Back to Rules

🧠 Claude Rule — Data Integrity, Transactions & Backend Database Design

OfficialPopular
ClaudePython Backend
claudepythondatabasedata-integritytransactionssqlalchemybackendbest-practices

You are a backend data engineer using Claude to build robust, consistent database layers for Python APIs.

🧱 Start with a Stable Data Model

  • Define clear domain boundaries: what belongs where
  • Normalize until clarity appears — denormalize only for performance
  • Use meaningful names that represent business reality

Reference: https://docs.sqlalchemy.org/en/20/core/metadata.html

🔑 Enforce Data Integrity Everywhere

  • Prefer database constraints over middleware checks
  • Validate external data at API edges using strict schemas
  • Let Claude detect & explain mismatches between DB schema and Python models

Reference: https://docs.pydantic.dev/

🔐 Transaction Safety and Atomicity

  • Wrap related changes in ACID transactions
  • Avoid long-lived transactions that block concurrency
  • Ensure compensations exist when distributed commits fail

Reference: https://en.wikipedia.org/wiki/ACID

🧲 Query Efficiency and Indexing

  • Use indexes to support frequent lookups, not every column
  • Benchmark slow queries before guessing root causes
  • Let Claude analyze logs to spot missing or unused indexes

Reference: https://use-the-index-luke.com/

🛰 Data Consistency Across Services

  • Choose async messaging or CDC (Change Data Capture) for cross-service sync
  • Avoid allowing multiple services to own the same entity
  • Validate consistency via periodic reports Claude can audit
  • Distributed data is a design choice, not a side effect

🧮 Schema Evolution & Migrations

  • Version every schema change — no hidden modifications
  • Zero-downtime rollout: forward-compatible first, cleanup later
  • Claude can draft migration plans with impact summaries

Reference: https://alembic.sqlalchemy.org/

🔎 Observability on the Data Path

  • Measure slow reads, blocked sessions, and retry churn
  • Log query plans during debugging sessions only
  • Alert when retry loops hide data correctness problems

♻ Data Governance as Collaboration

  • Track ownership per table/domain — clear accountability
  • Document meaning of each critical field as part of code reviews
  • Claude can maintain shared knowledge through automated summaries

🧠 Core Data Durability Values

  • Systems should never guess data intent
  • Data models evolve — version everything
  • Concurrency must be safe, not lucky
  • Claude helps uncover silent data corruption
  • Reliability beats cleverness every time
View Tool Page