Loom editor wiring audit — Fabric vs. Loom truth table

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.

Created 2026-05-26 after a Playwright side-by-side against casino-fabric-poc (F64 capacity, limitlessdata tenant). Auto-extracted backend calls from lib/editors/*.tsx via grep -oE "fetch\(...)".

Why this exists

The v3.x UAT harness (apps/fiab-console/e2e/editors.uat.ts) gave 87 A · 26 B · 0 F — but it only tested "page renders without crash" + "backend call returns a documented status". It did not test that:

• Workspace dropdowns are populated from Loom's workspace catalog (Cosmos) or from real Azure-native compute Loom deployed.
• The "Run" button hits the right Azure backend (Synapse Spark Livy / Databricks Jobs / ADX command / etc.) instead of Fabric's tenant API (which doesn't apply when Fabric isn't enabled — the whole reason Loom exists).
• The data shown in pickers represents resources that actually exist in the tenant's Azure footprint.

This doc is the honest gap between "render passes UAT" and "functional UAT works end-to-end with real Azure backing".

Reference workspace

casino-fabric-poc (F64 SKU, limitlessdata tenant) — a real medallion-architecture POC:

casino-fabric-poc/
├── data_pipelines/        Folder — data ingestion pipelines
├── lh_bronze/             Folder — raw landing zone
│   ├── notebooks/         Folder of 5 numbered ingestion notebooks
│   ├── lh_bronze          Lakehouse
│   └── lh_bronze          SQL Analytics Endpoint (auto-paired with Lakehouse)
├── lh_silver/             Folder — cleansed/conformed
├── lh_gold/               Folder — business-ready
├── reports/               Folder — Power BI reports
├── semantic_models/       Folder — Direct Lake / Import models
└── CopyJob_1              Copy job item

Key parity insight: in Fabric, a Lakehouse comes paired with a SQL Analytics Endpoint that exposes the lake's Delta tables for T-SQL. Loom's lakehouse editor + synapse-serverless-sql-pool editor are split — we should auto-pair them so creating a Loom Lakehouse spins up a paired Serverless SQL view set against the same ADLS path.

Fabric's master item taxonomy (the IA Loom should mirror)

From + New item → All categories → Expand all in Fabric:

Category	Purpose
Visualize data	Dashboard, Report, Semantic Model, Real-Time Dashboard, Paginated Report, Graph queryset, Graph model, Event Schema Set, Exploration, Map
Get data	Connectors / shortcuts to external sources
Mirror data	Mirrored Database (Azure SQL, Snowflake, Cosmos, etc.)
Store data	Lakehouse, Warehouse, KQL Database, SQL database
Prepare data	Dataflow Gen2, Copy job, Pipeline
Analyze and train data	Notebook, Spark job definition, ML Model, ML Experiment
Develop data	API for GraphQL, User data functions, Variable library
Track data	Activator (Reflex)
Distribute data	Eventstream, Eventhouse, KQL Queryset
Legacy	Older Power BI items
Others	Misc

Top-level UX patterns Fabric exposes (and Loom should match)

Pattern	Fabric	Loom today	Action
Workspace header	[Experience badge] · [Workspace breadcrumb] · [Capacity icons]	✅ matches	none
Workspace command bar	+ New item · New folder · Import · Migrate · [right-side: Recycle bin · Create deployment pipeline · Create app · Manage access · Workspace settings]	partial — has +New, missing Migrate, Deployment pipelines, App publish	gap
Workspace grid cols	Name · Status · Type · Task · Owner · Refreshed · Next refresh · Endorsement · Sensitivity · Included in app	partial	gap
+ New item dialog	Favorites tab + All items with 11 categories + filter + per-tenant favorites	partial — Loom has 80+ types in a flat list	gap
Folder support inside workspaces	Yes — drag/drop, subfolders	✅ matches	none
Lakehouse ↔ SQL endpoint pairing	Auto-paired	not wired	gap
Per-experience switcher	Power BI / Data Engineering / Data Science / Data Warehouse / Data Factory / Real-Time / Industry Solutions	Loom shows all in left rail	parity-deferred — Loom's flat list is arguably better, but losing the experience filter

Per-editor wiring truth (grep-extracted)

Verdict legend:

• 🟢 LOCAL-AZURE — calls a Loom BFF route that proxies to a real Azure service Loom deployed (Synapse, Databricks, ADF, Cosmos, ADLS, AI Search, ADX/Kusto, Foundry, APIM, Dataverse). End-to-end works.
• 🟡 LOOM-COSMOS-ONLY — calls a Loom BFF that only writes/reads Cosmos metadata; no real Azure execution wired yet. Item creation works, "Run" is a stub.
• 🔴 FABRIC-API — calls /api/fabric/... or per-type Fabric REST proxy (/api/items/notebook/..., etc.). Will succeed only in a tenant where Fabric is enabled — defeats Loom's whole purpose. MUST REWIRE.
• ⚪ MIXED — some calls to Azure backends, some to Fabric.

