Skip to content
CSA Loom — the Microsoft Fabric experience for Azure tenants where Fabric isn't yet available: lakehouses, warehouses, notebooks, semantic models, Activator rules, Data Agents, across Commercial, GCC, GCC-High, and DoD IL5

CSA Loom expanded scope — v2 backlog from 2026-05-23 walkthrough

Comparative positioning note

This document is written from the perspective of Microsoft Azure, Cloud Scale Analytics, and CSA Loom. Any description of third-party or competing products, services, pricing, or capabilities is derived from publicly available documentation and sources believed accurate at the time of writing, and is provided for general comparison only. We do not claim expertise in, or authority over, any non-Microsoft product or service; the respective vendor's official documentation is the authoritative source for their offerings, which may change over time. Nothing here is intended to disparage any vendor — where a competing product has genuine advantages, we aim to note them honestly. Verify all third-party details against the vendor's current official documentation before making decisions.

Direct capture of the user's expanded vision for CSA Loom, beyond the v1 work completed in this session (25 PRPs, 14 PRs merged, infra live in Azure). These items are tracked in this doc and will be turned into PRPs (PRP-26 through PRP-40+) before any code work begins.

Scope context: v1 (current) shipped infrastructure parity + 8 Console panes mirroring Fabric's surface. v2 below extends Loom into a full data platform with developer portal, marketplace, governance, and AI-driven operations — the items below are individually each PRP-sized.

New v2 capabilities

Data marketplace (PRP-26 candidate)

  • Catalog of internal data products (lakehouse tables, views, semantic models, KQL functions) with descriptions, freshness, sensitivity, ownership
  • Subscription + access-request workflow (admin approves, RLS applied on grant)
  • Usage metrics + lineage per product
  • Mirrors Fabric OneLake Catalog but with cross-DLZ federation

OneLake-equivalent unified namespace (PRP-27)

  • Logical mount of all DLZ lakehouses under one path (loom://)
  • Shortcut creation across DLZs + external sources (S3, GCS, Snowflake, databases) — see PRP-29 below for shortcut builder UI
  • Cross-workspace querying via either Databricks federated catalog or Synapse cross-database query

APIM API builder for data sharing (PRP-28)

  • Loom Console pane that generates REST/GraphQL endpoints from lakehouse tables or semantic measures
  • Auto-publishes spec to APIM (the Premium instance we already deploy)
  • Per-endpoint auth/rate-limit/quota policies driven by RBAC
  • Function App backend auto-generated; OBO identity throughout

Function API management

  • Console pane that lists Azure Functions across all DLZs + Admin Plane
  • Per-function: invocation metrics, app-settings editor, deployment slot promotion, log tail
  • App Insights integration already in place

AI/ML API management (PRP-30)

  • Console pane that wraps AI Foundry deployments behind APIM
  • Per-model: rate limits, content-safety enforcement, audit, cost controls
  • Bring-your-own-model registration (HuggingFace, custom AOAI deployment)

Developer portal + dev tools (PRP-31)

Comprehensive in-Console developer surface: - API playground (Swagger UI / GraphiQL embedded) - Bicep/Terraform template gallery for common Loom extensions - SDK + CLI documentation (Loom CLI, Loom SDK for Python/.NET/TS) - Sample code + tutorials inline - Code generators from data models → DTOs / API clients

Metadata-driven data source onboarding (PRP-32)

  • Self-service workflow in Console
  • Operator points at source (DB, API, file share) → Loom introspects schema → generates Bicep + Debezium connector config + Spark job
  • Auto-creates lineage entries in Purview/Atlas
  • Approval gate before production publish

Domain management (PRP-33)

  • First-class domain construct (Finance, HR, Mission Ops, etc.)
  • Per-domain: Entra group, Power BI workspace, DLZ assignment, data product catalog filter, custom tags
  • Mirrors Fabric Domains in admin portal

New DLZ deployment + setup (already partial)

  • Setup Wizard pane exists for this (PRP-04 shipped)
  • v2 extension: full self-service including pre-validation (subscription quota check, region availability, dependency check)

dbt builder + integration (PRP-34)

  • Loom Console pane for dbt project authoring
  • Visual model designer + raw SQL editor
  • Auto-deploys to Databricks SQL Warehouse or Synapse Serverless
  • Lineage flows into Purview/Atlas

Complete shortcut builder (PRP-35)

  • Like Fabric Lakehouse Shortcuts but expanded:
  • ADLS Gen2 shortcuts (current Fabric supports)
  • S3, GCS, Snowflake, on-prem SMB, REST API, GraphQL
  • Per-shortcut transformation (column rename, type cast, filter)
  • Cached vs live (with TTL)
  • Each shortcut becomes a queryable table in the lakehouse

Data virtualization builder + manager (PRP-36)

  • Cross-database federated query UI
  • Sources: Synapse Serverless EXTERNAL TABLES, Databricks Federated Catalog, PolyBase
  • Console pane for virtual schema design + query workbench

Complete observability — DMLZ + DLZ telemetry (PRP-37)

Already started (telemetry-everywhere standard), v2 extends: - Per-DLZ telemetry dashboard (already in Console catalog as panes) - Cross-DLZ rollup at DMLZ (Data Management Landing Zone) level - DataDog/New Relic export option (in addition to LAW) - Custom KPI definitions + alerting

Power BI reports — management + operations (PRP-38)

Pre-built suite of Power BI reports auto-deployed: - Capacity utilization (per DLZ, per workspace) - Cost attribution (by domain, by tenant) - User adoption funnel - Data freshness SLO compliance - Security events (failed auth, permission changes) - Top queries, top users, top error sources - All wired to the standardized LAW + cost-management APIs

Loom Copilot agent — does-anything (PRP-39)

Already started (apps/copilot + apps/fiab-data-agents), v2 expands: - Embedded in every Console pane (not just Data Agent pane) - Can perform ANY operation a user has permission for: - Create workspace, deploy DLZ, build shortcut, write dbt model, publish API, configure activator rule, refresh semantic model, grant access, etc. - Identity passthrough: agent acts as caller; nothing the user can't do themselves - Conversational workflow + approval gates for destructive ops - Multi-step task execution with progress streaming - Tool catalog dynamically scoped to user's RBAC

Sizing

Honest estimate: 12-15 additional PRPs, each roughly PRP-02 to PRP-09 sized (real Bicep + apps + tests). Probably 6-9 months engineering at the same pace as v1.

Tracking

  • v2 scope locked in this doc, dated 2026-05-23
  • PRP-26 through PRP-40 stubs to be authored in PRPs/active/csa-loom/ before any code begins
  • Each PRP gets a GitHub issue + label csa-loom-v2
  • v2 epic to be opened referencing this doc

Prerequisites before v2 starts

  1. Complete v1 end-to-end:
  2. Apps fully deployed + browser-validated (in flight; deploy iter J running now)
  3. GCC + GCC-High validation (operator action — bootstrap secrets)
  4. Build 2026 freshness rescan (auto-fires Jun 8)
  5. v2 walkthrough with brand + scope decisions analogous to the 2026-05-22 v1 walkthrough