warepact 0.1.3

The dbt of data contracts — define, enforce, and evolve data contracts

WarePact

The dbt of data contracts. Define, enforce, and evolve contracts between data producers and consumers — as code, in CI, with zero vendor lock-in.

pip install warepact[duckdb]

---

Quick start

# Scaffold a contracts/ directory
warepact init

# Check a contract against your warehouse
warepact check orders

# Check all contracts
warepact check --all

# Generate a contract from a live table
warepact generate analytics.core.orders --warehouse snowflake

# Watch contracts continuously
warepact watch --interval 300

---

Contract format

version: 1
name: orders
owner: [e-mail]
warehouse: snowflake
table: analytics.core.orders

schema:
  - column: order_id
    type: integer
    not_null: true
    unique: true
  - column: status
    type: string
    allowed_values: [pending, shipped, delivered, cancelled]
  - column: amount
    type: float

sla:
  freshness_hours: 6
  min_rows: 1000
  max_null_rate: 0.01

alerts:
  - channel: slack
    webhook_url: "${SLACK_WEBHOOK}"
    on: [breach, recovery]

custom_checks:
  - name: no_negative_amounts
    sql: "SELECT COUNT(*) FROM {table} WHERE amount < 0"
    expected: 0

---

Supported warehouses

Warehouse	Install extra
DuckDB	pip install warepact[duckdb]
Snowflake	pip install warepact[snowflake]
BigQuery	pip install warepact[bigquery]
Redshift	pip install warepact[redshift]
Postgres	pip install warepact[postgres]
Databricks	pip install warepact[databricks]

---

Configuration

WarePact reads warehouse credentials from environment variables. No secrets belong in contract YAML files.

Warehouse	Required env vars
snowflake	SNOWFLAKE_ACCOUNT, SNOWFLAKE_USER, SNOWFLAKE_PASSWORD, SNOWFLAKE_WAREHOUSE, SNOWFLAKE_DATABASE, SNOWFLAKE_SCHEMA
bigquery	GOOGLE_APPLICATION_CREDENTIALS, BIGQUERY_PROJECT
redshift	REDSHIFT_HOST, REDSHIFT_PORT, REDSHIFT_DATABASE, REDSHIFT_USER, REDSHIFT_PASSWORD
postgres	PGHOST, PGPORT, PGDATABASE, PGUSER, PGPASSWORD
databricks	DATABRICKS_HOST, DATABRICKS_TOKEN, DATABRICKS_HTTP_PATH
duckdb	DUCKDB_DATABASE (path to .duckdb file, or :memory:)

warepact init automatically detects your warehouse from these env vars and prints the full list of required variables.

Supported alert channels

• Slack — Block Kit webhook
• Email — SMTP
• PagerDuty — Events API v2
• Webhook — Generic JSON POST/PUT
• Teams — Adaptive Cards webhook

---

CLI reference

Command	Description
warepact init	Scaffold contracts/ with an example contract
warepact check <name>	Check one contract
warepact check --all	Check every contract in contracts/
warepact check --all --json	Output results as JSON
warepact generate <table>	Generate a contract YAML from a live table
warepact watch	Continuously re-check on a schedule
warepact diff <name> <fileA> <fileB>	Diff two contract versions
warepact report	Generate an HTML health dashboard
warepact publish <name>	Publish a contract to the local registry
warepact mcp	Start the MCP server for Claude/Cursor integration

---

Python API

from warepact import ContractEngine
from warepact.parsers import YAMLParser

parser = YAMLParser()
contract = parser.parse_file("contracts/orders.contract.yaml")

engine = ContractEngine()
result = engine.check(contract)

print(result.to_human_readable())
# Contract 'orders': PASSED
#   ✓ Schema matches expected definition
#   ✓ Row count 45231 is within bounds [1000, ∞)
#   ✓ Data is fresh (last updated 2.1 hours ago)
#   ✓ Null rates are within acceptable limits

---

MCP integration

Start the MCP server and connect it to Claude or Cursor:

warepact mcp

Available tools: check_contract, list_contracts, explain_breach, get_contract_health, suggest_contract.

---

Writing a custom adapter

from warepact.core.registry import PluginRegistry
from warepact.interfaces.warehouse import WarehouseAdapter

@PluginRegistry.register_warehouse("mydb")
class MyDBAdapter(WarehouseAdapter):
    def connect(self, credentials: dict) -> None: ...
    def get_schema(self, table: str) -> list[dict]: ...
    def get_row_count(self, table: str) -> int: ...
    def get_last_updated(self, table: str): ...
    def run_query(self, sql: str) -> list[dict]: ...
    def get_null_rates(self, table: str, columns: list[str]) -> dict: ...

Drop the file in warepact/plugins/ or install it as a package — PluginRegistry.autodiscover() will pick it up automatically.

---

License

Apache 2.0 — free forever, no monetization.