Editor	Backend call(s)	Verdict	What "Run / Create / Save" should actually do in Loom
notebook	/api/fabric/workspaces, /api/items/notebook/{id}/run?workspaceId=... (proxies Fabric Notebook v1 API)	🔴 FABRIC-API	Workspace dropdown → /api/workspaces (Cosmos). Compute dropdown → Synapse Spark pool list ∪ Databricks cluster list (both from ARM). Run → POST to Synapse Livy /livyApi/versions/2019-11-01-preview/spark/sessions OR Databricks Jobs /api/2.1/jobs/runs/submit.
data-pipeline	/api/fabric/workspaces, /api/items/data-pipeline/{id}/run?workspaceId=... (Fabric DI pipeline)	🔴 FABRIC-API	Use Loom workspace. Backend → Azure Data Factory pipeline runs. Already wired separately for adf-pipeline — data-pipeline should merge into that or be a higher-level wrapper.
dataflow	/api/fabric/workspaces, /api/items/dataflow/{id}/refresh	🔴 FABRIC-API	Dataflow Gen2 is a Fabric-native concept. Loom equivalent = Synapse pipeline with mapping data flow OR ADF mapping data flow. Re-target the Run button.
mirrored-database	/api/fabric/workspaces, /api/items/mirrored-database/{id} (Fabric Mirroring v1)	🔴 FABRIC-API	Loom has its own loom-mirroring-engine container app. Wire to that.
eventhouse	/api/items/eventhouse/{id} → returns Loom ADX cluster URI + databases	🟢 LOCAL-AZURE	Works — verified UAT
kql-database	/api/items/kql-database/{id}/query, /api/items/kql-database/{id}/tables	🟢 LOCAL-AZURE	Works — verified UAT (Sample geo + graph datasets loadable via /api/admin/load-sample-data)
kql-queryset	(via phase3-editors shared workspace picker → /api/fabric/workspaces)	⚪ MIXED	KQL backend is correct (ADX), but workspace picker calls Fabric. Re-target picker.
kql-dashboard	(same as above)	⚪ MIXED	Same fix
eventstream	/api/items/eventstream/{id} → Cosmos state only	🟡 LOOM-COSMOS-ONLY	Save/load works. Actually pushing events is not wired — needs Event Hubs + Kusto ingestion pipeline. v3 deferred this.
activator	(via phase3 picker) /api/items/activator?workspaceId=...	⚪ MIXED	Workspace picker calls Fabric. Backend is the Loom loom-activator-engine container which works. Re-target picker.
warehouse	(via phase3 picker) Fabric Warehouse REST	🔴 FABRIC-API	Loom equivalent = Synapse Dedicated SQL Pool. Already wired separately for synapse-dedicated-sql-pool. Merge or wrap.
semantic-model	(via phase3 picker) Fabric Semantic Model REST	🔴 FABRIC-API	Loom should use Power BI tenant API + the user's Power BI workspace, not Fabric.
report, dashboard, paginated-report, scorecard	(via phase3 picker) Fabric REST	🔴 FABRIC-API	All four are Power BI items. Re-target to Power BI tenant API.
synapse-serverless-sql-pool	/api/items/synapse-serverless-sql-pool/{id}/query	🟢 LOCAL-AZURE	Works — real T-SQL against syn-loom-default-eastus2
synapse-dedicated-sql-pool	/api/items/synapse-dedicated-sql-pool/{id}/query + resume/state	🟢 LOCAL-AZURE	Works — real T-SQL + auto-pause Logic App
synapse-spark-pool	/api/items/synapse-spark-pool/list, /api/items/synapse-spark-pool/{name}	🟢 LOCAL-AZURE	Works — real Spark pool ops
synapse-pipeline	/api/items/synapse-pipeline/{name}/runs + shared DAG view (v3.27)	🟢 LOCAL-AZURE (v3.27)	Works. v3.27: Graph tab added — PipelineDagView renders the activities array as a topologically-ranked DAG with success/failure/completion/skip edge legend. Shared component reused across synapse-pipeline / adf-pipeline / data-pipeline editors.
databricks-cluster	/api/items/databricks-cluster/{id} + events + state	🟢 LOCAL-AZURE	Works — real cluster ops
