Databricks Marketplace package

AI/BI Readiness Scanner for Databricks

This is the technical review hub for Brickcost: setup docs, security posture, connector contracts, MCP interface notes, legal terms, and support policy for the customer-contained Marketplace scanner.

Deployment Customer-contained scanner package
Access model Read-only telemetry, no production writes
Reviewer path Security, pricing, connectors, and MCP docs

Domain Roles

brickcost.sig.ai is the buyer-facing landing page for the cost-recovery and AI-readiness POC narrative. insights.sig.ai is the Databricks Marketplace information and trust center for evaluators, security reviewers, procurement, and technical onboarding.

Runs in the customer workspace

Scanner-lite notebooks run inside the customer's Databricks workspace and write diagnostic outputs to a customer-owned Unity Catalog schema.

Read-only scanner mode

Scanner-lite is read-only against operational metadata. Enterprise deployments can generate approved remediation actions only after explicit customer authorization.

Evidence-backed findings

Findings include confidence labels and supporting telemetry so platform, audit, and BI teams can validate them before taking action.

What Scanner-Lite Finds

  • Repeated SQL and dashboard intent that suggests missing metric views or data products.
  • Candidate metric conflicts across dashboards, aliases, and SQL expressions.
  • Shadow or sandbox assets used by recurring business workflows.
  • Risky Genie or AI/BI domains where definitions, ownership, or freshness are weak.
  • Genie benchmark questions with domain risk, evidence level, and owner-review guidance.
  • Materialized-view candidates with query-count, dashboard-count, and bytes-read evidence; dashboard dependency hot spots; and experiment-plan text.
  • Serverless attribution-policy gaps, hot-table maintenance candidates, and Lakeflow tier-review prompts.

Generated metric-view DDL and validation SQL, data-product contracts, dollar-denominated savings/payback economics, a 90-day remediation backlog, warehouse sizing, job compute, DBR runtime, Photon, compute-policy, cluster-restart, instance-pool, detailed cost-attribution, and billing-reconciliation detectors belong to the separately deployed private Databricks App or enterprise diagnostic. They are not included in the free scanner-lite notebooks and require the relevant customer telemetry and deployment validation.

Marketplace Readiness Statement

The free scanner-lite package does not send query history, lineage, table names, user activity, or business metadata to Brickcost or any external service. Optional enterprise deployments may support customer-contained Databricks App mode or SaaS-assisted inference only with explicit customer approval.

Review Documents