Models & vector stores Pluggable providers

Public architecture

Application authority, made inspectable.

AI Fabric embeds inside the Spring Boot application and coordinates only the optional capabilities that the application enables.

Open architecture

AI Fabric architecture showing the Spring Boot application as trusted authority, embedded orchestration, optional capabilities, and pluggable model, embedding, and vector database providers — Application data and authorization stay authoritative; vector state is derived and provider boundaries remain replaceable.

Ordered request pipeline

Every governed request follows an explicit contract.

Protection, context resolution, intent handling, result normalization and recording happen in a stable order, with optional steps activated only when their modules are present.

Open pipeline

Ordered AI Fabric orchestration pipeline from security and access control through intent handling, result normalization, response sanitization, and optional conversation persistence — The order is an extension contract: authority gates fail closed while optional intelligence and persistence stages remain modular.

The problem

Useful AI needs software boundaries, not only prompts.

Useful AI features need more than a model call. They need application context, current data, observable evidence and action boundaries that survive real software lifecycle changes.

The product value

A reusable layer with an explicit host contract.

AI Fabric makes those boundaries explicit and reusable while staying close to Spring Boot and Spring AI. Teams can adopt the capabilities they need without moving their domain into a second application platform.

Ownership boundary

Product responsibility stops before application authority.

Framework owns

Orchestration contracts and pipeline behavior
Retrieval, evidence and structured result contracts
Action proposals, confirmation state and execution hand-off
Provider-neutral configuration and extension points

Host application owns

User identity, sessions and tenant authorization
Business rules, resource ownership and source-of-truth data
Action implementation and final permission checks
Product UI, customer promises and operational policy

Quick start

Begin with one application-owned path.

Start from the public documentation, wire one bounded scenario and keep authorization and source data in the host.

Developer guide

<dependencyManagement>
  <dependency>
    <groupId>io.github.loom-ai-labs</groupId>
    <artifactId>ai-fabric-bom</artifactId>
    <version>0.4.0</version>
    <type>pom</type>
    <scope>import</scope>
  </dependency>
</dependencyManagement>

Live proof