databricks-job, databricks-notebook	/api/items/databricks-job/{id}/run, /api/items/databricks-notebook/{id}/run	🟢 LOCAL-AZURE	Works — real Jobs API
databricks-sql-warehouse	(covered by databricks-editors)	🟢 LOCAL-AZURE	Works
adf-pipeline, adf-dataset, adf-trigger	/api/items/adf-pipeline/{name}/run, etc. + shared DAG view (v3.27)	🟢 LOCAL-AZURE (v3.27)	Works — real ADF REST. v3.27: Graph tab added (same component as synapse-pipeline).
usql-job	(no live backend — ADLA retired 2024-02-29)	🟡 DEPRECATION-SURFACE (v3.27)	Editor rewired as a deprecation page: error-intent MessageBar explaining ADLA retirement, U-SQL → PySpark heuristic translator (covers EXTRACT/SELECT/GROUP BY/OUTPUT). No more fake AU/cost estimates. See usql-job-parity-spec.md.
apim-api, apim-product, apim-policy	/api/items/apim-*	🟢 LOCAL-AZURE (v3.27)	Works — real APIM REST. v3.27: APIM Policy Operation-scope route added (apis/{aid}/operations/{oid} shape). All four scopes now wired: Global / API / Product / Operation.
data-product	/api/cosmos-items/data-product/{id} (state) + /api/items/apim-product (publish mirror)	🟡 LOOM-COSMOS-ONLY (v3.27 — was F)	Cleared F-vaporware: hardcoded sample state removed; editor is now Cosmos-state-backed with displayName/domain/owner/sla/bundle[] fields. Purview UC publish still pending — MessageBar gate explains this and the "Publish to APIM" action is now correctly labeled as the API consumer surface, not the Purview Data Product registration.
lakehouse	/api/lakehouse/containers, /paths, /preview, /upload; /api/items/synapse-serverless-sql-pool/{id}/query	🟢 LOCAL-AZURE	Works — real ADLS Gen2 browse + Synapse Serverless query
azure-sql-server, azure-sql-database, azure-sql-managed-instance, sql-server-2025-vector-index	/api/items/azure-sql-*/query etc.	🟢 LOCAL-AZURE	Works — real TDS + ARM
cosmos-gremlin-graph	/api/items/cosmos-gremlin-graph/{id}/query	🟢 LOCAL-AZURE (v3.27)	Real Gremlin query works. v3.27: ribbon Edges / Vertices buttons now wired — load quick g.V() / g.E() queries with label + name projection and dispatch immediately. Graph viz still deferred.
cypher-graph	/api/items/kql-database/{id}/query (Cypher routed to ADX make-graph)	🟢 LOCAL-AZURE	Works — verified UAT. No Azure-native Cypher backend exists; KQL substitution is the honest path.
gql-graph	/api/items/gql-graph/{id}/query (501 deferred until LOOM_FABRIC_GRAPH_WORKSPACE bound)	🟡 LOOM-COSMOS-ONLY (v3.27 — was F)	Cleared F-vaporware: Run button now dispatches to a real route. Three backends in picker: persist-only (saves to item state), fabric-graph (returns 501 with documented gate), cosmos-gremlin-translate (best-effort GQL→Gremlin). See gql-graph-parity-spec.md.
vector-store	/api/items/vector-store (Cosmos state)	🟡 LOOM-COSMOS-ONLY (v3.27)	Spec is saved, similarity test not wired. v3.27: cosmos-nosql backend added to picker (DiskANN + VectorEmbeddingPolicy — Microsoft-recommended). Live creation REST still deferred.
ai-foundry-hub, ai-foundry-project, prompt-flow, evaluation, content-safety, tracing, ai-search-index, compute, dataset	/api/items/{type}/{id} → Foundry REST	🟢 LOCAL-AZURE	Works — real Azure ML / Foundry / Content Safety / AI Search
copilot-studio-agent, -knowledge, -topic, -action, -channel, -analytics, copilot-template-library	/api/items/copilot-studio-*?envId=... → Dataverse msdyn_copilots etc.	🟢 LOCAL-AZURE (after v3.19 + AppUser + Copilot Studio per-env enable)	Works when env has Dataverse + Copilot Studio enabled
