How microsoft/fabric-architecture-review works
Fabric Architecture Review Accelerator (FAR) collects Fabric metadata, configuration, source definitions and metrics—not customer business rows—then analyzes the evidence into prioritized findings and reports. Its main workflow is scoped collection, rule-based review and report generation; Fabric deployments add Gold-backed governance reporting. The README also documents targeted reviews, owner-isolated reports and agents, and an interactive app. Paths for some documented subsystems are directory-level because their implementation was not sampled; the graph omits uncertain internal wiring.
Main entry points: runtime.py Review runtime · reports Written reports · gold_layer.py Governance Gold · analyzers Review analyzers
Components 22
Review workflow
- Review runtime
orchestration/runtime.py - Workspace scope
collectors/workspace_scope.py - Review analyzers
analyzers - Prioritized findings
Evidence collection
- Evidence collectors
collectors - Raw evidence
- Pipeline definitions
collectors/pipeline_definitions.py - Execution history
collectors/pipelines_notebooks.py
Reports and access
- Written reports
reports - Governance Gold
reports/gold_layer.py - Owner report model
reports/owner - Workspace Owner Agent
reports/agent/owner_agent.py - Central Data Agent
reports/agent/data_agent.py
Interactive app
- Review workbench
fabric/app/src/components/review-workbench.tsx - Review data context
fabric/app/src/hooks/review-data.context.tsx - Semantic model queries
fabric/app/src/hooks/use-semantic-model-query.ts
Fabric operations
- Targeted review selection
orchestration/sources.py - Fabric deployment
orchestration/deployment.py
Other components
- Review operator
- Fabric APIs
- Workspace owner
- Fabric administrator
Connections 31
- Review operator → Review runtime: starts review
- Review runtime → Workspace scope: sets scope
- Review runtime → Evidence collectors: runs collection
- Evidence collectors → Fabric APIs: requests metadata
- Evidence collectors → Raw evidence: writes evidence
- Review runtime → Pipeline definitions: collects definitions
- Pipeline definitions → Fabric APIs: fetches definitions
- Review runtime → Execution history: collects history
- Execution history → Fabric APIs: reads job history
- Pipeline definitions → Raw evidence: writes definitions
- Execution history → Raw evidence: writes executions
- Review runtime → Review analyzers: runs analysis
- Review analyzers → Raw evidence: reads evidence
- Review analyzers → Prioritized findings: produces findings
- Review runtime → Written reports: generates reports
- Prioritized findings → Written reports: supports findings
- Fabric deployment → Governance Gold: creates artifacts
- Written reports → Governance Gold: populates review data
- Written reports → Owner report model: supports subset
- Owner report model → Workspace Owner Agent: grounds questions
- Workspace owner → Workspace Owner Agent: asks questions
- Fabric administrator → Targeted review selection: configures targeting
- Targeted review selection → Review runtime: selects workspaces
- Review operator → Review workbench: opens workbench
- Review workbench → Review data context: reads review data
- Review data context → Governance Gold: loads review data
- Review workbench → Semantic model queries: requests model query
- Semantic model queries → Fabric APIs: queries semantic model
- Fabric administrator → Fabric deployment: deploys artifacts
- Governance Gold → Central Data Agent: grounds questions
- Review operator → Central Data Agent: asks questions