Product Requirements Summary¶
This page summarizes the product intent behind the current implementation. It is not a replacement for the full PRD; it gives engineers, operators, and evaluators a shared product lens while reading the docs.
Product problem¶
Data platform teams often assemble infrastructure, governance, access control, and operating procedures through a mixture of scripts, tickets, notebooks, manual UI changes, and provider-specific conventions. That creates drift and makes it hard to answer basic questions:
- What should exist?
- Who approved the change?
- What changed between environments?
- Which resources are governed by platform policy?
- Which resources are manually managed exceptions?
DataMuru addresses this by making data platform intent declarative, reviewable, and executable through a Python-first control layer.
Primary users¶
| User | Job to be done |
|---|---|
| Platform engineer | Define repeatable data platform resources and apply them safely. |
| Data governance lead | Connect taxonomy, RBAC, and masking intent to platform resources. |
| Security reviewer | Understand permissions before they are applied. |
| Data product team | Request or review governed platform changes without learning every provider API. |
| Enterprise operator | Run validated, auditable changes across controlled environments. |
Product principles¶
- Provider-agnostic core: provider-specific APIs belong behind adapters.
- Azure-first implementation: Databricks on Azure is the first live path, but it must not become the product boundary.
- Declarative configuration: YAML describes desired state; commands execute a predictable lifecycle.
- Governance by design: RBAC, classification, taxonomy, and masking are first-class model concerns.
- Safety before convenience: validate, doctor, plan, saved plans, targets, and explicit destroy confirmation are part of the product contract.
- OSS and Enterprise clarity: OSS owns the shared package, CLI, schemas, docs, and core contracts; Enterprise extends those contracts.
Current alpha scope¶
The current alpha focuses on:
- project scaffolding;
- validation and diagnostics;
- local state;
- deterministic planning;
- targeted apply and destroy;
- saved-plan safety checks;
- Databricks catalog and schema live operations;
- Databricks default-storage catalog creation through SQL warehouses;
- RBAC grant compilation for Unity Catalog;
- import discovery, YAML generation, and conservative state adoption;
- basic Enterprise identity lifecycle hooks where account SCIM is available.
Out of scope today¶
These remain roadmap items:
- production cloud state backends;
- full multi-workspace orchestration;
- AWS and GCP live parity;
- ingestion pipeline management;
- modeling workflow management;
- observability integrations;
- compliance report generation;
- transactional rollback for every provider operation;
- hosted control plane.
Evaluation promise¶
A successful alpha evaluation should prove that DataMuru can:
- express platform intent in a readable configuration layout;
- validate that layout before provider calls;
- compare desired and observed state;
- create a small live target safely;
- re-plan idempotently;
- preserve enough structured output for audit and troubleshooting.
Do not evaluate the alpha as if every PRD capability is complete. Evaluate whether the foundation is credible, safe, and extensible.