powerplatform-environment, dataverse-table, power-app, power-page, power-automate-flow, ai-builder-model	/api/items/{type}?envId=...	🟢 LOCAL-AZURE (after Dataverse setup)	Works post-AppUser registration
geo-map, geo-dataset, geo-query, geo-pipeline	(via Synapse Serverless / ADX route + Loom Cosmos state)	⚪ MIXED	Backends correct; map tile rendering is iframe to bing-maps which requires Azure Maps key — not configured (B-grade)
cross-item-copilot	/api/copilot/orchestrate, /api/copilot/sessions, /api/copilot/tools	🟢 LOCAL-AZURE (v3.27)	Works — Loom's own orchestrator + Azure OpenAI. 32 tools wired. v3.27: ribbon buttons now wired — New / Refresh dispatch loom-copilot:session CustomEvents the embedded console listens for, View registry opens raw /api/copilot/tools JSON in a new tab.
ml-model, ml-experiment	/api/items/ml-model/{id}, /api/items/ml-experiment/{id} → Azure ML / Foundry	🟢 LOCAL-AZURE	Works — verified UAT (Foundry-backed)
graphql-api	/api/items/graphql-api/{id}/publish	🟡 LOOM-COSMOS-ONLY	Schema is persisted; actually exposing the GraphQL endpoint needs APIM provisioning. Spec done, runtime not wired.
user-data-function	/api/items/user-data-function/...	🟡 LOOM-COSMOS-ONLY	Body persisted; runtime hosting deferred.
variable-library	/api/items/variable-library/...	🟡 LOOM-COSMOS-ONLY	Cosmos store works; consumers (notebooks/pipelines reading vars) not wired.
ontology	/api/items/ontology/{id} + POST to /api/items/graph-model (v3.27)	⚪ MIXED (v3.27 — was 🟡)	Cosmos persistence works. v3.27: Materialize as graph-model button parses the class hierarchy → creates a graph-model item with one node type per class + IS_A edge type. Downstream graph-model → ADX path then lights up. Lakehouse entity binding + Activator triggers still deferred.
graph-model	/api/items/graph-model/{id} + /materialize to ADX	🟢 LOCAL-AZURE	Real ADX materialize via KQL DDL.
plan	/api/items/plan/{id} (Cosmos)	⚪ MIXED (v3.27 — was 🟡)	Cosmos persistence works. v3.27: progress meter + status badge counts (todo/doing/done/overdue) surface real values. Power Automate approval workflow + semantic-model writeback still deferred.
map	/api/items/map/{id} (Cosmos GeoJSON) + Azure Maps Static API (v3.27)	⚪ MIXED (v3.27 — was 🟡)	Cosmos persistence + GeoJSON validation + bbox autocompute. v3.27: Azure Maps tile preview (Static REST API) renders when NEXT_PUBLIC_LOOM_AZURE_MAPS_KEY is set; bbox-driven zoom + center; falls back to MessageBar gate otherwise. Vector overlay rendering of features still v2.x.
operations-agent, data-agent	/api/items/{type}/{id} (Cosmos)	🟡 LOOM-COSMOS-ONLY	Cosmos config persists. Foundry Agent runtime + per-source typed picker pending — see operations-agent / data-agent parity specs.
data-product-template	/api/items/data-product-template/{slug}/instantiate	🟢 LOCAL-AZURE	Instantiate creates real Loom items from a template — works. B-grade. Gap: Purview governance domain linkage + access policy editor. See data-product-template-parity-spec.md.
data-product-instance	(Cosmos state + per-component refs)	🟢 LOCAL-AZURE (v3.27)	Components table is real. v3.27: Health column added — peeks at each child via /api/cosmos-items/<slug>/<id>, classifies as OK (<7d) / Stale (<30d) / Missing / Unknown. Ribbon Health button now wired to refresh. Access/Quality/Lineage/Activity tabs still pending.
copy-job, dbt-job, spark-job-definition, environment	/api/items/{type}/{id}/run	⚪ MIXED	spark-job-definition → Synapse Spark batch (works); copy-job, dbt-job → Cosmos state + Synapse pipeline run; environment → Cosmos only

