Technology Stack

AI Fabrix is engineered to run entirely inside the customer’s Azure tenant with a modular, cloud-native stack. The platform favors open standards (OpenAPI, OpenTelemetry, SCIM, OAuth2/OIDC) and a plugin-first approach to integrations.

Azure Services

• Network & Edge

  • Azure Virtual Network (VNet), Private Endpoints, Private DNS
  • Azure Front Door (global entry), Azure Application Gateway (L7)

• Identity & Secrets

  • Microsoft Entra ID (SSO, RBAC), SCIM provisioning
  • Azure Key Vault (keys, secrets, certificates)

• Compute & Containers

  • Azure App Service / Container Apps (default)
  • Azure Kubernetes Service (AKS) for large-scale or regulated workloads
  • Azure Container Registry (ACR)

• Data & Storage

  • Azure Database for PostgreSQL Flexible Server (pgvector)
  • Azure Cache for Redis
  • Azure Storage (Blob, Files) for artifacts and document stores

• AI & Search

  • Azure OpenAI Service (optional GPT family and embeddings)
  • Azure AI Search (optional hybrid with pgvector)

• Monitoring & Security

  • Azure Monitor, Log Analytics, Application Insights
  • Microsoft Defender for Cloud, Microsoft Sentinel (optional)

AI/ML Technologies

• LLM Access

  • Azure OpenAI or OpenAI (managed foundation models)
  • Optional: model gateways (e.g., vLLM/Ollama) in private clusters where required

• RAG & Reasoning

  • Flowise pipelines for retrieval-augmented generation
  • Metadata-aware retrieval (policy filters, permissions context)

• Embeddings & Vector Ops

  • Azure OpenAI or OpenAI embeddings
  • pgvector in PostgreSQL as primary vector store; optional Azure AI Search vectors

• Safety & Governance

  • Prompt governance patterns, guardrails, audit of model inputs/outputs

Database Technologies

• Relational & Vector

  • PostgreSQL (transactional data, configuration, audit) + pgvector (semantic index)

• Cache & Queues

  • Azure Cache for Redis (sessions, conversation state, short-lived context)
  • Azure Storage Queues / Event Grid or Service Bus (asynchronous orchestration)

• Document Storage

  • Azure Blob Storage (ingested files, embeddings artifacts, exports)

Frontend Technologies

• User Interface

  • OpenWebUI (hardened for enterprise deployment)
  • React-based components, role-aware UX (RBAC via Entra ID groups)

• Design System

  • Tailwind CSS and utility-class theming for admin/operator consoles
  • Accessibility and enterprise theming support

Backend Technologies

• Core & Services

  • Node.js/TypeScript services for Core APIs and plugin runtime
  • Python for selected connectors and data preparation tasks (where appropriate)

• Miso (Control Layer)

  • Controller services for identity, policy, provisioning (Bicep-based orchestration)
  • Policy-as-code for environment promotion (Dev → Test → Prod)

• APIs

  • OpenAPI/JSON Schema contracts; OAuth2/OIDC; SCIM for user/group lifecycle

• Packaging

  • Containerized microservices; IaC-driven release artifacts

Integration Technologies

• Enterprise Connectors

  • Microsoft Graph (SharePoint/OneDrive/Teams/Users/Groups)
  • Dynamics 365/Dataverse, SAP (OData/REST), Salesforce, ServiceNow, Jira
  • CRM/ERP/HR/Finance systems via REST, OData, Webhooks

• Plugin Framework

  • Flowise Enterprise Plugin Framework (server-side execution only)
  • Developer SDK with dynamic fields, dependent pickers, output schemas

• Data Ingestion

  • SharePoint REST/Graph for hierarchical metadata and document sync
  • Webhooks/Event Grid for event-driven flows

Development Tools

• Source & CI/CD

  • GitHub (Repos, Actions), GitHub Container Registry/ACR
  • IaC with Bicep (reference) and optional Terraform (customer choice)

• Build & Local Dev

  • Docker/Compose, Node.js/TypeScript toolchain, Python venv/poetry (where used)
  • Cursor-guided development rules (governed AI coding, review gates)

• Release Management

  • Environment promotion workflows, policy checks, SBOM generation

Monitoring and Observability

• Telemetry

  • OpenTelemetry instrumentation (traces/metrics/logs)
  • Azure Monitor & Application Insights dashboards and alerts

• Security & Compliance

  • Centralized audit logs (immutable retention policies)
  • Defender for Cloud posture checks; Sentinel analytics and playbooks (optional)

• Operational Health

  • Health endpoints, structured logs, workload SLOs/SLAs
  • Cost observability (per-service tagging, chargeback/FinOps readiness)