Aggregate verdict

	Count	Editors
🟢 LOCAL-AZURE (works)	~45	All Synapse SQL, Databricks, ADF, APIM, AI Foundry, Power Platform (post-setup), Lakehouse, Cosmos Gremlin, Cypher/ADX, Azure SQL
🟡 LOOM-COSMOS-ONLY (UI saves config, runtime not wired)	~8	eventstream, vector-store, graphql-api, user-data-function, variable-library, ontology, plan, operations-agent
🔴 FABRIC-API (wires to wrong backend)	~12	notebook, data-pipeline, dataflow, mirrored-database, warehouse, semantic-model, report, dashboard, paginated-report, scorecard, plus partial breakage in kql-queryset/kql-dashboard/activator workspace pickers
⚪ MIXED	~4	geo-*, kql-queryset, kql-dashboard, activator (correct backend, wrong workspace picker)

Honest summary: ~12 of 85 editors are pointing at Fabric tenant APIs that won't work in Loom's "Fabric-on-Azure-native" model. The user explicitly hit one of these (notebook) when picking a workspace.

Phased remediation plan (proposed)

Phase A (must-fix before "shippable for customers") — 12 editors

For each 🔴, rewire:

1. Workspace dropdown → /api/workspaces (Loom Cosmos)
2. Compute / target dropdown → real ARM-discovered resources in Loom's RGs:
3. notebook: synapse-spark-pool list + databricks-cluster list
4. data-pipeline + dataflow: adf-pipeline list (consolidate, since ADF is the Azure-native equivalent)
5. mirrored-database: loom-mirroring-engine jobs API
6. warehouse: pre-existing synapse-dedicated-sql-pool editor (wrap or redirect)
7. semantic-model + report + dashboard + paginated-report + scorecard: Power BI tenant API + workspace picker → user's actual Power BI workspaces
8. Primary action (Run/Refresh/Publish) → corresponding Azure REST call
9. Add a per-editor UAT spec that:
10. Creates a real Lakehouse / Spark pool / cluster / etc.
11. Opens the editor, asserts the dropdown contains it
12. Clicks Run, asserts the job appears in the real Azure backend's job list

Phase B — 8 editors

Wire the 🟡 LOOM-COSMOS-ONLY ones to real runtimes. Examples: - eventstream: deploy Event Hubs namespace via bicep + wire to Kusto ingestion - graphql-api: provision APIM with the saved schema - vector-store: wire to AI Search vector indexer

Phase C — Quality of life

• Lakehouse ↔ SQL Analytics Endpoint auto-pairing
• + New item dialog: Favorites tab + 11-category grouping matching Fabric's IA
• Workspace grid: Endorsement + Sensitivity columns + "Included in app" column

Phase A — sized estimates (rough)

Editor	Effort	Reason
notebook	M	new compute picker + 2 backend code paths
data-pipeline	S	redirect to existing adf-pipeline wiring
dataflow	M	map Dataflow Gen2 actions to ADF Mapping Data Flow REST
mirrored-database	M	wire to existing loom-mirroring-engine API
warehouse	XS	redirect to synapse-dedicated-sql-pool editor
semantic-model	M	Power BI tenant API + DAX runtime
report, dashboard, paginated-report, scorecard	M each	Power BI REST + embed token
kql-queryset, kql-dashboard, activator	S each	swap workspace picker only

Total: ~3-4 weeks of focused engineering, not a single PR.

What this means for the v3 release

The v3 UAT "113/113 pass" claim is technically true for the rubric we used (render + documented-gate handling) but not sufficient for customer-shippable. Honest release notes should say:

• 45 editors are end-to-end functional against real Azure backings ✅
• 8 editors persist their config to Cosmos but the runtime they describe isn't deployed yet — they should carry a "Preview" badge ⚠️
• 12 editors call Fabric tenant APIs that don't apply outside a Fabric-enabled tenant — remove from the catalog or block creation until Phase A re-wire ❌

Until Phase A lands, the catalog should hide or "Preview"-gate those 12 types.

---

This audit was produced by tracing real fetch() calls in apps/fiab-console/lib/editors/*.tsx and visiting Fabric's casino-fabric-poc workspace via Playwright. Re-run by:

cd apps/fiab-console
for f in lib/editors/*.tsx; do
  echo "=== $(basename $f .tsx) ==="
  grep -oE "fetch\(['\`][^'\`]+['\`]" "$f" | sed -E "s/fetch\(['\`]//" | sort -u
done